{"before_sent": "Language models have become a key step to achieve state-of-the-art results in many different Natural Language Processing (NLP) tasks.", "before_sent_with_intent": " Language models have become a key step to achieve state-of-the-art results in many different Natural Language Processing (NLP) tasks.", "after_sent": "Language models have become a key step to achieve state-of-the art results in many different Natural Language Processing (NLP) tasks.", "labels": "fluency", "doc_id": "1912.05372", "revision_depth": 1} {"before_sent": " Different versions of FlauBERT as well as a unified evaluation protocol for the downstream tasks are shared with the research community for further reproducible experiments in French NLP.", "before_sent_with_intent": " Different versions of FlauBERT as well as a unified evaluation protocol for the downstream tasks are shared with the research community for further reproducible experiments in French NLP.", "after_sent": " Different versions of FlauBERT as well as a unified evaluation protocol for the downstream tasks , called FLUE (French Language Understanding Evaluation), are shared to the research community for further reproducible experiments in French NLP.", "labels": "meaning-changed", "doc_id": "1912.05372", "revision_depth": 1} {"before_sent": " This has been widely demonstrated for English using contextualized word representations such as OpenAI GPT (Radford et al., 2018 ), BERT ( Devlin et al., 2019 ), or XLNet ( Yang et al., 2019b).", "before_sent_with_intent": " This has been widely demonstrated for English using contextualized word representations such as OpenAI GPT (Radford et al., 2018 ), BERT ( Devlin et al., 2019 ), or XLNet ( Yang et al., 2019b).", "after_sent": " This has been widely demonstrated for English using contextualized representations (Dai and Le, 2015; Peters et al., 2018 ), BERT ( Devlin et al., 2019 ), or XLNet ( Yang et al., 2019b).", "labels": "meaning-changed", "doc_id": "1912.05372", "revision_depth": 2} {"before_sent": " This has been widely demonstrated for English using contextualized word representations such as OpenAI GPT (Radford et al., 2018 ), BERT ( Devlin et al., 2019 ), or XLNet ( Yang et al., 2019b).", "before_sent_with_intent": " This has been widely demonstrated for English using contextualized word representations such as OpenAI GPT (Radford et al., 2018 ), BERT ( Devlin et al., 2019 ), or XLNet ( Yang et al., 2019b).", "after_sent": " This has been widely demonstrated for English using contextualized word representations such as OpenAI GPT (Radford et al., 2018 ; Howard and Ruder, 2018; Radford et al., 2018; Devlin et al., 2019 ), or XLNet ( Yang et al., 2019b).", "labels": "meaning-changed", "doc_id": "1912.05372", "revision_depth": 2} {"before_sent": " This has been widely demonstrated for English using contextualized word representations such as OpenAI GPT (Radford et al., 2018 ), BERT ( Devlin et al., 2019 ), or XLNet ( Yang et al., 2019b).", "before_sent_with_intent": " This has been widely demonstrated for English using contextualized word representations such as OpenAI GPT (Radford et al., 2018 ), BERT ( Devlin et al., 2019 ), or XLNet ( Yang et al., 2019b).", "after_sent": " This has been widely demonstrated for English using contextualized word representations such as OpenAI GPT (Radford et al., 2018 ), BERT ( Devlin et al., 2019 ; Yang et al., 2019b).", "labels": "clarity", "doc_id": "1912.05372", "revision_depth": 2} {"before_sent": " We apply our French language models to complex NLP tasks ( natural language inference, parsing, word sense disambiguation) and show that most of the time they outperform other pre-training approaches.", "before_sent_with_intent": " We apply our French language models to complex NLP tasks ( natural language inference, parsing, word sense disambiguation) and show that most of the time they outperform other pre-training approaches.", "after_sent": " We apply our French language models to diverse NLP tasks ( natural language inference, parsing, word sense disambiguation) and show that most of the time they outperform other pre-training approaches.", "labels": "clarity", "doc_id": "1912.05372", "revision_depth": 2} {"before_sent": " We apply our French language models to complex NLP tasks ( natural language inference, parsing, word sense disambiguation) and show that most of the time they outperform other pre-training approaches.", "before_sent_with_intent": " We apply our French language models to complex NLP tasks ( natural language inference, parsing, word sense disambiguation) and show that most of the time they outperform other pre-training approaches.", "after_sent": " We apply our French language models to complex NLP tasks ( text classification, paraphrasing, natural language inference, parsing, word sense disambiguation) and show that most of the time they outperform other pre-training approaches.", "labels": "meaning-changed", "doc_id": "1912.05372", "revision_depth": 2} {"before_sent": "An effective method to generate a large number of parallel sentences for training improved neural machine translation (NMT) systems is the use of back-translations of the target-side monolingual data.", "before_sent_with_intent": " An effective method to generate a large number of parallel sentences for training improved neural machine translation (NMT) systems is the use of back-translations of the target-side monolingual data.", "after_sent": "An effective method to generate a large number of parallel sentences for training improved neural machine translation (NMT) systems is the use of the back-translations of the target-side monolingual data.", "labels": "fluency", "doc_id": "1912.10514", "revision_depth": 2} {"before_sent": " The method was not able to utilize the available huge amount of monolingual data because of the inability of models to differentiate between the authentic and synthetic parallel data .", "before_sent_with_intent": " The method was not able to utilize the available huge amount of monolingual data because of the inability of models to differentiate between the authentic and synthetic parallel data .", "after_sent": " The standard back-translation method has been shown to be unable to efficiently utilize the available huge amount of monolingual data because of the inability of models to differentiate between the authentic and synthetic parallel data .", "labels": "meaning-changed", "doc_id": "1912.10514", "revision_depth": 2} {"before_sent": " The method was not able to utilize the available huge amount of monolingual data because of the inability of models to differentiate between the authentic and synthetic parallel data .", "before_sent_with_intent": " The method was not able to utilize the available huge amount of monolingual data because of the inability of models to differentiate between the authentic and synthetic parallel data .", "after_sent": " The method was not able to utilize the available huge amount of existing monolingual data because of the inability of models to differentiate between the authentic and synthetic parallel data .", "labels": "clarity", "doc_id": "1912.10514", "revision_depth": 2} {"before_sent": " The method was not able to utilize the available huge amount of monolingual data because of the inability of models to differentiate between the authentic and synthetic parallel data .", "before_sent_with_intent": " The method was not able to utilize the available huge amount of monolingual data because of the inability of models to differentiate between the authentic and synthetic parallel data .", "after_sent": " The method was not able to utilize the available huge amount of monolingual data because of the inability of translation models to differentiate between the authentic and synthetic parallel data .", "labels": "meaning-changed", "doc_id": "1912.10514", "revision_depth": 2} {"before_sent": " The method was not able to utilize the available huge amount of monolingual data because of the inability of models to differentiate between the authentic and synthetic parallel data .", "before_sent_with_intent": " The method was not able to utilize the available huge amount of monolingual data because of the inability of models to differentiate between the authentic and synthetic parallel data .", "after_sent": " The method was not able to utilize the available huge amount of monolingual data because of the inability of models to differentiate between the authentic and synthetic parallel data during training .", "labels": "clarity", "doc_id": "1912.10514", "revision_depth": 2} {"before_sent": " Tagging, or using gates, has been used to enable translation models to distinguish between synthetic and authentic data, improving standard back-translation and also enabling the use of iterative back-translation on language pairs that under-performed using standard back-translation.", "before_sent_with_intent": " Tagging, or using gates, has been used to enable translation models to distinguish between synthetic and authentic data, improving standard back-translation and also enabling the use of iterative back-translation on language pairs that under-performed using standard back-translation.", "after_sent": " Tagging, or using gates, has been used to enable translation models to distinguish between synthetic and authentic data, improving standard back-translation and also enabling the use of iterative back-translation on language pairs that underperformed using standard back-translation.", "labels": "fluency", "doc_id": "1912.10514", "revision_depth": 2} {"before_sent": " This workpresents pre-training and fine-tuning as a simplified but more effective approach of differentiating between the two data.", "before_sent_with_intent": " This workpresents pre-training and fine-tuning as a simplified but more effective approach of differentiating between the two data.", "after_sent": " In this work, we approach back-translation as a domain adaptation problem, eliminating the need for explicit tagging. In the approach -- pre-training and fine-tuning as a simplified but more effective approach of differentiating between the two data.", "labels": "meaning-changed", "doc_id": "1912.10514", "revision_depth": 2} {"before_sent": " This workpresents pre-training and fine-tuning as a simplified but more effective approach of differentiating between the two data.", "before_sent_with_intent": " This workpresents pre-training and fine-tuning as a simplified but more effective approach of differentiating between the two data.", "after_sent": " This workpresentstag-less back-translation pre-training and fine-tuning as a simplified but more effective approach of differentiating between the two data.", "labels": "meaning-changed", "doc_id": "1912.10514", "revision_depth": 2} {"before_sent": " This workpresents pre-training and fine-tuning as a simplified but more effective approach of differentiating between the two data.", "before_sent_with_intent": " This workpresents pre-training and fine-tuning as a simplified but more effective approach of differentiating between the two data.", "after_sent": " This workpresents -- the synthetic and authentic parallel data are treated as out-of-domain and in-domain data respectively and, through pre-training and fine-tuning as a simplified but more effective approach of differentiating between the two data.", "labels": "meaning-changed", "doc_id": "1912.10514", "revision_depth": 2} {"before_sent": " This workpresents pre-training and fine-tuning as a simplified but more effective approach of differentiating between the two data. The approach - tag-less back-translation - trains the model on the synthetic data and fine-tunes it on the authentic data.", "before_sent_with_intent": " This workpresents pre-training and fine-tuning as a simplified but more effective approach of differentiating between the two data. The approach - tag-less back-translation - trains the model on the synthetic data and fine-tunes it on the authentic data.", "after_sent": " This workpresents pre-training and fine-tuning, the translation model is shown to be able to learn more efficiently from them during training. Experimental results have shown that the approach outperforms the standard and tagged back-translation - trains the model on the synthetic data and fine-tunes it on the authentic data.", "labels": "meaning-changed", "doc_id": "1912.10514", "revision_depth": 2} {"before_sent": " The approach - tag-less back-translation - trains the model on the synthetic data and fine-tunes it on the authentic data. Experiments have shown the approach to outperform the baseline and standard back-translation by 4.0 and 0.7 BLEU respectively on low resource English-Vietnamese NMT. While the need for tagging (noising) the dataset has been removed, the technique outperformed tagged back-translation by 0.4 BLEU.", "before_sent_with_intent": " The approach - tag-less back-translation - trains the model on the synthetic data and fine-tunes it on the authentic data. Experiments have shown the approach to outperform the baseline and standard back-translation by 4.0 and 0.7 BLEU respectively on low resource English-Vietnamese NMT. While the need for tagging (noising) the dataset has been removed, the technique outperformed tagged back-translation by 0.4 BLEU.", "after_sent": " The approach - tag-less back-translation approaches on low resource English-Vietnamese NMT. While the need for tagging (noising) the dataset has been removed, the technique outperformed tagged back-translation by 0.4 BLEU.", "labels": "clarity", "doc_id": "1912.10514", "revision_depth": 2} {"before_sent": " Experiments have shown the approach to outperform the baseline and standard back-translation by 4.0 and 0.7 BLEU respectively on low resource English-Vietnamese NMT. While the need for tagging (noising) the dataset has been removed, the technique outperformed tagged back-translation by 0.4 BLEU. The approach reached the best scores in less training time than the standard and tagged back-translation approaches .", "before_sent_with_intent": " Experiments have shown the approach to outperform the baseline and standard back-translation by 4.0 and 0.7 BLEU respectively on low resource English-Vietnamese NMT. While the need for tagging (noising) the dataset has been removed, the technique outperformed tagged back-translation by 0.4 BLEU. The approach reached the best scores in less training time than the standard and tagged back-translation approaches .", "after_sent": " Experiments have shown the approach to outperform the baseline and standard back-translation by 4.0 and 0.7 BLEU respectively on low resource English-Vietnamese and English-German neural machine translation .", "labels": "clarity", "doc_id": "1912.10514", "revision_depth": 2} {"before_sent": " While deep learning methods have been applied to classification-based approaches, current similarity-based methods only embody static notions of similarity.", "before_sent_with_intent": " While deep learning methods have been applied to classification-based approaches, current similarity-based methods only embody static notions of similarity.", "after_sent": " While deep learning methods have been applied to classification-based approaches, applications to similarity-based methods only embody static notions of similarity.", "labels": "clarity", "doc_id": "1912.10616", "revision_depth": 1} {"before_sent": " While deep learning methods have been applied to classification-based approaches, current similarity-based methods only embody static notions of similarity.", "before_sent_with_intent": " While deep learning methods have been applied to classification-based approaches, current similarity-based methods only embody static notions of similarity.", "after_sent": " While deep learning methods have been applied to classification-based approaches, current similarity-based applications have been limited, and most similarity-based methods only embody static notions of similarity.", "labels": "meaning-changed", "doc_id": "1912.10616", "revision_depth": 1} {"before_sent": " Siamese networks have been used to develop learned notions of similarity in one-shot image tasks, and also for tasks of semantic relatedness in NLP.", "before_sent_with_intent": " Siamese networks have been used to develop learned notions of similarity in one-shot image tasks, and also for tasks of semantic relatedness in NLP.", "after_sent": " Siamese networks have been used to develop learned notions of similarity in one-shot image tasks, and also for tasks of mostly semantic relatedness in NLP.", "labels": "clarity", "doc_id": "1912.10616", "revision_depth": 1} {"before_sent": " We examine their application to the stylistic task of authorship attribution , and show that they can substantially outperform both classification- and existing similarity-based approaches on datasets with large numbers of authors .", "before_sent_with_intent": " We examine their application to the stylistic task of authorship attribution , and show that they can substantially outperform both classification- and existing similarity-based approaches on datasets with large numbers of authors .", "after_sent": " We examine their application to the stylistic task of authorship attribution on datasets with large numbers of authors, looking at multiple energy functions and neural network architectures, and show that they can substantially outperform both classification- and existing similarity-based approaches on datasets with large numbers of authors .", "labels": "meaning-changed", "doc_id": "1912.10616", "revision_depth": 1} {"before_sent": " We examine their application to the stylistic task of authorship attribution , and show that they can substantially outperform both classification- and existing similarity-based approaches on datasets with large numbers of authors .", "before_sent_with_intent": " We examine their application to the stylistic task of authorship attribution , and show that they can substantially outperform both classification- and existing similarity-based approaches on datasets with large numbers of authors .", "after_sent": " We examine their application to the stylistic task of authorship attribution , and show that they can substantially outperform both classification- and existing similarity-based approaches . We also find an unexpected relationship between choice of energy function and number of authors, in terms of performance .", "labels": "meaning-changed", "doc_id": "1912.10616", "revision_depth": 1} {"before_sent": " Classification-based approaches work well for small numbers of candidate authors, but only similarity-based methods are applicable for larger numbers of authors or for authors beyond the training set .", "before_sent_with_intent": " Classification-based approaches work well for small numbers of candidate authors, but only similarity-based methods are applicable for larger numbers of authors or for authors beyond the training set .", "after_sent": " Approaches to tackling it have been conventionally divided into classification-based ones, which work well for small numbers of candidate authors, but only similarity-based methods are applicable for larger numbers of authors or for authors beyond the training set .", "labels": "coherence", "doc_id": "1912.10616", "revision_depth": 2} {"before_sent": " Classification-based approaches work well for small numbers of candidate authors, but only similarity-based methods are applicable for larger numbers of authors or for authors beyond the training set .", "before_sent_with_intent": " Classification-based approaches work well for small numbers of candidate authors, but only similarity-based methods are applicable for larger numbers of authors or for authors beyond the training set .", "after_sent": " Classification-based approaches work well for small numbers of candidate authors, and similarity-based methods are applicable for larger numbers of authors or for authors beyond the training set .", "labels": "coherence", "doc_id": "1912.10616", "revision_depth": 2} {"before_sent": " Classification-based approaches work well for small numbers of candidate authors, but only similarity-based methods are applicable for larger numbers of authors or for authors beyond the training set .", "before_sent_with_intent": " Classification-based approaches work well for small numbers of candidate authors, but only similarity-based methods are applicable for larger numbers of authors or for authors beyond the training set .", "after_sent": " Classification-based approaches work well for small numbers of candidate authors, but only similarity-based methods , which are applicable for larger numbers of authors or for authors beyond the training set .", "labels": "coherence", "doc_id": "1912.10616", "revision_depth": 2} {"before_sent": " Classification-based approaches work well for small numbers of candidate authors, but only similarity-based methods are applicable for larger numbers of authors or for authors beyond the training set . While deep learning methodshave been applied to classification-based approaches, applications to similarity-based applications have been limited, and most similarity-based methods only embody static notions of similarity .", "before_sent_with_intent": " Classification-based approaches work well for small numbers of candidate authors, but only similarity-based methods are applicable for larger numbers of authors or for authors beyond the training set . While deep learning methodshave been applied to classification-based approaches, applications to similarity-based applications have been limited, and most similarity-based methods only embody static notions of similarity .", "after_sent": " Classification-based approaches work well for small numbers of candidate authors, but only similarity-based methods are applicable for larger numbers of authors or for authors beyond the training set ; these existing similarity-based applications have been limited, and most similarity-based methods only embody static notions of similarity .", "labels": "clarity", "doc_id": "1912.10616", "revision_depth": 2} {"before_sent": " While deep learning methodshave been applied to classification-based approaches, applications to similarity-based applications have been limited, and most similarity-based methods only embody static notions of similarity .", "before_sent_with_intent": " While deep learning methodshave been applied to classification-based approaches, applications to similarity-based applications have been limited, and most similarity-based methods only embody static notions of similarity .", "after_sent": " While deep learning methodshave been applied to classification-based approaches, applications to similarity-based methods have only embodied static notions of similarity. Deep learning methods, which blur the boundaries between classification-based and similarity-based approaches, are promising in terms of ability to learn a notion of similarity, but have previously only been used in a conventional small-closed-class classification setup .", "labels": "meaning-changed", "doc_id": "1912.10616", "revision_depth": 2} {"before_sent": " We examine their application to the stylistic task of authorship attribution on datasets with large numbers of authors, looking at multiple energy functions and neural network architectures, and show that they can substantially outperform both classification- and existing similarity-based approaches. We also find an unexpected relationship between choice of energy function and number of authors, in terms of performance .", "before_sent_with_intent": " We examine their application to the stylistic task of authorship attribution on datasets with large numbers of authors, looking at multiple energy functions and neural network architectures, and show that they can substantially outperform both classification- and existing similarity-based approaches. We also find an unexpected relationship between choice of energy function and number of authors, in terms of performance .", "after_sent": " We examine their application to the stylistic task of authorship attribution on datasets with large numbers of authors, looking at multiple energy functions and neural network architectures, and show that they can substantially outperform previous approaches .", "labels": "clarity", "doc_id": "1912.10616", "revision_depth": 2} {"before_sent": " While many algorithms exploit this fact in summary generation, it has a detrimental effect on teaching the model to discriminate and extract important information .", "before_sent_with_intent": " While many algorithms exploit this fact in summary generation, it has a detrimental effect on teaching the model to discriminate and extract important information .", "after_sent": " While many algorithms exploit this fact in summary generation, it has a detrimental effect on teaching the model to discriminate and extract important information in general .", "labels": "clarity", "doc_id": "1912.11602", "revision_depth": 1} {"before_sent": " We propose that the lead bias can be leveraged in a simple and effective way in our favor to pretrain abstractive news summarization models on large-scale unlabeled corpus : predicting the leading sentences using the rest of an article.", "before_sent_with_intent": " We propose that the lead bias can be leveraged in a simple and effective way in our favor to pretrain abstractive news summarization models on large-scale unlabeled corpus : predicting the leading sentences using the rest of an article.", "after_sent": " We propose that the lead bias can be leveraged in our favor in a simple and effective way in our favor to pretrain abstractive news summarization models on large-scale unlabeled corpus : predicting the leading sentences using the rest of an article.", "labels": "meaning-changed", "doc_id": "1912.11602", "revision_depth": 1} {"before_sent": " We propose that the lead bias can be leveraged in a simple and effective way in our favor to pretrain abstractive news summarization models on large-scale unlabeled corpus : predicting the leading sentences using the rest of an article.", "before_sent_with_intent": " We propose that the lead bias can be leveraged in a simple and effective way in our favor to pretrain abstractive news summarization models on large-scale unlabeled corpus : predicting the leading sentences using the rest of an article.", "after_sent": " We propose that the lead bias can be leveraged in a simple and effective way to pre-train abstractive news summarization models on large-scale unlabeled corpus : predicting the leading sentences using the rest of an article.", "labels": "clarity", "doc_id": "1912.11602", "revision_depth": 1} {"before_sent": " We propose that the lead bias can be leveraged in a simple and effective way in our favor to pretrain abstractive news summarization models on large-scale unlabeled corpus : predicting the leading sentences using the rest of an article.", "before_sent_with_intent": " We propose that the lead bias can be leveraged in a simple and effective way in our favor to pretrain abstractive news summarization models on large-scale unlabeled corpus : predicting the leading sentences using the rest of an article.", "after_sent": " We propose that the lead bias can be leveraged in a simple and effective way in our favor to pretrain abstractive news summarization models on large-scale unlabeled news corpora : predicting the leading sentences using the rest of an article.", "labels": "clarity", "doc_id": "1912.11602", "revision_depth": 1} {"before_sent": " Via careful data cleaning and filtering , our transformer-based pretrained model without any finetuning achieves remarkable results over various news summarization tasks.", "before_sent_with_intent": " Via careful data cleaning and filtering , our transformer-based pretrained model without any finetuning achieves remarkable results over various news summarization tasks.", "after_sent": " We collect a massive news corpus and conduct data cleaning and filtering , our transformer-based pretrained model without any finetuning achieves remarkable results over various news summarization tasks.", "labels": "clarity", "doc_id": "1912.11602", "revision_depth": 1} {"before_sent": " Via careful data cleaning and filtering , our transformer-based pretrained model without any finetuning achieves remarkable results over various news summarization tasks. With further finetuning, our model outperforms many competitive baseline models. Human evaluations further show the effectiveness of our method .", "before_sent_with_intent": " Via careful data cleaning and filtering , our transformer-based pretrained model without any finetuning achieves remarkable results over various news summarization tasks. With further finetuning, our model outperforms many competitive baseline models. Human evaluations further show the effectiveness of our method .", "after_sent": " Via careful data cleaning and filtering via statistical analysis. We then apply the proposed self-supervised pre-training to existing generation models BART and T5 for domain adaptation. Via extensive experiments on six benchmark datasets, we show that this approach can dramatically improve the summarization quality and achieve state-of-the-art results for zero-shot news summarization without any fine-tuning. For example, in the DUC2003 dataset, the ROUGE-1 score of BART increases 13.7\\% after the lead-bias pre-training. We deploy the model in Microsoft News and provide public APIs as well as a demo website for multi-lingual news summarization .", "labels": "meaning-changed", "doc_id": "1912.11602", "revision_depth": 1} {"before_sent": "Lead bias is a common phenomenon in news summarization, where early parts of an article often contain the most salient information .", "before_sent_with_intent": " Lead bias is a common phenomenon in news summarization, where early parts of an article often contain the most salient information .", "after_sent": "A typical journalistic convention in news articles is to deliver the most salient information .", "labels": "clarity", "doc_id": "1912.11602", "revision_depth": 2} {"before_sent": "Lead bias is a common phenomenon in news summarization, where early parts of an article often contain the most salient information . While many algorithms exploit this fact in summary generation , it has a detrimental effect on teaching the model to discriminate and extract important information in general.", "before_sent_with_intent": " Lead bias is a common phenomenon in news summarization, where early parts of an article often contain the most salient information . While many algorithms exploit this fact in summary generation , it has a detrimental effect on teaching the model to discriminate and extract important information in general.", "after_sent": "Lead bias is a common phenomenon in news summarization, where early parts of an article often contain the most salient information in the beginning, also known as the lead bias. While this phenomenon can be exploited in generating a summary , it has a detrimental effect on teaching the model to discriminate and extract important information in general.", "labels": "clarity", "doc_id": "1912.11602", "revision_depth": 2} {"before_sent": " While many algorithms exploit this fact in summary generation , it has a detrimental effect on teaching the model to discriminate and extract important information in general.", "before_sent_with_intent": " While many algorithms exploit this fact in summary generation , it has a detrimental effect on teaching the model to discriminate and extract important information in general.", "after_sent": " While many algorithms exploit this fact in summary generation , it has a detrimental effect on teaching a model to discriminate and extract important information in general.", "labels": "clarity", "doc_id": "1912.11602", "revision_depth": 2} {"before_sent": " We propose that the lead bias can be leveraged in our favor in a simple and effective way to pre-train abstractive news summarization models on large-scale unlabeled news corpora: predicting the leading sentences using the rest of an article.", "before_sent_with_intent": " We propose that the lead bias can be leveraged in our favor in a simple and effective way to pre-train abstractive news summarization models on large-scale unlabeled news corpora: predicting the leading sentences using the rest of an article.", "after_sent": " We propose that this lead bias can be leveraged in our favor in a simple and effective way to pre-train abstractive news summarization models on large-scale unlabeled news corpora: predicting the leading sentences using the rest of an article.", "labels": "fluency", "doc_id": "1912.11602", "revision_depth": 2} {"before_sent": " We then apply the proposed self-supervised pre-training to existing generation models BART and T5 for domain adaptation.", "before_sent_with_intent": " We then apply the proposed self-supervised pre-training to existing generation models BART and T5 for domain adaptation.", "after_sent": " We then apply self-supervised pre-training to existing generation models BART and T5 for domain adaptation.", "labels": "clarity", "doc_id": "1912.11602", "revision_depth": 2} {"before_sent": " We then apply the proposed self-supervised pre-training to existing generation models BART and T5 for domain adaptation.", "before_sent_with_intent": " We then apply the proposed self-supervised pre-training to existing generation models BART and T5 for domain adaptation.", "after_sent": " We then apply the proposed self-supervised pre-training on this dataset to existing generation models BART and T5 for domain adaptation.", "labels": "meaning-changed", "doc_id": "1912.11602", "revision_depth": 2} {"before_sent": " Despite the wide spread of pre-training models for NLP applications, they almost focused on text-level manipulation, while neglecting the layout and style information that is vital for document image understanding.", "before_sent_with_intent": " Despite the wide spread of pre-training models for NLP applications, they almost focused on text-level manipulation, while neglecting the layout and style information that is vital for document image understanding.", "after_sent": " Despite the widespread of pre-training models for NLP applications, they almost focused on text-level manipulation, while neglecting the layout and style information that is vital for document image understanding.", "labels": "fluency", "doc_id": "1912.13318", "revision_depth": 1} {"before_sent": " In this paper, we propose textbf LayoutLM to jointly model the interaction between text and layout information across scanned document images, which is beneficial for a great number of real-world document image understanding tasks such as information extraction from scanned documents.", "before_sent_with_intent": " In this paper, we propose textbf LayoutLM to jointly model the interaction between text and layout information across scanned document images, which is beneficial for a great number of real-world document image understanding tasks such as information extraction from scanned documents.", "after_sent": " In this paper, we propose LayoutLM to jointly model the interaction between text and layout information across scanned document images, which is beneficial for a great number of real-world document image understanding tasks such as information extraction from scanned documents.", "labels": "others", "doc_id": "1912.13318", "revision_depth": 1} {"before_sent": " In this paper, we propose textbf LayoutLM to jointly model the interaction between text and layout information across scanned document images, which is beneficial for a great number of real-world document image understanding tasks such as information extraction from scanned documents.", "before_sent_with_intent": " In this paper, we propose textbf LayoutLM to jointly model the interaction between text and layout information across scanned document images, which is beneficial for a great number of real-world document image understanding tasks such as information extraction from scanned documents.", "after_sent": " In this paper, we propose textbf the LayoutLM to jointly model the interaction between text and layout information across scanned document images, which is beneficial for a great number of real-world document image understanding tasks such as information extraction from scanned documents.", "labels": "meaning-changed", "doc_id": "1912.13318", "revision_depth": 1} {"before_sent": " We also leverage the image features to incorporate the style information of words in LayoutLM.", "before_sent_with_intent": " We also leverage the image features to incorporate the style information of words in LayoutLM.", "after_sent": " Furthermore, we also leverage the image features to incorporate the style information of words in LayoutLM.", "labels": "coherence", "doc_id": "1912.13318", "revision_depth": 1} {"before_sent": " We also leverage the image features to incorporate the style information of words in LayoutLM.", "before_sent_with_intent": " We also leverage the image features to incorporate the style information of words in LayoutLM.", "after_sent": " We also leverage the image features to incorporate the visual information of words in LayoutLM.", "labels": "clarity", "doc_id": "1912.13318", "revision_depth": 1} {"before_sent": " We also leverage the image features to incorporate the style information of words in LayoutLM.", "before_sent_with_intent": " We also leverage the image features to incorporate the style information of words in LayoutLM.", "after_sent": " We also leverage the image features to incorporate the style information of words into LayoutLM.", "labels": "fluency", "doc_id": "1912.13318", "revision_depth": 1} {"before_sent": " To the best of our knowledge, this is the first time that text and layout are jointly learned in a single framework for document-level pre-training , leading to significant performance improvement in downstream tasks for document image understanding.", "before_sent_with_intent": " To the best of our knowledge, this is the first time that text and layout are jointly learned in a single framework for document-level pre-training , leading to significant performance improvement in downstream tasks for document image understanding.", "after_sent": " To the best of our knowledge, this is the first time that text and layout are jointly learned in a single framework for document-level pre-training . It achieves new state-of-the-art results in several downstream tasks, including receipt understanding (from 94.02 to 95.24) and document image classification (from 93.07 to 94.42). The code and pre-trained LayoutLM models will be available soon at URL", "labels": "meaning-changed", "doc_id": "1912.13318", "revision_depth": 1} {"before_sent": " It achieves new state-of-the-art results in several downstream tasks, including receipt understanding (from 94.02 to 95.24) and document image classification (from 93.07 to 94.42).", "before_sent_with_intent": " It achieves new state-of-the-art results in several downstream tasks, including receipt understanding (from 94.02 to 95.24) and document image classification (from 93.07 to 94.42).", "after_sent": " It achieves new state-of-the-art results in several downstream tasks, including form understanding (from 70.72 to 79.27), receipt understanding (from 94.02 to 95.24) and document image classification (from 93.07 to 94.42).", "labels": "meaning-changed", "doc_id": "1912.13318", "revision_depth": 2} {"before_sent": " The code and pre-trained LayoutLM models will be available soon at URL", "before_sent_with_intent": " The code and pre-trained LayoutLM models will be available soon at URL", "after_sent": " The code and pre-trained LayoutLM models are publicly available at URL", "labels": "meaning-changed", "doc_id": "1912.13318", "revision_depth": 2} {"before_sent": " Specifically, first, we curate a massive, deduplicated corpus of 6M Python files from GitHub, which we use to pre-train CuBERT, an open-sourced code-understanding BERT model;", "before_sent_with_intent": " Specifically, first, we curate a massive, deduplicated corpus of 6M Python files from GitHub, which we use to pre-train CuBERT, an open-sourced code-understanding BERT model;", "after_sent": " Specifically, first, we curate a massive, deduplicated corpus of 7.4M Python files from GitHub, which we use to pre-train CuBERT, an open-sourced code-understanding BERT model;", "labels": "meaning-changed", "doc_id": "2001.00059", "revision_depth": 2} {"before_sent": " Future work on source-code embedding can benefit from reusing our benchmark, and comparing against CuBERT models as a strong baseline.", "before_sent_with_intent": " Future work on source-code embedding can benefit from reusing our benchmark, and comparing against CuBERT models as a strong baseline.", "after_sent": " Future work on source-code embedding can benefit from reusing our benchmark, and from comparing against CuBERT models as a strong baseline.", "labels": "fluency", "doc_id": "2001.00059", "revision_depth": 2} {"before_sent": " In this paper, we develop variants of layer-wise relevance backpropagation (LRP) and gradient backpropagation, tailored to image captioning with attention. The result provides simultaneously pixel-wise image explanation and linguistic explanation for each word in the captions.", "before_sent_with_intent": " In this paper, we develop variants of layer-wise relevance backpropagation (LRP) and gradient backpropagation, tailored to image captioning with attention. The result provides simultaneously pixel-wise image explanation and linguistic explanation for each word in the captions.", "after_sent": " In this paper, we develop variants of layer-wise relevance backpropagation (LRP) and gradient backpropagation, tailored to image captioning models with attention mechanisms. The explanations provide simultaneously pixel-wise image explanation and linguistic explanation for each word in the captions.", "labels": "clarity", "doc_id": "2001.01037", "revision_depth": 1} {"before_sent": " We show that given a word in the caption to be explained, explanation methods such as LRP reveal supporting and opposing pixels as well as words.", "before_sent_with_intent": " We show that given a word in the caption to be explained, explanation methods such as LRP reveal supporting and opposing pixels as well as words.", "after_sent": " We show that given a word in the caption to be explained, explanation methods such as LRP reveal supporting and opposing pixels as well as preceding words.", "labels": "clarity", "doc_id": "2001.01037", "revision_depth": 1} {"before_sent": " We show that explanation methods, firstly, correlate to object locations with higher precision than attention, secondly, are able to identify object words that are unsupported by image content, and thirdly, provide guidance to debias and improve the model.", "before_sent_with_intent": " We show that explanation methods, firstly, correlate to object locations with higher precision than attention, secondly, are able to identify object words that are unsupported by image content, and thirdly, provide guidance to debias and improve the model.", "after_sent": " We show that explanation methods, firstly, correlate to object locations with higher precision than attention, secondly, are able to identify object words that are unsupported by image content, and thirdly, provide guidance to improve and de-bias the model.", "labels": "clarity", "doc_id": "2001.01037", "revision_depth": 1} {"before_sent": " Results are reported for image captioning using two different attention models trained with Flickr30K and MSCOCO2017 datasets.", "before_sent_with_intent": " Results are reported for image captioning using two different attention models trained with Flickr30K and MSCOCO2017 datasets.", "after_sent": " Results are reported using two different attention models trained with Flickr30K and MSCOCO2017 datasets.", "labels": "clarity", "doc_id": "2001.01037", "revision_depth": 1} {"before_sent": " Results are reported for image captioning using two different attention models trained with Flickr30K and MSCOCO2017 datasets.", "before_sent_with_intent": " Results are reported for image captioning using two different attention models trained with Flickr30K and MSCOCO2017 datasets.", "after_sent": " Results are reported for image captioning using two different image captioning attention models trained with Flickr30K and MSCOCO2017 datasets.", "labels": "clarity", "doc_id": "2001.01037", "revision_depth": 1} {"before_sent": "This paper explains predictions of image captioning models with attention mechanisms beyond visualizing the attention itself.", "before_sent_with_intent": " This paper explains predictions of image captioning models with attention mechanisms beyond visualizing the attention itself.", "after_sent": "This paper interprets the predictions of image captioning models with attention mechanisms beyond visualizing the attention itself.", "labels": "clarity", "doc_id": "2001.01037", "revision_depth": 2} {"before_sent": " In this paper, we develop variants of layer-wise relevance backpropagation (LRP) and gradient backpropagation , tailored to image captioning models with attention mechanisms.", "before_sent_with_intent": " In this paper, we develop variants of layer-wise relevance backpropagation (LRP) and gradient backpropagation , tailored to image captioning models with attention mechanisms.", "after_sent": " In this paper, we develop variants of layer-wise relevance propagation (LRP) and gradient backpropagation , tailored to image captioning models with attention mechanisms.", "labels": "clarity", "doc_id": "2001.01037", "revision_depth": 2} {"before_sent": " In this paper, we develop variants of layer-wise relevance backpropagation (LRP) and gradient backpropagation , tailored to image captioning models with attention mechanisms.", "before_sent_with_intent": " In this paper, we develop variants of layer-wise relevance backpropagation (LRP) and gradient backpropagation , tailored to image captioning models with attention mechanisms.", "after_sent": " In this paper, we develop variants of layer-wise relevance backpropagation (LRP) and gradient-based explanation methods , tailored to image captioning models with attention mechanisms.", "labels": "clarity", "doc_id": "2001.01037", "revision_depth": 2} {"before_sent": " The explanations provide simultaneously pixel-wise image explanation and linguistic explanation for each word in the captions. We show that given a word in the caption to be explained, explanation methods such as LRP reveal supporting and opposing pixels as well as preceding words. We compare the properties of attention heatmaps systematically against those computed with explanation methods such as LRP, Grad-CAM and Guided Grad-CAM.", "before_sent_with_intent": " The explanations provide simultaneously pixel-wise image explanation and linguistic explanation for each word in the captions. We show that given a word in the caption to be explained, explanation methods such as LRP reveal supporting and opposing pixels as well as preceding words. We compare the properties of attention heatmaps systematically against those computed with explanation methods such as LRP, Grad-CAM and Guided Grad-CAM.", "after_sent": " We compare the interpretability of attention heatmaps systematically against those computed with explanation methods such as LRP, Grad-CAM and Guided Grad-CAM.", "labels": "clarity", "doc_id": "2001.01037", "revision_depth": 2} {"before_sent": " We compare the properties of attention heatmaps systematically against those computed with explanation methods such as LRP, Grad-CAM and Guided Grad-CAM.", "before_sent_with_intent": " We compare the properties of attention heatmaps systematically against those computed with explanation methods such as LRP, Grad-CAM and Guided Grad-CAM.", "after_sent": " We compare the properties of attention heatmaps systematically against the explanations computed with explanation methods such as LRP, Grad-CAM and Guided Grad-CAM.", "labels": "clarity", "doc_id": "2001.01037", "revision_depth": 2} {"before_sent": " We compare the properties of attention heatmaps systematically against those computed with explanation methods such as LRP, Grad-CAM and Guided Grad-CAM.", "before_sent_with_intent": " We compare the properties of attention heatmaps systematically against those computed with explanation methods such as LRP, Grad-CAM and Guided Grad-CAM.", "after_sent": " We compare the properties of attention heatmaps systematically against those computed with explanation methods such as LRP, Grad-CAM , and Guided Grad-CAM.", "labels": "fluency", "doc_id": "2001.01037", "revision_depth": 2} {"before_sent": " We show that explanation methods , firstly, correlate to object locations with higher precisionthan attention, secondly, are able to identify object wordsthat are unsupported by image content, and thirdly, provide guidance to improve and de-bias the model.", "before_sent_with_intent": " We show that explanation methods , firstly, correlate to object locations with higher precisionthan attention, secondly, are able to identify object wordsthat are unsupported by image content, and thirdly, provide guidance to improve and de-bias the model.", "after_sent": " We show that explanation methods provide simultaneously pixel-wise image explanation (supporting and opposing pixels of the input image) and linguistic explanation (supporting and opposing words of the preceding sequence) for each word in the predicted captions. We demonstrate with extensive experiments that explanation methods can 1) reveal more related evidence used by the model to make decisions than attention; 2) correlate to object locations with higher precisionthan attention, secondly, are able to identify object wordsthat are unsupported by image content, and thirdly, provide guidance to improve and de-bias the model.", "labels": "meaning-changed", "doc_id": "2001.01037", "revision_depth": 2} {"before_sent": " We show that explanation methods , firstly, correlate to object locations with higher precisionthan attention, secondly, are able to identify object wordsthat are unsupported by image content, and thirdly, provide guidance to improve and de-bias the model. Results are reported using two different image captioning attention models trained with Flickr30K and MSCOCO2017 datasets. Experimental analyses show the strength of explanation methods for understanding image captioning attention models .", "before_sent_with_intent": " We show that explanation methods , firstly, correlate to object locations with higher precisionthan attention, secondly, are able to identify object wordsthat are unsupported by image content, and thirdly, provide guidance to improve and de-bias the model. Results are reported using two different image captioning attention models trained with Flickr30K and MSCOCO2017 datasets. Experimental analyses show the strength of explanation methods for understanding image captioning attention models .", "after_sent": " We show that explanation methods , firstly, correlate to object locations with high precision; 3) is helpful to `debug' the model such as analyzing the reasons for hallucinated object words. With the observed properties of explanations, we further design an LRP-inference fine-tuning strategy that can alleviate the object hallucination of image captioning models, meanwhile, maintain the sentence fluency. We conduct experiments with two widely used attention mechanisms: the adaptive attention mechanism calculated with the additive attention and the multi-head attention calculated with the scaled dot product .", "labels": "meaning-changed", "doc_id": "2001.01037", "revision_depth": 2} {"before_sent": " Experimental results show ProphetNet achieves the best performance on both abstractive summarization and question generation tasks compared to the models using the same base scale pre-training dataset.", "before_sent_with_intent": " Experimental results show ProphetNet achieves the best performance on both abstractive summarization and question generation tasks compared to the models using the same base scale pre-training dataset.", "after_sent": " Then we conduct experiments on CNN/DailyMail, Gigaword, and SQuAD 1.1 benchmarks for abstractive summarization and question generation tasks compared to the models using the same base scale pre-training dataset.", "labels": "meaning-changed", "doc_id": "2001.04063", "revision_depth": 1} {"before_sent": " Experimental results show ProphetNet achieves the best performance on both abstractive summarization and question generation tasks compared to the models using the same base scale pre-training dataset. For the large scale dataset pre-training, ProphetNet achieves new state-of-the-art results on Gigaword and comparable results on CNN/DailyMail using only about 1/5 pre-training epochs of the previous model .", "before_sent_with_intent": " Experimental results show ProphetNet achieves the best performance on both abstractive summarization and question generation tasks compared to the models using the same base scale pre-training dataset. For the large scale dataset pre-training, ProphetNet achieves new state-of-the-art results on Gigaword and comparable results on CNN/DailyMail using only about 1/5 pre-training epochs of the previous model .", "after_sent": " Experimental results show ProphetNet achieves the best performance on both abstractive summarization and question generation tasks . Experimental results show that ProphetNet achieves new state-of-the-art results on Gigaword and comparable results on CNN/DailyMail using only about 1/5 pre-training epochs of the previous model .", "labels": "clarity", "doc_id": "2001.04063", "revision_depth": 1} {"before_sent": " For the large scale dataset pre-training, ProphetNet achieves new state-of-the-art results on Gigaword and comparable results on CNN/DailyMail using only about 1/5 pre-training epochs of the previous model .", "before_sent_with_intent": " For the large scale dataset pre-training, ProphetNet achieves new state-of-the-art results on Gigaword and comparable results on CNN/DailyMail using only about 1/5 pre-training epochs of the previous model .", "after_sent": " For the large scale dataset pre-training, ProphetNet achieves new state-of-the-art results on all these datasets compared to the models using the same scale pre-training epochs of the previous model .", "labels": "clarity", "doc_id": "2001.04063", "revision_depth": 1} {"before_sent": " For the large scale dataset pre-training, ProphetNet achieves new state-of-the-art results on Gigaword and comparable results on CNN/DailyMail using only about 1/5 pre-training epochs of the previous model .", "before_sent_with_intent": " For the large scale dataset pre-training, ProphetNet achieves new state-of-the-art results on Gigaword and comparable results on CNN/DailyMail using only about 1/5 pre-training epochs of the previous model .", "after_sent": " For the large scale dataset pre-training, ProphetNet achieves new state-of-the-art results on Gigaword and comparable results on CNN/DailyMail using only about 1/5 pre-training corpus .", "labels": "clarity", "doc_id": "2001.04063", "revision_depth": 1} {"before_sent": " As pictographs, Chinese characters contain latent glyph information , which is often overlooked.", "before_sent_with_intent": " As pictographs, Chinese characters contain latent glyph information , which is often overlooked.", "after_sent": " As pictographs, Chinese characters contain latent glyph infor-mation , which is often overlooked.", "labels": "fluency", "doc_id": "2001.05272", "revision_depth": 1} {"before_sent": " We propose the FGN , Fusion Glyph Network for Chinese NER.", "before_sent_with_intent": " We propose the FGN , Fusion Glyph Network for Chinese NER.", "after_sent": " In this paper, we propose the FGN , Fusion Glyph Network for Chinese NER.", "labels": "clarity", "doc_id": "2001.05272", "revision_depth": 1} {"before_sent": " This method may offer glyph informationfor fusion representation learning with BERT . The major innovations of FGN include:", "before_sent_with_intent": " This method may offer glyph informationfor fusion representation learning with BERT . The major innovations of FGN include:", "after_sent": " Except for adding glyph information, this method may also add extra interactive infor-mation with the fusion mechanism . The major innovations of FGN include:", "labels": "clarity", "doc_id": "2001.05272", "revision_depth": 1} {"before_sent": " This method may offer glyph informationfor fusion representation learning with BERT . The major innovations of FGN include:", "before_sent_with_intent": " This method may offer glyph informationfor fusion representation learning with BERT . The major innovations of FGN include:", "after_sent": " This method may offer glyph informationfor fusion representation learning with BERT . The major in-novations of FGN include:", "labels": "fluency", "doc_id": "2001.05272", "revision_depth": 1} {"before_sent": " (1) a novel CNN structure called CGS-CNN is proposed to capture glyph information from both character graphs and their neighboring graphs . (2) we provide a method with sliding window and Slice-Attention to extract interactive information between BERT representation and glyph representation . Experiments are conducted on four NER datasets, showing that FGN with LSTM-CRF as tagger achieves new state-of-the-arts performance for Chinese NER.", "before_sent_with_intent": " (1) a novel CNN structure called CGS-CNN is proposed to capture glyph information from both character graphs and their neighboring graphs . (2) we provide a method with sliding window and Slice-Attention to extract interactive information between BERT representation and glyph representation . Experiments are conducted on four NER datasets, showing that FGN with LSTM-CRF as tagger achieves new state-of-the-arts performance for Chinese NER.", "after_sent": " (1) a novel CNN struc-ture called CGS-CNN is proposed to capture glyph information from both character graphs and their neighboring graphs . (2) we provide a method with sliding window and Slice-Attention to extract interactive information between BERT representation and glyph representation . Experiments are conducted on four NER datasets, showing that FGN with LSTM-CRF as tagger achieves new state-of-the-arts performance for Chinese NER.", "labels": "others", "doc_id": "2001.05272", "revision_depth": 1} {"before_sent": " (1) a novel CNN structure called CGS-CNN is proposed to capture glyph information from both character graphs and their neighboring graphs . (2) we provide a method with sliding window and Slice-Attention to extract interactive information between BERT representation and glyph representation . Experiments are conducted on four NER datasets, showing that FGN with LSTM-CRF as tagger achieves new state-of-the-arts performance for Chinese NER.", "before_sent_with_intent": " (1) a novel CNN structure called CGS-CNN is proposed to capture glyph information from both character graphs and their neighboring graphs . (2) we provide a method with sliding window and Slice-Attention to extract interactive information between BERT representation and glyph representation . Experiments are conducted on four NER datasets, showing that FGN with LSTM-CRF as tagger achieves new state-of-the-arts performance for Chinese NER.", "after_sent": " (1) a novel CNN structure called CGS-CNN is proposed to capture both glyph information and interactive information between glyphs from neighboring characters . (2) we provide a method with sliding window and Slice-Attention to extract interactive information between BERT representation and glyph representation . Experiments are conducted on four NER datasets, showing that FGN with LSTM-CRF as tagger achieves new state-of-the-arts performance for Chinese NER.", "labels": "clarity", "doc_id": "2001.05272", "revision_depth": 1} {"before_sent": " (1) a novel CNN structure called CGS-CNN is proposed to capture glyph information from both character graphs and their neighboring graphs . (2) we provide a method with sliding window and Slice-Attention to extract interactive information between BERT representation and glyph representation . Experiments are conducted on four NER datasets, showing that FGN with LSTM-CRF as tagger achieves new state-of-the-arts performance for Chinese NER.", "before_sent_with_intent": " (1) a novel CNN structure called CGS-CNN is proposed to capture glyph information from both character graphs and their neighboring graphs . (2) we provide a method with sliding window and Slice-Attention to extract interactive information between BERT representation and glyph representation . Experiments are conducted on four NER datasets, showing that FGN with LSTM-CRF as tagger achieves new state-of-the-arts performance for Chinese NER.", "after_sent": " (1) a novel CNN structure called CGS-CNN is proposed to capture glyph information from both character graphs and their neighboring graphs . (2) we provide a method with sliding window and Slice-Attention to fuse the BERT representation and glyph representation . Experiments are conducted on four NER datasets, showing that FGN with LSTM-CRF as tagger achieves new state-of-the-arts performance for Chinese NER.", "labels": "clarity", "doc_id": "2001.05272", "revision_depth": 1} {"before_sent": " (1) a novel CNN structure called CGS-CNN is proposed to capture glyph information from both character graphs and their neighboring graphs . (2) we provide a method with sliding window and Slice-Attention to extract interactive information between BERT representation and glyph representation . Experiments are conducted on four NER datasets, showing that FGN with LSTM-CRF as tagger achieves new state-of-the-arts performance for Chinese NER.", "before_sent_with_intent": " (1) a novel CNN structure called CGS-CNN is proposed to capture glyph information from both character graphs and their neighboring graphs . (2) we provide a method with sliding window and Slice-Attention to extract interactive information between BERT representation and glyph representation . Experiments are conducted on four NER datasets, showing that FGN with LSTM-CRF as tagger achieves new state-of-the-arts performance for Chinese NER.", "after_sent": " (1) a novel CNN structure called CGS-CNN is proposed to capture glyph information from both character graphs and their neighboring graphs . (2) we provide a method with sliding window and Slice-Attention to extract interactive information between BERT representation and glyph representation for a character, which may capture potential interactive knowledge be-tween context and glyph . Experiments are conducted on four NER datasets, showing that FGN with LSTM-CRF as tagger achieves new state-of-the-arts performance for Chinese NER.", "labels": "meaning-changed", "doc_id": "2001.05272", "revision_depth": 1} {"before_sent": " (1) a novel CNN structure called CGS-CNN is proposed to capture glyph information from both character graphs and their neighboring graphs . (2) we provide a method with sliding window and Slice-Attention to extract interactive information between BERT representation and glyph representation . Experiments are conducted on four NER datasets, showing that FGN with LSTM-CRF as tagger achieves new state-of-the-arts performance for Chinese NER.", "before_sent_with_intent": " (1) a novel CNN structure called CGS-CNN is proposed to capture glyph information from both character graphs and their neighboring graphs . (2) we provide a method with sliding window and Slice-Attention to extract interactive information between BERT representation and glyph representation . Experiments are conducted on four NER datasets, showing that FGN with LSTM-CRF as tagger achieves new state-of-the-arts performance for Chinese NER.", "after_sent": " (1) a novel CNN structure called CGS-CNN is proposed to capture glyph information from both character graphs and their neighboring graphs . (2) we provide a method with sliding window and Slice-Attention to extract interactive information between BERT representation and glyph representation . Experiments are con-ducted on four NER datasets, showing that FGN with LSTM-CRF as tagger achieves new state-of-the-arts performance for Chinese NER.", "labels": "fluency", "doc_id": "2001.05272", "revision_depth": 1} {"before_sent": " Further, more experiments are conducted to investigate the influences of various components and settings in FGN.", "before_sent_with_intent": " Further, more experiments are conducted to investigate the influences of various components and settings in FGN.", "after_sent": " Further, more experiments are conducted to inves-tigate the influences of various components and settings in FGN.", "labels": "others", "doc_id": "2001.05272", "revision_depth": 1} {"before_sent": " As pictographs, Chinese characters contain latent glyph infor-mation , which is often overlooked.", "before_sent_with_intent": " As pictographs, Chinese characters contain latent glyph infor-mation , which is often overlooked.", "after_sent": " As pictographs, Chinese characters contain latent glyph information , which is often overlooked.", "labels": "fluency", "doc_id": "2001.05272", "revision_depth": 2} {"before_sent": " Except for adding glyph information, this method may also add extra interactive infor-mation with the fusion mechanism.", "before_sent_with_intent": " Except for adding glyph information, this method may also add extra interactive infor-mation with the fusion mechanism.", "after_sent": " Except for adding glyph information, this method may also add extra interactive information with the fusion mechanism.", "labels": "fluency", "doc_id": "2001.05272", "revision_depth": 2} {"before_sent": " The major in-novations of FGN include: (1) a novel CNN struc-ture called CGS-CNN is proposed to capture both glyph information and interactive information between glyphs from neighboring characters.", "before_sent_with_intent": " The major in-novations of FGN include: (1) a novel CNN struc-ture called CGS-CNN is proposed to capture both glyph information and interactive information between glyphs from neighboring characters.", "after_sent": " The major innovations of FGN include: (1) a novel CNN struc-ture called CGS-CNN is proposed to capture both glyph information and interactive information between glyphs from neighboring characters.", "labels": "fluency", "doc_id": "2001.05272", "revision_depth": 2} {"before_sent": " The major in-novations of FGN include: (1) a novel CNN struc-ture called CGS-CNN is proposed to capture both glyph information and interactive information between glyphs from neighboring characters.", "before_sent_with_intent": " The major in-novations of FGN include: (1) a novel CNN struc-ture called CGS-CNN is proposed to capture both glyph information and interactive information between glyphs from neighboring characters.", "after_sent": " The major in-novations of FGN include: (1) a novel CNN structure called CGS-CNN is proposed to capture both glyph information and interactive information between glyphs from neighboring characters.", "labels": "fluency", "doc_id": "2001.05272", "revision_depth": 2} {"before_sent": " (2) we provide a method with sliding window and Slice-Attention to fuse the BERT representation and glyph representation for a character, which may capture potential interactive knowledge be-tween context and glyph.", "before_sent_with_intent": " (2) we provide a method with sliding window and Slice-Attention to fuse the BERT representation and glyph representation for a character, which may capture potential interactive knowledge be-tween context and glyph.", "after_sent": " (2) we provide a method with sliding window and Slice-Attention to fuse the BERT representation and glyph representation for a character, which may capture potential interactive knowledge between context and glyph.", "labels": "fluency", "doc_id": "2001.05272", "revision_depth": 2} {"before_sent": " Experiments are con-ducted on four NER datasets, showing that FGN with LSTM-CRF as tagger achieves new state-of-the-arts performance for Chinese NER.", "before_sent_with_intent": " Experiments are con-ducted on four NER datasets, showing that FGN with LSTM-CRF as tagger achieves new state-of-the-arts performance for Chinese NER.", "after_sent": " Experiments are conducted on four NER datasets, showing that FGN with LSTM-CRF as tagger achieves new state-of-the-arts performance for Chinese NER.", "labels": "fluency", "doc_id": "2001.05272", "revision_depth": 2} {"before_sent": " Further, more experiments are conducted to inves-tigate the influences of various components and settings in FGN.", "before_sent_with_intent": " Further, more experiments are conducted to inves-tigate the influences of various components and settings in FGN.", "after_sent": " Further, more experiments are conducted to investigate the influences of various components and settings in FGN.", "labels": "fluency", "doc_id": "2001.05272", "revision_depth": 2} {"before_sent": "Although over 95 million people worldwide speak the Vietnamese language , there are not many research studies on Vietnamese machine reading comprehension (MRC), the task of understanding a text and answering questions about it.", "before_sent_with_intent": " Although over 95 million people worldwide speak the Vietnamese language , there are not many research studies on Vietnamese machine reading comprehension (MRC), the task of understanding a text and answering questions about it.", "after_sent": "Although Vietnamese is the 17th most popular native-speaker language in the world , there are not many research studies on Vietnamese machine reading comprehension (MRC), the task of understanding a text and answering questions about it.", "labels": "meaning-changed", "doc_id": "2001.05687", "revision_depth": 3} {"before_sent": " In this work, we construct a dataset which consists of 417 Vietnamese texts and 2,783 pairs of multiple-choice questions and answers .", "before_sent_with_intent": " In this work, we construct a dataset which consists of 417 Vietnamese texts and 2,783 pairs of multiple-choice questions and answers .", "after_sent": " In this work, we construct a dataset which consists of 2,783 pairs of multiple-choice questions and answers .", "labels": "clarity", "doc_id": "2001.05687", "revision_depth": 3} {"before_sent": " In this work, we construct a dataset which consists of 417 Vietnamese texts and 2,783 pairs of multiple-choice questions and answers . The texts are commonly used for teaching reading comprehension for elementary school pupils.", "before_sent_with_intent": " In this work, we construct a dataset which consists of 417 Vietnamese texts and 2,783 pairs of multiple-choice questions and answers . The texts are commonly used for teaching reading comprehension for elementary school pupils.", "after_sent": " In this work, we construct a dataset which consists of 417 Vietnamese texts and 2,783 pairs of multiple-choice questions and answers based on 417 Vietnamese texts which are commonly used for teaching reading comprehension for elementary school pupils.", "labels": "clarity", "doc_id": "2001.05687", "revision_depth": 3} {"before_sent": " In addition, we propose a lexical-based MRC technique that utilizes semantic similarity measures and external knowledge sources to analyze questions and extract answers from the given text.", "before_sent_with_intent": " In addition, we propose a lexical-based MRC technique that utilizes semantic similarity measures and external knowledge sources to analyze questions and extract answers from the given text.", "after_sent": " In addition, we propose a lexical-based MRC method that utilizes semantic similarity measures and external knowledge sources to analyze questions and extract answers from the given text.", "labels": "clarity", "doc_id": "2001.05687", "revision_depth": 3} {"before_sent": " We compare the performance of the proposed model with several lexical-based and neural network-based baseline models.", "before_sent_with_intent": " We compare the performance of the proposed model with several lexical-based and neural network-based baseline models.", "after_sent": " We compare the performance of the proposed model with several baseline lexical-based and neural network-based baseline models.", "labels": "meaning-changed", "doc_id": "2001.05687", "revision_depth": 3} {"before_sent": " We compare the performance of the proposed model with several lexical-based and neural network-based baseline models.", "before_sent_with_intent": " We compare the performance of the proposed model with several lexical-based and neural network-based baseline models.", "after_sent": " We compare the performance of the proposed model with several lexical-based and neural network-based models.", "labels": "clarity", "doc_id": "2001.05687", "revision_depth": 3} {"before_sent": " Our proposed technique achieves 61.81\\% in accuracy, which is 5.51\\% higher than the best baseline model.", "before_sent_with_intent": " Our proposed technique achieves 61.81\\% in accuracy, which is 5.51\\% higher than the best baseline model.", "after_sent": " Our proposed method achieves 61.81\\% in accuracy, which is 5.51\\% higher than the best baseline model.", "labels": "clarity", "doc_id": "2001.05687", "revision_depth": 3} {"before_sent": " Our proposed technique achieves 61.81\\% in accuracy, which is 5.51\\% higher than the best baseline model.", "before_sent_with_intent": " Our proposed technique achieves 61.81\\% in accuracy, which is 5.51\\% higher than the best baseline model.", "after_sent": " Our proposed technique achieves 61.81\\% by accuracy, which is 5.51\\% higher than the best baseline model.", "labels": "clarity", "doc_id": "2001.05687", "revision_depth": 3} {"before_sent": " We also measure human performance on our dataset and find that there is a big gap between human and model performances.", "before_sent_with_intent": " We also measure human performance on our dataset and find that there is a big gap between human and model performances.", "after_sent": " We also measure human performance on our dataset and find that there is a big gap between machine-model and human performances.", "labels": "clarity", "doc_id": "2001.05687", "revision_depth": 3} {"before_sent": " The dataset is freely available at our website for research purposes.", "before_sent_with_intent": " The dataset is freely available at our website for research purposes.", "after_sent": " The dataset is freely available on our website for research purposes.", "labels": "fluency", "doc_id": "2001.05687", "revision_depth": 3} {"before_sent": " Finally, regular supervised training is performed on the resulting training set.", "before_sent_with_intent": " Finally, regular supervised training is performed on the resulting training set.", "after_sent": " Finally, standard supervised training is performed on the resulting training set.", "labels": "clarity", "doc_id": "2001.07676", "revision_depth": 2} {"before_sent": " For several tasks and languages, PET outperforms both supervised training and unsupervised approaches in low-resource settings by a large margin.", "before_sent_with_intent": " For several tasks and languages, PET outperforms both supervised training and unsupervised approaches in low-resource settings by a large margin.", "after_sent": " For several tasks and languages, PET outperforms supervised training and unsupervised approaches in low-resource settings by a large margin.", "labels": "clarity", "doc_id": "2001.07676", "revision_depth": 2} {"before_sent": " For several tasks and languages, PET outperforms both supervised training and unsupervised approaches in low-resource settings by a large margin.", "before_sent_with_intent": " For several tasks and languages, PET outperforms both supervised training and unsupervised approaches in low-resource settings by a large margin.", "after_sent": " For several tasks and languages, PET outperforms both supervised training and strong semi-supervised approaches in low-resource settings by a large margin.", "labels": "meaning-changed", "doc_id": "2001.07676", "revision_depth": 2} {"before_sent": " In this work, we extend this approach to the task of dialogue state tracking for goal-oriented dialogues, in which the data naturally exhibits a hierarchical structure over utterances and related annotations .", "before_sent_with_intent": " In this work, we extend this approach to the task of dialogue state tracking for goal-oriented dialogues, in which the data naturally exhibits a hierarchical structure over utterances and related annotations .", "after_sent": " In this work, we extend this approach to the task of dialog state tracking for goal-oriented dialogues, in which the data naturally exhibits a hierarchical structure over utterances and related annotations .", "labels": "fluency", "doc_id": "2001.08604", "revision_depth": 1} {"before_sent": " In this work, we extend this approach to the task of dialogue state tracking for goal-oriented dialogues, in which the data naturally exhibits a hierarchical structure over utterances and related annotations .", "before_sent_with_intent": " In this work, we extend this approach to the task of dialogue state tracking for goal-oriented dialogues, in which the data naturally exhibits a hierarchical structure over utterances and related annotations .", "after_sent": " In this work, we extend this approach to the task of dialogue state tracking for goal-oriented dialogs. Since, goal-oriented dialogs naturally exhibit a hierarchical structure over utterances and related annotations .", "labels": "coherence", "doc_id": "2001.08604", "revision_depth": 1} {"before_sent": " In this work, we extend this approach to the task of dialogue state tracking for goal-oriented dialogues, in which the data naturally exhibits a hierarchical structure over utterances and related annotations . Deep generative data augmentation for dialogue state tracking requires the generative model to be aware of the hierarchically structured data .", "before_sent_with_intent": " In this work, we extend this approach to the task of dialogue state tracking for goal-oriented dialogues, in which the data naturally exhibits a hierarchical structure over utterances and related annotations . Deep generative data augmentation for dialogue state tracking requires the generative model to be aware of the hierarchically structured data .", "after_sent": " In this work, we extend this approach to the task of dialogue state tracking for goal-oriented dialogues, in which the data naturally exhibits a hierarchical structure over utterances and related annotations , deep generative data augmentation for dialogue state tracking requires the generative model to be aware of the hierarchically structured data .", "labels": "fluency", "doc_id": "2001.08604", "revision_depth": 1} {"before_sent": " Deep generative data augmentation for dialogue state tracking requires the generative model to be aware of the hierarchically structured data .", "before_sent_with_intent": " Deep generative data augmentation for dialogue state tracking requires the generative model to be aware of the hierarchically structured data .", "after_sent": " Deep generative data augmentation for the task requires the generative model to be aware of the hierarchically structured data .", "labels": "clarity", "doc_id": "2001.08604", "revision_depth": 1} {"before_sent": " Deep generative data augmentation for dialogue state tracking requires the generative model to be aware of the hierarchically structured data .", "before_sent_with_intent": " Deep generative data augmentation for dialogue state tracking requires the generative model to be aware of the hierarchically structured data .", "after_sent": " Deep generative data augmentation for dialogue state tracking requires the generative model to be aware of the hierarchical nature .", "labels": "clarity", "doc_id": "2001.08604", "revision_depth": 1} {"before_sent": " We propose Variational Hierarchical Dialog Autoencoder (VHDA) for modeling various aspects of goal-oriented dialogues , including linguistic and underlying annotation structures.", "before_sent_with_intent": " We propose Variational Hierarchical Dialog Autoencoder (VHDA) for modeling various aspects of goal-oriented dialogues , including linguistic and underlying annotation structures.", "after_sent": " We propose the Variational Hierarchical Dialog Autoencoder (VHDA) for modeling various aspects of goal-oriented dialogues , including linguistic and underlying annotation structures.", "labels": "fluency", "doc_id": "2001.08604", "revision_depth": 1} {"before_sent": " We propose Variational Hierarchical Dialog Autoencoder (VHDA) for modeling various aspects of goal-oriented dialogues , including linguistic and underlying annotation structures.", "before_sent_with_intent": " We propose Variational Hierarchical Dialog Autoencoder (VHDA) for modeling various aspects of goal-oriented dialogues , including linguistic and underlying annotation structures.", "after_sent": " We propose Variational Hierarchical Dialog Autoencoder (VHDA) for modeling complete aspects of goal-oriented dialogues , including linguistic and underlying annotation structures.", "labels": "meaning-changed", "doc_id": "2001.08604", "revision_depth": 1} {"before_sent": " We propose Variational Hierarchical Dialog Autoencoder (VHDA) for modeling various aspects of goal-oriented dialogues , including linguistic and underlying annotation structures.", "before_sent_with_intent": " We propose Variational Hierarchical Dialog Autoencoder (VHDA) for modeling various aspects of goal-oriented dialogues , including linguistic and underlying annotation structures.", "after_sent": " We propose Variational Hierarchical Dialog Autoencoder (VHDA) for modeling various aspects of goal-oriented dialogs , including linguistic and underlying annotation structures.", "labels": "fluency", "doc_id": "2001.08604", "revision_depth": 1} {"before_sent": " We propose Variational Hierarchical Dialog Autoencoder (VHDA) for modeling various aspects of goal-oriented dialogues , including linguistic and underlying annotation structures. Our experiments show that our model is able to generate realistic and novel samples that improve the robustness of state-of-the-art dialogue state trackers, ultimately improving their final dialogue state tracking performances on several datasets .", "before_sent_with_intent": " We propose Variational Hierarchical Dialog Autoencoder (VHDA) for modeling various aspects of goal-oriented dialogues , including linguistic and underlying annotation structures. Our experiments show that our model is able to generate realistic and novel samples that improve the robustness of state-of-the-art dialogue state trackers, ultimately improving their final dialogue state tracking performances on several datasets .", "after_sent": " We propose Variational Hierarchical Dialog Autoencoder (VHDA) for modeling various aspects of goal-oriented dialogues , including linguistic features and underlying structured annotations, namely dialog acts and goals. We also propose two training policies to mitigate issues that arise with training VAE-based models. Experiments show that our model is able to generate realistic and novel samples that improve the robustness of state-of-the-art dialogue state trackers, ultimately improving their final dialogue state tracking performances on several datasets .", "labels": "meaning-changed", "doc_id": "2001.08604", "revision_depth": 1} {"before_sent": " Our experiments show that our model is able to generate realistic and novel samples that improve the robustness of state-of-the-art dialogue state trackers, ultimately improving their final dialogue state tracking performances on several datasets .", "before_sent_with_intent": " Our experiments show that our model is able to generate realistic and novel samples that improve the robustness of state-of-the-art dialogue state trackers, ultimately improving their final dialogue state tracking performances on several datasets .", "after_sent": " Our experiments show that our hierarchical model is able to generate realistic and novel samples that improve the robustness of state-of-the-art dialogue state trackers, ultimately improving their final dialogue state tracking performances on several datasets .", "labels": "meaning-changed", "doc_id": "2001.08604", "revision_depth": 1} {"before_sent": " Our experiments show that our model is able to generate realistic and novel samples that improve the robustness of state-of-the-art dialogue state trackers, ultimately improving their final dialogue state tracking performances on several datasets .", "before_sent_with_intent": " Our experiments show that our model is able to generate realistic and novel samples that improve the robustness of state-of-the-art dialogue state trackers, ultimately improving their final dialogue state tracking performances on several datasets .", "after_sent": " Our experiments show that our model is able to generate realistic and novel samples that improve the robustness of state-of-the-art dialog state trackers, ultimately improving their final dialogue state tracking performances on several datasets .", "labels": "fluency", "doc_id": "2001.08604", "revision_depth": 1} {"before_sent": " Our experiments show that our model is able to generate realistic and novel samples that improve the robustness of state-of-the-art dialogue state trackers, ultimately improving their final dialogue state tracking performances on several datasets .", "before_sent_with_intent": " Our experiments show that our model is able to generate realistic and novel samples that improve the robustness of state-of-the-art dialogue state trackers, ultimately improving their final dialogue state tracking performances on several datasets .", "after_sent": " Our experiments show that our model is able to generate realistic and novel samples that improve the robustness of state-of-the-art dialogue state trackers, ultimately improving the dialog state tracking performances on several datasets .", "labels": "clarity", "doc_id": "2001.08604", "revision_depth": 1} {"before_sent": " Our experiments show that our model is able to generate realistic and novel samples that improve the robustness of state-of-the-art dialogue state trackers, ultimately improving their final dialogue state tracking performances on several datasets .", "before_sent_with_intent": " Our experiments show that our model is able to generate realistic and novel samples that improve the robustness of state-of-the-art dialogue state trackers, ultimately improving their final dialogue state tracking performances on several datasets .", "after_sent": " Our experiments show that our model is able to generate realistic and novel samples that improve the robustness of state-of-the-art dialogue state trackers, ultimately improving their final dialogue state tracking performances on various dialog domains. Surprisingly, the ability to jointly generate dialog features enables our model to outperform previous state-of-the-arts in related subtasks, such as language generation and user simulation .", "labels": "meaning-changed", "doc_id": "2001.08604", "revision_depth": 1} {"before_sent": "Recent works have shown that generative data augmentation, where synthetic samples generated from deep generative models are used to augment the training dataset, benefit certain NLP tasks.", "before_sent_with_intent": " Recent works have shown that generative data augmentation, where synthetic samples generated from deep generative models are used to augment the training dataset, benefit certain NLP tasks.", "after_sent": "Recent works have shown that generative data augmentation, where synthetic samples generated from deep generative models complement the training dataset, benefit certain NLP tasks.", "labels": "clarity", "doc_id": "2001.08604", "revision_depth": 2} {"before_sent": "Recent works have shown that generative data augmentation, where synthetic samples generated from deep generative models are used to augment the training dataset, benefit certain NLP tasks.", "before_sent_with_intent": " Recent works have shown that generative data augmentation, where synthetic samples generated from deep generative models are used to augment the training dataset, benefit certain NLP tasks.", "after_sent": "Recent works have shown that generative data augmentation, where synthetic samples generated from deep generative models are used to augment the training dataset, benefit NLP tasks.", "labels": "clarity", "doc_id": "2001.08604", "revision_depth": 2} {"before_sent": " Since, goal-oriented dialogs naturally exhibit a hierarchical structure over utterances and related annotations, deep generative data augmentation for the task requires the generative model to be aware of the hierarchical nature .", "before_sent_with_intent": " Since, goal-oriented dialogs naturally exhibit a hierarchical structure over utterances and related annotations, deep generative data augmentation for the task requires the generative model to be aware of the hierarchical nature .", "after_sent": " Due to the inherent hierarchical structure of goal-oriented dialogs naturally exhibit a hierarchical structure over utterances and related annotations, deep generative data augmentation for the task requires the generative model to be aware of the hierarchical nature .", "labels": "clarity", "doc_id": "2001.08604", "revision_depth": 2} {"before_sent": " Since, goal-oriented dialogs naturally exhibit a hierarchical structure over utterances and related annotations, deep generative data augmentation for the task requires the generative model to be aware of the hierarchical nature .", "before_sent_with_intent": " Since, goal-oriented dialogs naturally exhibit a hierarchical structure over utterances and related annotations, deep generative data augmentation for the task requires the generative model to be aware of the hierarchical nature .", "after_sent": " Since, goal-oriented dialogs over utterances and related annotations, deep generative data augmentation for the task requires the generative model to be aware of the hierarchical nature .", "labels": "clarity", "doc_id": "2001.08604", "revision_depth": 2} {"before_sent": " Since, goal-oriented dialogs naturally exhibit a hierarchical structure over utterances and related annotations, deep generative data augmentation for the task requires the generative model to be aware of the hierarchical nature .", "before_sent_with_intent": " Since, goal-oriented dialogs naturally exhibit a hierarchical structure over utterances and related annotations, deep generative data augmentation for the task requires the generative model to be aware of the hierarchical nature .", "after_sent": " Since, goal-oriented dialogs naturally exhibit a hierarchical structure over utterances and related annotations, the deep generative model must be capable of capturing the coherence among different hierarchies and types of dialog features .", "labels": "clarity", "doc_id": "2001.08604", "revision_depth": 2} {"before_sent": " We propose the Variational Hierarchical Dialog Autoencoder (VHDA) for modeling complete aspects of goal-oriented dialogs, including linguistic features and underlying structured annotations, namely dialog acts and goals.", "before_sent_with_intent": " We propose the Variational Hierarchical Dialog Autoencoder (VHDA) for modeling complete aspects of goal-oriented dialogs, including linguistic features and underlying structured annotations, namely dialog acts and goals.", "after_sent": " We propose the Variational Hierarchical Dialog Autoencoder (VHDA) for modeling the complete aspects of goal-oriented dialogs, including linguistic features and underlying structured annotations, namely dialog acts and goals.", "labels": "fluency", "doc_id": "2001.08604", "revision_depth": 2} {"before_sent": " We propose the Variational Hierarchical Dialog Autoencoder (VHDA) for modeling complete aspects of goal-oriented dialogs, including linguistic features and underlying structured annotations, namely dialog acts and goals.", "before_sent_with_intent": " We propose the Variational Hierarchical Dialog Autoencoder (VHDA) for modeling complete aspects of goal-oriented dialogs, including linguistic features and underlying structured annotations, namely dialog acts and goals.", "after_sent": " We propose the Variational Hierarchical Dialog Autoencoder (VHDA) for modeling complete aspects of goal-oriented dialogs, including linguistic features and underlying structured annotations, namely speaker information, dialog acts, and goals.", "labels": "clarity", "doc_id": "2001.08604", "revision_depth": 2} {"before_sent": " We also propose two training policies to mitigate issues that arise with training VAE-based models.", "before_sent_with_intent": " We also propose two training policies to mitigate issues that arise with training VAE-based models.", "after_sent": " The proposed architecture is designed to model each aspect of goal-oriented dialogs using inter-connected latent variables and learns to generate coherent goal-oriented dialogs from the latent spaces. To overcome training issues that arise with training VAE-based models.", "labels": "meaning-changed", "doc_id": "2001.08604", "revision_depth": 2} {"before_sent": " We also propose two training policies to mitigate issues that arise with training VAE-based models. Experiments show that our hierarchical model is able to generate realistic and novel samples that improve the robustness of state-of-the-art dialog state trackers, ultimately improving the dialog state tracking performances on various dialog domains.", "before_sent_with_intent": " We also propose two training policies to mitigate issues that arise with training VAE-based models. Experiments show that our hierarchical model is able to generate realistic and novel samples that improve the robustness of state-of-the-art dialog state trackers, ultimately improving the dialog state tracking performances on various dialog domains.", "after_sent": " We also propose two training policies to mitigate issues that arise from training complex variational models, we propose appropriate training strategies. Experiments on various dialog datasets show that our hierarchical model is able to generate realistic and novel samples that improve the robustness of state-of-the-art dialog state trackers, ultimately improving the dialog state tracking performances on various dialog domains.", "labels": "clarity", "doc_id": "2001.08604", "revision_depth": 2} {"before_sent": " Experiments show that our hierarchical model is able to generate realistic and novel samples that improve the robustness of state-of-the-art dialog state trackers, ultimately improving the dialog state tracking performances on various dialog domains. Surprisingly, the ability to jointly generate dialog features enables our model to outperform previous state-of-the-arts in related subtasks, such as language generation and user simulation .", "before_sent_with_intent": " Experiments show that our hierarchical model is able to generate realistic and novel samples that improve the robustness of state-of-the-art dialog state trackers, ultimately improving the dialog state tracking performances on various dialog domains. Surprisingly, the ability to jointly generate dialog features enables our model to outperform previous state-of-the-arts in related subtasks, such as language generation and user simulation .", "after_sent": " Experiments show that our model improves the downstream dialog trackers' robustness via generative data augmentation. We also discover additional benefits of our unified approach to modeling goal-oriented dialogs: dialog response generation and user simulation .", "labels": "clarity", "doc_id": "2001.08604", "revision_depth": 2} {"before_sent": " Surprisingly, the ability to jointly generate dialog features enables our model to outperform previous state-of-the-arts in related subtasks, such as language generation and user simulation .", "before_sent_with_intent": " Surprisingly, the ability to jointly generate dialog features enables our model to outperform previous state-of-the-arts in related subtasks, such as language generation and user simulation .", "after_sent": " Surprisingly, the ability to jointly generate dialog features enables our model to outperform previous state-of-the-arts in related subtasks, such as language generation and user simulation , where our model outperforms previous strong baselines .", "labels": "meaning-changed", "doc_id": "2001.08604", "revision_depth": 2} {"before_sent": " How can neural models make sample-efficient generalizations from task-language combinations with available data to low-resource ones?", "before_sent_with_intent": " How can neural models make sample-efficient generalizations from task-language combinations with available data to low-resource ones?", "after_sent": " How can neural models make sample-efficient generalizations from task--language combinations with available data to low-resource ones?", "labels": "fluency", "doc_id": "2001.11453", "revision_depth": 1} {"before_sent": " We infer the posteriors over such latent variables based on data from seen task-language combinations through variational inference.", "before_sent_with_intent": " We infer the posteriors over such latent variables based on data from seen task-language combinations through variational inference.", "after_sent": " We infer the posteriors over such latent variables based on data from seen task--language combinations through variational inference.", "labels": "fluency", "doc_id": "2001.11453", "revision_depth": 1} {"before_sent": " In particular, we experiment with a typologically diverse sample of 33 languages from 4 continents and 11 families, and show that our model yields comparable or better results than state-of-the-art, zero-shot cross-lingual transfer methods ; it increases performance by 4.49 points for POS tagging and 7.73 points for NER on average compared to the strongest baseline .", "before_sent_with_intent": " In particular, we experiment with a typologically diverse sample of 33 languages from 4 continents and 11 families, and show that our model yields comparable or better results than state-of-the-art, zero-shot cross-lingual transfer methods ; it increases performance by 4.49 points for POS tagging and 7.73 points for NER on average compared to the strongest baseline .", "after_sent": " In particular, we experiment with a typologically diverse sample of 33 languages from 4 continents and 11 families, and show that our model yields comparable or better results than state-of-the-art, zero-shot cross-lingual transfer methods . Moreover, we demonstrate that approximate Bayesian model averaging results in smoother predictive distributions, whose entropy strongly correlates with accuracy. Hence, the proposed framework also offers robust estimates of uncertainty .", "labels": "meaning-changed", "doc_id": "2001.11453", "revision_depth": 1} {"before_sent": " How can neural models make sample-efficient generalizations from task--language combinations with available data to low-resource ones?", "before_sent_with_intent": " How can neural models make sample-efficient generalizations from task--language combinations with available data to low-resource ones?", "after_sent": " How can neural models make sample-efficient generalizations from task-language combinations with available data to low-resource ones?", "labels": "fluency", "doc_id": "2001.11453", "revision_depth": 2} {"before_sent": " We infer the posteriors over such latent variables based on data from seen task--language combinations through variational inference.", "before_sent_with_intent": " We infer the posteriors over such latent variables based on data from seen task--language combinations through variational inference.", "after_sent": " We infer the posteriors over such latent variables based on data from seen task-language combinations through variational inference.", "labels": "fluency", "doc_id": "2001.11453", "revision_depth": 2} {"before_sent": " Moreover, we demonstrate that approximate Bayesian model averaging results in smoother predictive distributions, whose entropy strongly correlates with accuracy.", "before_sent_with_intent": " Moreover, we demonstrate that approximate Bayesian model averaging results in smoother predictive distributions, whose entropy strongly correlates with accuracy.", "after_sent": " Moreover, we demonstrate that approximate Bayesian model averaging results in smoother predictive distributions, whose entropy inversely correlates with accuracy.", "labels": "meaning-changed", "doc_id": "2001.11453", "revision_depth": 2} {"before_sent": " Hence, the proposed framework also offers robust estimates of uncertainty.", "before_sent_with_intent": " Hence, the proposed framework also offers robust estimates of uncertainty.", "after_sent": " Hence, the proposed framework also offers robust estimates of prediction uncertainty. Our code is located at github.com/cambridgeltl/parameter-factorization", "labels": "meaning-changed", "doc_id": "2001.11453", "revision_depth": 2} {"before_sent": "We propose UniViLM: a Unified Video and Language pre-training Model for multimodal understanding and generation. Motivated by the recent success of BERT based pre-training technique for NLP and image-language tasks, VideoBERT and CBT are proposed to exploit BERT model for video and language pre-training using narrated instructional videos.", "before_sent_with_intent": " We propose UniViLM: a Unified Video and Language pre-training Model for multimodal understanding and generation. Motivated by the recent success of BERT based pre-training technique for NLP and image-language tasks, VideoBERT and CBT are proposed to exploit BERT model for video and language pre-training using narrated instructional videos.", "after_sent": "With the recent success of BERT based pre-training technique for NLP and image-language tasks, VideoBERT and CBT are proposed to exploit BERT model for video and language pre-training using narrated instructional videos.", "labels": "coherence", "doc_id": "2002.06353", "revision_depth": 1} {"before_sent": " Motivated by the recent success of BERT based pre-training technique for NLP and image-language tasks, VideoBERT and CBT are proposed to exploit BERT model for video and language pre-training using narrated instructional videos.", "before_sent_with_intent": " Motivated by the recent success of BERT based pre-training technique for NLP and image-language tasks, VideoBERT and CBT are proposed to exploit BERT model for video and language pre-training using narrated instructional videos.", "after_sent": " Motivated by the recent success of pre-training technique for NLP and image-language tasks, VideoBERT and CBT are proposed to exploit BERT model for video and language pre-training using narrated instructional videos.", "labels": "coherence", "doc_id": "2002.06353", "revision_depth": 1} {"before_sent": " Motivated by the recent success of BERT based pre-training technique for NLP and image-language tasks, VideoBERT and CBT are proposed to exploit BERT model for video and language pre-training using narrated instructional videos.", "before_sent_with_intent": " Motivated by the recent success of BERT based pre-training technique for NLP and image-language tasks, VideoBERT and CBT are proposed to exploit BERT model for video and language pre-training using narrated instructional videos.", "after_sent": " Motivated by the recent success of BERT based pre-training technique for NLP and image-linguistic tasks, there are still few works on video-linguistic pre-training using narrated instructional videos.", "labels": "clarity", "doc_id": "2002.06353", "revision_depth": 1} {"before_sent": " Motivated by the recent success of BERT based pre-training technique for NLP and image-language tasks, VideoBERT and CBT are proposed to exploit BERT model for video and language pre-training using narrated instructional videos. Different from their works which only pre-train understanding task, we propose a unified video-language pre-training model for both understanding and generation tasks .", "before_sent_with_intent": " Motivated by the recent success of BERT based pre-training technique for NLP and image-language tasks, VideoBERT and CBT are proposed to exploit BERT model for video and language pre-training using narrated instructional videos. Different from their works which only pre-train understanding task, we propose a unified video-language pre-training model for both understanding and generation tasks .", "after_sent": " Motivated by the recent success of BERT based pre-training technique for NLP and image-language tasks, VideoBERT and CBT are proposed to exploit BERT model for video and language pre-training . Besides, most of the existing multimodal models are pre-trained for understanding task, we propose a unified video-language pre-training model for both understanding and generation tasks .", "labels": "coherence", "doc_id": "2002.06353", "revision_depth": 1} {"before_sent": " Different from their works which only pre-train understanding task, we propose a unified video-language pre-training model for both understanding and generation tasks .", "before_sent_with_intent": " Different from their works which only pre-train understanding task, we propose a unified video-language pre-training model for both understanding and generation tasks .", "after_sent": " Different from their works which only pre-train understanding task, which leads to a pretrain-finetune discrepency for generation tasks. In this paper, we propose UniViLM: a Unified Video and Language pre-training model for both understanding and generation tasks .", "labels": "meaning-changed", "doc_id": "2002.06353", "revision_depth": 1} {"before_sent": " Different from their works which only pre-train understanding task, we propose a unified video-language pre-training model for both understanding and generation tasks .", "before_sent_with_intent": " Different from their works which only pre-train understanding task, we propose a unified video-language pre-training model for both understanding and generation tasks .", "after_sent": " Different from their works which only pre-train understanding task, we propose a unified video-language pre-training Model for both multimodal understanding and generation tasks .", "labels": "clarity", "doc_id": "2002.06353", "revision_depth": 1} {"before_sent": " Different from their works which only pre-train understanding task, we propose a unified video-language pre-training model for both understanding and generation tasks .", "before_sent_with_intent": " Different from their works which only pre-train understanding task, we propose a unified video-language pre-training model for both understanding and generation tasks .", "after_sent": " Different from their works which only pre-train understanding task, we propose a unified video-language pre-training model for both understanding and generation .", "labels": "clarity", "doc_id": "2002.06353", "revision_depth": 1} {"before_sent": " IMAGINE learns to represent goals by jointly learning a language model and a goal-conditioned reward function.", "before_sent_with_intent": " IMAGINE learns to represent goals by jointly learning a language model and a goal-conditioned reward function.", "after_sent": " IMAGINE learns to represent goals by jointly learning a language encoder and a goal-conditioned reward function.", "labels": "clarity", "doc_id": "2002.09253", "revision_depth": 1} {"before_sent": " Just like humans, our agent uses language compositionality to generate new goals by composing known ones . Leveraging modular model architectures based on Deep Sets and gated-attention mechanisms, IMAGINE autonomously builds a repertoire of behaviors and shows good zero-shot generalization properties for various types of generalization.", "before_sent_with_intent": " Just like humans, our agent uses language compositionality to generate new goals by composing known ones . Leveraging modular model architectures based on Deep Sets and gated-attention mechanisms, IMAGINE autonomously builds a repertoire of behaviors and shows good zero-shot generalization properties for various types of generalization.", "after_sent": " Just like humans, our agent uses language compositionality to generate new goals by composing known ones , using an algorithm grounded in construction grammar models of child language acquisition . Leveraging modular model architectures based on Deep Sets and gated-attention mechanisms, IMAGINE autonomously builds a repertoire of behaviors and shows good zero-shot generalization properties for various types of generalization.", "labels": "meaning-changed", "doc_id": "2002.09253", "revision_depth": 1} {"before_sent": " Just like humans, our agent uses language compositionality to generate new goals by composing known ones . Leveraging modular model architectures based on Deep Sets and gated-attention mechanisms, IMAGINE autonomously builds a repertoire of behaviors and shows good zero-shot generalization properties for various types of generalization.", "before_sent_with_intent": " Just like humans, our agent uses language compositionality to generate new goals by composing known ones . Leveraging modular model architectures based on Deep Sets and gated-attention mechanisms, IMAGINE autonomously builds a repertoire of behaviors and shows good zero-shot generalization properties for various types of generalization.", "after_sent": " Just like humans, our agent uses language compositionality to generate new goals by composing known ones . Leveraging modular model architectures based on deepsets and gated attention mechanisms, IMAGINE autonomously builds a repertoire of behaviors and shows good zero-shot generalization properties for various types of generalization.", "labels": "fluency", "doc_id": "2002.09253", "revision_depth": 1} {"before_sent": "Autonomous reinforcement learning agents must be intrinsically motivated to explore their environment, discover potential goals, represent them and learn how to achieve them.", "before_sent_with_intent": " Autonomous reinforcement learning agents must be intrinsically motivated to explore their environment, discover potential goals, represent them and learn how to achieve them.", "after_sent": "Developmental machine learning studies how artificial agents can model the way children learn open-ended repertoires of skills. Such agents need to create and represent goals, select which ones to pursue and learn to achieve them.", "labels": "meaning-changed", "doc_id": "2002.09253", "revision_depth": 2} {"before_sent": " As children do the same, they benefit from exposure to language, using it to formulate goals and imagine new ones as they learn their meaning. In our proposed learning architecture (IMAGINE), the agent freely explores its environment and turns natural language descriptions of interesting interactions from a social partner into potential goals. IMAGINE learns to represent goalsby jointly learning a language encoder and a goal-conditioned reward function . Just like humans, our agent uses language compositionality to generate new goals by composing known ones, using an algorithm grounded in construction grammar models of child language acquisition. Leveraging modular model architectures based on deepsets and gated attention mechanisms, IMAGINE autonomously builds a repertoire of behaviors and shows good zero-shot generalization properties for various types of generalization. When imagining its own goals, the agent leverages zero-shot generalization of the reward function to further train on imagined goals and refine its behavior. We present experiments in a simulated domain where the agent interacts with procedurally generated scenes containing objects of various types and colors, discovers goals, imagines others and learns to achieve them .", "before_sent_with_intent": " As children do the same, they benefit from exposure to language, using it to formulate goals and imagine new ones as they learn their meaning. In our proposed learning architecture (IMAGINE), the agent freely explores its environment and turns natural language descriptions of interesting interactions from a social partner into potential goals. IMAGINE learns to represent goalsby jointly learning a language encoder and a goal-conditioned reward function . Just like humans, our agent uses language compositionality to generate new goals by composing known ones, using an algorithm grounded in construction grammar models of child language acquisition. Leveraging modular model architectures based on deepsets and gated attention mechanisms, IMAGINE autonomously builds a repertoire of behaviors and shows good zero-shot generalization properties for various types of generalization. When imagining its own goals, the agent leverages zero-shot generalization of the reward function to further train on imagined goals and refine its behavior. We present experiments in a simulated domain where the agent interacts with procedurally generated scenes containing objects of various types and colors, discovers goals, imagines others and learns to achieve them .", "after_sent": " Recent approaches have considered goal spaces that were either fixed and hand-defined or learned using generative models of states. This limited agents to sample goals within the distribution of known effects. We argue that the ability to imagine out-of-distribution goals is key to enable creative discoveries and open-ended learning. Children do so by leveraging the compositionality of language as a tool to imagine descriptions of outcomes they never experienced before, targeting them as goals during play. We introduce Imagine, an intrinsically motivated deep reinforcement learning architecture that models this ability. Such imaginative agents, like children, benefit from the guidance of a social peer who provides language descriptions. To take advantage of goal imagination, agents must be able to leverage these descriptions to interpret their imagined out-of-distribution goals. This generalization is made possible by modularity: a decomposition between learned goal-achievement reward function and policy relying on deep sets, gated attention and object-centered representations. We introduce the Playground environment and study how this form of goal imagination improves generalization and exploration over agents lacking this capacity. In addition, we identify the properties of goal imagination that enable these results and study the impacts of modularity and social interactions .", "labels": "meaning-changed", "doc_id": "2002.09253", "revision_depth": 2} {"before_sent": "Producing natural and accurate responses like human beings is the ultimate goal of intelligent dialogue agents. So far, most of the past works concentrate on selecting or generating one pertinent and fluent response according to current query and its context. These models work on a one-to-one environment, making one response to one utterance each round.However, in real human-human conversations, human often sequentially sends several short messages for readability instead of a long message in one turn.", "before_sent_with_intent": " Producing natural and accurate responses like human beings is the ultimate goal of intelligent dialogue agents. So far, most of the past works concentrate on selecting or generating one pertinent and fluent response according to current query and its context. These models work on a one-to-one environment, making one response to one utterance each round.However, in real human-human conversations, human often sequentially sends several short messages for readability instead of a long message in one turn.", "after_sent": "Different people have different habits of describing their intents in conversations. Some people may tend to deliberate their full intents in several successive utterances, i.e., they use several consistent messages for readability instead of a long message in one turn.", "labels": "clarity", "doc_id": "2002.09616", "revision_depth": 1} {"before_sent": "However, in real human-human conversations, human often sequentially sends several short messages for readability instead of a long message in one turn. Thus messages will not end with an explicit ending signal, which is crucial for agents to decide when to reply. So the first step for an intelligent dialogue agent is not replying but deciding if it should reply at the moment. To address this issue, in this paper , we propose a novel Imagine-then-Arbitrate (ITA) neural dialogue model to help the agent decide whether to wait or to make a response directly.", "before_sent_with_intent": " However, in real human-human conversations, human often sequentially sends several short messages for readability instead of a long message in one turn. Thus messages will not end with an explicit ending signal, which is crucial for agents to decide when to reply. So the first step for an intelligent dialogue agent is not replying but deciding if it should reply at the moment. To address this issue, in this paper , we propose a novel Imagine-then-Arbitrate (ITA) neural dialogue model to help the agent decide whether to wait or to make a response directly.", "after_sent": "However, in real human-human conversations, human often sequentially sends several short messages for readability instead of a long sentence to express their question. This creates a predicament faced by dialogue systems' application, especially in real-world industrial scenarios, in which the dialogue system is unsure that whether it should answer the user's query immediately or wait for users' further supplementary input. Motivated by such interesting quandary, we define a novel task: Wait-or-Answer to better tackle this dilemma faced by dialogue systems. We shed light on a new research topic about how the dialogue system can be more competent to behave in this Wait-or-Answer quandary. Further , we propose a novel Imagine-then-Arbitrate (ITA) neural dialogue model to help the agent decide whether to wait or to make a response directly.", "labels": "meaning-changed", "doc_id": "2002.09616", "revision_depth": 1} {"before_sent": " To address this issue, in this paper , we propose a novel Imagine-then-Arbitrate (ITA) neural dialogue model to help the agent decide whether to wait or to make a response directly.", "before_sent_with_intent": " To address this issue, in this paper , we propose a novel Imagine-then-Arbitrate (ITA) neural dialogue model to help the agent decide whether to wait or to make a response directly.", "after_sent": " To address this issue, in this paper , we propose a predictive approach dubbed Imagine-then-Arbitrate (ITA) neural dialogue model to help the agent decide whether to wait or to make a response directly.", "labels": "clarity", "doc_id": "2002.09616", "revision_depth": 1} {"before_sent": " To address this issue, in this paper , we propose a novel Imagine-then-Arbitrate (ITA) neural dialogue model to help the agent decide whether to wait or to make a response directly.", "before_sent_with_intent": " To address this issue, in this paper , we propose a novel Imagine-then-Arbitrate (ITA) neural dialogue model to help the agent decide whether to wait or to make a response directly.", "after_sent": " To address this issue, in this paper , we propose a novel Imagine-then-Arbitrate (ITA) to resolve this Wait-or-Answer task. More specifically, we take advantage of an arbitrator model to help the agent decide whether to wait or to make a response directly.", "labels": "meaning-changed", "doc_id": "2002.09616", "revision_depth": 1} {"before_sent": " To address this issue, in this paper , we propose a novel Imagine-then-Arbitrate (ITA) neural dialogue model to help the agent decide whether to wait or to make a response directly.", "before_sent_with_intent": " To address this issue, in this paper , we propose a novel Imagine-then-Arbitrate (ITA) neural dialogue model to help the agent decide whether to wait or to make a response directly.", "after_sent": " To address this issue, in this paper , we propose a novel Imagine-then-Arbitrate (ITA) neural dialogue model to help the dialogue system decide to wait or to make a response directly.", "labels": "clarity", "doc_id": "2002.09616", "revision_depth": 1} {"before_sent": " To address this issue, in this paper , we propose a novel Imagine-then-Arbitrate (ITA) neural dialogue model to help the agent decide whether to wait or to make a response directly. Our method has two imaginator modules and an arbitrator module. The two imaginators will learn the agent's and user's speaking style respectively, generate possible utterances as the input of the arbitrator, combining with dialogue history. And the arbitrator decides whether to wait or to make a response to the user directly. To verify the performance and effectiveness of our method, we prepared two dialogue datasets and compared our approach with several popular models. Experimental results show that our model performs well on addressing ending prediction issue and outperforms baseline models .", "before_sent_with_intent": " To address this issue, in this paper , we propose a novel Imagine-then-Arbitrate (ITA) neural dialogue model to help the agent decide whether to wait or to make a response directly. Our method has two imaginator modules and an arbitrator module. The two imaginators will learn the agent's and user's speaking style respectively, generate possible utterances as the input of the arbitrator, combining with dialogue history. And the arbitrator decides whether to wait or to make a response to the user directly. To verify the performance and effectiveness of our method, we prepared two dialogue datasets and compared our approach with several popular models. Experimental results show that our model performs well on addressing ending prediction issue and outperforms baseline models .", "after_sent": " To address this issue, in this paper , we propose a novel Imagine-then-Arbitrate (ITA) neural dialogue model to help the agent decide whether to wait or answer. The arbitrator's decision is made with the assistance of two ancillary imaginator models: a wait imaginator and an answer imaginator. The wait imaginator tries to predict what the user would supplement and use its prediction to persuade the arbitrator that the user has some information to add, so the dialogue system should wait. The answer imaginator, nevertheless, struggles to predict the answer of the dialogue system and convince the arbitrator that it's a superior choice to answer the users' query immediately. To our best knowledge, our paper is the first work to explicitly define the Wait-or-Answer task in the dialogue system. Additionally, our proposed ITA approach significantly outperforms the existing models in solving this Wait-or-Answer problem .", "labels": "meaning-changed", "doc_id": "2002.09616", "revision_depth": 1} {"before_sent": "Different people have different habits of describing their intents in conversations. Some people may tend to deliberate their full intents in several successive utterances, i. e., they use several consistent messages for readability instead of a long sentence to express their question.", "before_sent_with_intent": " Different people have different habits of describing their intents in conversations. Some people may tend to deliberate their full intents in several successive utterances, i. e., they use several consistent messages for readability instead of a long sentence to express their question.", "after_sent": "Producing natural and accurate responses like human beings is the ultimate goal of intelligent dialogue agents. So far, most of the past works concentrate on selecting or generating one pertinent and fluent response according to current query and its context. These models work on a one-to-one environment, making one response to one utterance each round. However, in real human-human conversations, human often sequentially sends several short messages for readability instead of a long sentence to express their question.", "labels": "meaning-changed", "doc_id": "2002.09616", "revision_depth": 2} {"before_sent": " Some people may tend to deliberate their full intents in several successive utterances, i. e., they use several consistent messages for readability instead of a long sentence to express their question. This creates a predicament faced by dialogue systems' application, especially in real-world industrial scenarios, in which the dialogue system is unsure that whether it should answer the user's query immediately or wait for users' further supplementary input. Motivated by such interesting quandary, we define a novel task: Wait-or-Answer to better tackle this dilemma faced by dialogue systems. We shed light on a new research topic about how the dialogue system can be more competent to behave in this Wait-or-Answer quandary. Further , we propose a predictive approach dubbed Imagine-then-Arbitrate (ITA) to resolve this Wait-or-Answer task.", "before_sent_with_intent": " Some people may tend to deliberate their full intents in several successive utterances, i. e., they use several consistent messages for readability instead of a long sentence to express their question. This creates a predicament faced by dialogue systems' application, especially in real-world industrial scenarios, in which the dialogue system is unsure that whether it should answer the user's query immediately or wait for users' further supplementary input. Motivated by such interesting quandary, we define a novel task: Wait-or-Answer to better tackle this dilemma faced by dialogue systems. We shed light on a new research topic about how the dialogue system can be more competent to behave in this Wait-or-Answer quandary. Further , we propose a predictive approach dubbed Imagine-then-Arbitrate (ITA) to resolve this Wait-or-Answer task.", "after_sent": " Some people may tend to deliberate their full intents in several successive utterances, i. e., they use several consistent messages for readability instead of a long message in one turn. Thus messages will not end with an explicit ending signal, which is crucial for agents to decide when to reply. So the first step for an intelligent dialogue agent is not replying but deciding if it should reply at the moment. To address this issue, in this paper , we propose a predictive approach dubbed Imagine-then-Arbitrate (ITA) to resolve this Wait-or-Answer task.", "labels": "meaning-changed", "doc_id": "2002.09616", "revision_depth": 2} {"before_sent": " Further , we propose a predictive approach dubbed Imagine-then-Arbitrate (ITA) to resolve this Wait-or-Answer task.", "before_sent_with_intent": " Further , we propose a predictive approach dubbed Imagine-then-Arbitrate (ITA) to resolve this Wait-or-Answer task.", "after_sent": " Further , we propose a novel Imagine-then-Arbitrate (ITA) to resolve this Wait-or-Answer task.", "labels": "clarity", "doc_id": "2002.09616", "revision_depth": 2} {"before_sent": " Further , we propose a predictive approach dubbed Imagine-then-Arbitrate (ITA) to resolve this Wait-or-Answer task. More specifically, we take advantage of an arbitrator model to help the dialogue system decide to wait or answer.", "before_sent_with_intent": " Further , we propose a predictive approach dubbed Imagine-then-Arbitrate (ITA) to resolve this Wait-or-Answer task. More specifically, we take advantage of an arbitrator model to help the dialogue system decide to wait or answer.", "after_sent": " Further , we propose a predictive approach dubbed Imagine-then-Arbitrate (ITA) neural dialogue model to help the dialogue system decide to wait or answer.", "labels": "coherence", "doc_id": "2002.09616", "revision_depth": 2} {"before_sent": " More specifically, we take advantage of an arbitrator model to help the dialogue system decide to wait or answer.", "before_sent_with_intent": " More specifically, we take advantage of an arbitrator model to help the dialogue system decide to wait or answer.", "after_sent": " More specifically, we take advantage of an arbitrator model to help the agent decide whether to wait or answer.", "labels": "clarity", "doc_id": "2002.09616", "revision_depth": 2} {"before_sent": " More specifically, we take advantage of an arbitrator model to help the dialogue system decide to wait or answer. The arbitrator's decision is made with the assistance of two ancillary imaginator models: a wait imaginator and an answer imaginator. The wait imaginator tries to predict what the user would supplement and use its prediction to persuade the arbitrator that the userhas some information to add, so the dialogue system should wait. The answer imaginator, nevertheless, struggles to predict the answer of the dialogue system and convince the arbitrator that it's a superior choice to answer the users' query immediately. To our best knowledge, our paper is the first work to explicitly define the Wait-or-Answer task in the dialogue system. Additionally, our proposed ITA approach significantly outperforms the existing modelsin solving this Wait-or-Answer problem .", "before_sent_with_intent": " More specifically, we take advantage of an arbitrator model to help the dialogue system decide to wait or answer. The arbitrator's decision is made with the assistance of two ancillary imaginator models: a wait imaginator and an answer imaginator. The wait imaginator tries to predict what the user would supplement and use its prediction to persuade the arbitrator that the userhas some information to add, so the dialogue system should wait. The answer imaginator, nevertheless, struggles to predict the answer of the dialogue system and convince the arbitrator that it's a superior choice to answer the users' query immediately. To our best knowledge, our paper is the first work to explicitly define the Wait-or-Answer task in the dialogue system. Additionally, our proposed ITA approach significantly outperforms the existing modelsin solving this Wait-or-Answer problem .", "after_sent": " More specifically, we take advantage of an arbitrator model to help the dialogue system decide to wait or to make a response directly. Our method has two imaginator modules and an arbitrator module. The two imaginators will learn the agent's and user's speaking style respectively, generate possible utterances as the input of the arbitrator, combining with dialogue history. And the arbitrator decides whether to wait or to make a response to the user directly. To verify the performance and effectiveness of our method, we prepared two dialogue datasets and compared our approach with several popular models. Experimental results show that our model performs well on addressing ending prediction issue and outperforms baseline models .", "labels": "meaning-changed", "doc_id": "2002.09616", "revision_depth": 2} {"before_sent": " To this end, we used data gathered by the CrowdSource team at Google Research in 2019 and fine-tuned pre-trained BERT model on our problem.", "before_sent_with_intent": " To this end, we used data gathered by the CrowdSource team at Google Research in 2019 and fine-tuned pre-trained BERT model on our problem.", "after_sent": " To this end, we used data gathered by the CrowdSource team at Google Research in 2019 and a fine-tuned pre-trained BERT model on our problem.", "labels": "fluency", "doc_id": "2002.10107", "revision_depth": 2} {"before_sent": " Based on evaluation by Mean-Squared-Error (MSE), model achieved the value of 0.046 after 2 epochs of training, which did not improve substantially in the next ones.", "before_sent_with_intent": " Based on evaluation by Mean-Squared-Error (MSE), model achieved the value of 0.046 after 2 epochs of training, which did not improve substantially in the next ones.", "after_sent": " Based on the evaluation by Mean-Squared-Error (MSE), model achieved the value of 0.046 after 2 epochs of training, which did not improve substantially in the next ones.", "labels": "fluency", "doc_id": "2002.10107", "revision_depth": 2} {"before_sent": " Based on evaluation by Mean-Squared-Error (MSE), model achieved the value of 0.046 after 2 epochs of training, which did not improve substantially in the next ones.", "before_sent_with_intent": " Based on evaluation by Mean-Squared-Error (MSE), model achieved the value of 0.046 after 2 epochs of training, which did not improve substantially in the next ones.", "after_sent": " Based on evaluation by Mean-Squared-Error (MSE), the model achieved a value of 0.046 after 2 epochs of training, which did not improve substantially in the next ones.", "labels": "clarity", "doc_id": "2002.10107", "revision_depth": 2} {"before_sent": " Previous attempts to learn variational auto-encoders for language data ? have had mixed success, with empirical performance well below state-of-the-art auto-regressive models, a key barrier being the ? occurrence of posterior collapse with VAEs.", "before_sent_with_intent": " Previous attempts to learn variational auto-encoders for language data ? have had mixed success, with empirical performance well below state-of-the-art auto-regressive models, a key barrier being the ? occurrence of posterior collapse with VAEs.", "after_sent": " Previous attempts to learn variational auto-encoders for language data have had mixed success, with empirical performance well below state-of-the-art auto-regressive models, a key barrier being the ? occurrence of posterior collapse with VAEs.", "labels": "fluency", "doc_id": "2003.02645", "revision_depth": 1} {"before_sent": " Previous attempts to learn variational auto-encoders for language data ? have had mixed success, with empirical performance well below state-of-the-art auto-regressive models, a key barrier being the ? occurrence of posterior collapse with VAEs.", "before_sent_with_intent": " Previous attempts to learn variational auto-encoders for language data ? have had mixed success, with empirical performance well below state-of-the-art auto-regressive models, a key barrier being the ? occurrence of posterior collapse with VAEs.", "after_sent": " Previous attempts to learn variational auto-encoders for language data ? have had mixed success, with empirical performance well below state-of-the-art auto-regressive models, a key barrier being the occurrence of posterior collapse with VAEs.", "labels": "fluency", "doc_id": "2003.02645", "revision_depth": 1} {"before_sent": "We introduce sentenceMIM, a probabilistic auto-encoder for language modelling , trained with Mutual Information Machine (MIM) learning . Previous attempts to learn variational auto-encoders for language data have had mixed success, with empirical performance well below state-of-the-art auto-regressive models, a key barrier being the occurrence of posterior collapsewith VAEs.", "before_sent_with_intent": " We introduce sentenceMIM, a probabilistic auto-encoder for language modelling , trained with Mutual Information Machine (MIM) learning . Previous attempts to learn variational auto-encoders for language data have had mixed success, with empirical performance well below state-of-the-art auto-regressive models, a key barrier being the occurrence of posterior collapsewith VAEs.", "after_sent": "SentenceMIM is a probabilistic auto-encoder for language modelling , trained with Mutual Information Machine (MIM) learning . Previous attempts to learn variational auto-encoders for language data have had mixed success, with empirical performance well below state-of-the-art auto-regressive models, a key barrier being the occurrence of posterior collapsewith VAEs.", "labels": "clarity", "doc_id": "2003.02645", "revision_depth": 2} {"before_sent": "We introduce sentenceMIM, a probabilistic auto-encoder for language modelling , trained with Mutual Information Machine (MIM) learning . Previous attempts to learn variational auto-encoders for language data have had mixed success, with empirical performance well below state-of-the-art auto-regressive models, a key barrier being the occurrence of posterior collapsewith VAEs.", "before_sent_with_intent": " We introduce sentenceMIM, a probabilistic auto-encoder for language modelling , trained with Mutual Information Machine (MIM) learning . Previous attempts to learn variational auto-encoders for language data have had mixed success, with empirical performance well below state-of-the-art auto-regressive models, a key barrier being the occurrence of posterior collapsewith VAEs.", "after_sent": "We introduce sentenceMIM, a probabilistic auto-encoder for language data , trained with Mutual Information Machine (MIM) learning . Previous attempts to learn variational auto-encoders for language data have had mixed success, with empirical performance well below state-of-the-art auto-regressive models, a key barrier being the occurrence of posterior collapsewith VAEs.", "labels": "clarity", "doc_id": "2003.02645", "revision_depth": 2} {"before_sent": "We introduce sentenceMIM, a probabilistic auto-encoder for language modelling , trained with Mutual Information Machine (MIM) learning . Previous attempts to learn variational auto-encoders for language data have had mixed success, with empirical performance well below state-of-the-art auto-regressive models, a key barrier being the occurrence of posterior collapsewith VAEs.", "before_sent_with_intent": " We introduce sentenceMIM, a probabilistic auto-encoder for language modelling , trained with Mutual Information Machine (MIM) learning . Previous attempts to learn variational auto-encoders for language data have had mixed success, with empirical performance well below state-of-the-art auto-regressive models, a key barrier being the occurrence of posterior collapsewith VAEs.", "after_sent": "We introduce sentenceMIM, a probabilistic auto-encoder for language modelling , trained with Mutual Information Machine (MIM) learning to provide a fixed length representation of variable length language observations (ie, similar to VAE) . Previous attempts to learn variational auto-encoders for language data have had mixed success, with empirical performance well below state-of-the-art auto-regressive models, a key barrier being the occurrence of posterior collapsewith VAEs.", "labels": "meaning-changed", "doc_id": "2003.02645", "revision_depth": 2} {"before_sent": "We introduce sentenceMIM, a probabilistic auto-encoder for language modelling , trained with Mutual Information Machine (MIM) learning . Previous attempts to learn variational auto-encoders for language data have had mixed success, with empirical performance well below state-of-the-art auto-regressive models, a key barrier being the occurrence of posterior collapsewith VAEs.", "before_sent_with_intent": " We introduce sentenceMIM, a probabilistic auto-encoder for language modelling , trained with Mutual Information Machine (MIM) learning . Previous attempts to learn variational auto-encoders for language data have had mixed success, with empirical performance well below state-of-the-art auto-regressive models, a key barrier being the occurrence of posterior collapsewith VAEs.", "after_sent": "We introduce sentenceMIM, a probabilistic auto-encoder for language modelling , trained with Mutual Information Machine (MIM) learning . Previous attempts to learn VAEs for language data have had mixed success, with empirical performance well below state-of-the-art auto-regressive models, a key barrier being the occurrence of posterior collapsewith VAEs.", "labels": "clarity", "doc_id": "2003.02645", "revision_depth": 2} {"before_sent": "We introduce sentenceMIM, a probabilistic auto-encoder for language modelling , trained with Mutual Information Machine (MIM) learning . Previous attempts to learn variational auto-encoders for language data have had mixed success, with empirical performance well below state-of-the-art auto-regressive models, a key barrier being the occurrence of posterior collapsewith VAEs. The recently proposed MIM framework encourages high mutual information between observations and latent variables, and is more robust against posterior collapse.", "before_sent_with_intent": " We introduce sentenceMIM, a probabilistic auto-encoder for language modelling , trained with Mutual Information Machine (MIM) learning . Previous attempts to learn variational auto-encoders for language data have had mixed success, with empirical performance well below state-of-the-art auto-regressive models, a key barrier being the occurrence of posterior collapsewith VAEs. The recently proposed MIM framework encourages high mutual information between observations and latent variables, and is more robust against posterior collapse.", "after_sent": "We introduce sentenceMIM, a probabilistic auto-encoder for language modelling , trained with Mutual Information Machine (MIM) learning . Previous attempts to learn variational auto-encoders for language data faced challenges due to posterior collapse. MIM learning encourages high mutual information between observations and latent variables, and is more robust against posterior collapse.", "labels": "clarity", "doc_id": "2003.02645", "revision_depth": 2} {"before_sent": " The recently proposed MIM framework encourages high mutual information between observations and latent variables, and is more robust against posterior collapse.", "before_sent_with_intent": " The recently proposed MIM framework encourages high mutual information between observations and latent variables, and is more robust against posterior collapse.", "after_sent": " The recently proposed MIM framework encourages high mutual information between observations and latent variables, and is robust against posterior collapse.", "labels": "clarity", "doc_id": "2003.02645", "revision_depth": 2} {"before_sent": " This paper formulates a MIM model for text data, along with a corresponding learning algorithm. We demonstrate excellent perplexity (PPL) results on several datasets, and show that the framework learns a rich latent space, allowing for interpolation between sentences of different lengths with a fixed-dimensional latent representation.", "before_sent_with_intent": " This paper formulates a MIM model for text data, along with a corresponding learning algorithm. We demonstrate excellent perplexity (PPL) results on several datasets, and show that the framework learns a rich latent space, allowing for interpolation between sentences of different lengths with a fixed-dimensional latent representation.", "after_sent": " As such, it learns informative representations whose dimension can be an order of magnitude higher than existing language VAEs. Importantly, the SentenceMIM loss has no hyper-parameters, simplifying optimization. We compare sentenceMIM with VAE, and AE on multiple datasets. SentenceMIM yields excellent reconstruction, comparable to AEs, with a rich structured latent space, allowing for interpolation between sentences of different lengths with a fixed-dimensional latent representation.", "labels": "meaning-changed", "doc_id": "2003.02645", "revision_depth": 2} {"before_sent": " We demonstrate excellent perplexity (PPL) results on several datasets, and show that the framework learns a rich latent space, allowing for interpolation between sentences of different lengths with a fixed-dimensional latent representation.", "before_sent_with_intent": " We demonstrate excellent perplexity (PPL) results on several datasets, and show that the framework learns a rich latent space, allowing for interpolation between sentences of different lengths with a fixed-dimensional latent representation.", "after_sent": " We demonstrate excellent perplexity (PPL) results on several datasets, and show that the framework learns a rich latent space, comparable to VAEs. The structured latent representation is demonstrated with interpolation between sentences of different lengths with a fixed-dimensional latent representation.", "labels": "meaning-changed", "doc_id": "2003.02645", "revision_depth": 2} {"before_sent": " We demonstrate excellent perplexity (PPL) results on several datasets, and show that the framework learns a rich latent space, allowing for interpolation between sentences of different lengths with a fixed-dimensional latent representation. We also demonstrate the versatility of sentenceMIM by utilizing a trained model for question-answering , a transfer learningtask , without fine-tuning .", "before_sent_with_intent": " We demonstrate excellent perplexity (PPL) results on several datasets, and show that the framework learns a rich latent space, allowing for interpolation between sentences of different lengths with a fixed-dimensional latent representation. We also demonstrate the versatility of sentenceMIM by utilizing a trained model for question-answering , a transfer learningtask , without fine-tuning .", "after_sent": " We demonstrate excellent perplexity (PPL) results on several datasets, and show that the framework learns a rich latent space, allowing for interpolation between sentences of different lengths . We demonstrate the versatility of sentenceMIM by utilizing a trained model for question-answering , a transfer learningtask , without fine-tuning .", "labels": "clarity", "doc_id": "2003.02645", "revision_depth": 2} {"before_sent": " We also demonstrate the versatility of sentenceMIM by utilizing a trained model for question-answering , a transfer learningtask , without fine-tuning .", "before_sent_with_intent": " We also demonstrate the versatility of sentenceMIM by utilizing a trained model for question-answering , a transfer learningtask , without fine-tuning .", "after_sent": " We also demonstrate the versatility of sentenceMIM by utilizing a trained model for question-answering and transfer learning , without fine-tuning .", "labels": "clarity", "doc_id": "2003.02645", "revision_depth": 2} {"before_sent": " We also demonstrate the versatility of sentenceMIM by utilizing a trained model for question-answering , a transfer learningtask , without fine-tuning . To the best of our knowledge, this is the first latent variable model (LVM) for text modelling that achieves competitive performance with non-LVM models .", "before_sent_with_intent": " We also demonstrate the versatility of sentenceMIM by utilizing a trained model for question-answering , a transfer learningtask , without fine-tuning . To the best of our knowledge, this is the first latent variable model (LVM) for text modelling that achieves competitive performance with non-LVM models .", "after_sent": " We also demonstrate the versatility of sentenceMIM by utilizing a trained model for question-answering , a transfer learningtask , without fine-tuning , outperforming VAE and AE with similar architectures .", "labels": "meaning-changed", "doc_id": "2003.02645", "revision_depth": 2} {"before_sent": "Empathetic conversational models have been shown to improve user satisfaction and task outcomes in numerous domains.", "before_sent_with_intent": " Empathetic conversational models have been shown to improve user satisfaction and task outcomes in numerous domains.", "after_sent": "Empathetic dialogue systems have been shown to improve user satisfaction and task outcomes in numerous domains.", "labels": "clarity", "doc_id": "2004.12316", "revision_depth": 1} {"before_sent": " In addition, our empirical analysis also suggests that persona plays an important role in empathetic conversations .", "before_sent_with_intent": " In addition, our empirical analysis also suggests that persona plays an important role in empathetic conversations .", "after_sent": " In addition, our empirical analysis also suggests that persona plays an important role in empathetic dialogues .", "labels": "clarity", "doc_id": "2004.12316", "revision_depth": 1} {"before_sent": " To this end, we propose a new task towards persona-based empathetic conversations and present the first empirical study on the impacts of persona on empathetic responding.", "before_sent_with_intent": " To this end, we propose a new task towards persona-based empathetic conversations and present the first empirical study on the impacts of persona on empathetic responding.", "after_sent": " To this end, we propose a new task to endow empathetic dialogue systems with personas and present the first empirical study on the impacts of persona on empathetic responding.", "labels": "clarity", "doc_id": "2004.12316", "revision_depth": 1} {"before_sent": " Specifically, we first present a novel large-scale multi-domain dataset for persona-based empathetic conversations .", "before_sent_with_intent": " Specifically, we first present a novel large-scale multi-domain dataset for persona-based empathetic conversations .", "after_sent": " Specifically, we first present a novel large-scale multi-domain dataset for empathetic dialogues with personas .", "labels": "clarity", "doc_id": "2004.12316", "revision_depth": 1} {"before_sent": " Notably, our results show that persona improves empathetic responding more when CoBERT is trained on empathetic conversations than non-empathetic ones, establishing an empirical link between persona and empathy in human conversations .", "before_sent_with_intent": " Notably, our results show that persona improves empathetic responding more when CoBERT is trained on empathetic conversations than non-empathetic ones, establishing an empirical link between persona and empathy in human conversations .", "after_sent": " Notably, our results show that persona improves empathetic responding more when CoBERT is trained on empathetic dialogues than non-empathetic ones, establishing an empirical link between persona and empathy in human conversations .", "labels": "clarity", "doc_id": "2004.12316", "revision_depth": 1} {"before_sent": " Notably, our results show that persona improves empathetic responding more when CoBERT is trained on empathetic conversations than non-empathetic ones, establishing an empirical link between persona and empathy in human conversations .", "before_sent_with_intent": " Notably, our results show that persona improves empathetic responding more when CoBERT is trained on empathetic conversations than non-empathetic ones, establishing an empirical link between persona and empathy in human conversations .", "after_sent": " Notably, our results show that persona improves empathetic responding more when CoBERT is trained on empathetic conversations than non-empathetic ones, establishing an empirical link between persona and empathy in human dialogues .", "labels": "clarity", "doc_id": "2004.12316", "revision_depth": 1} {"before_sent": "Empathetic dialogue systems have been shown to improve user satisfaction and task outcomes in numerous domains.", "before_sent_with_intent": " Empathetic dialogue systems have been shown to improve user satisfaction and task outcomes in numerous domains.", "after_sent": "Empathetic conversational models have been shown to improve user satisfaction and task outcomes in numerous domains.", "labels": "clarity", "doc_id": "2004.12316", "revision_depth": 2} {"before_sent": " In addition, our empirical analysis also suggests that persona plays an important role in empathetic dialogues .", "before_sent_with_intent": " In addition, our empirical analysis also suggests that persona plays an important role in empathetic dialogues .", "after_sent": " In addition, our empirical analysis also suggests that persona plays an important role in empathetic conversations .", "labels": "clarity", "doc_id": "2004.12316", "revision_depth": 2} {"before_sent": " To this end, we propose a new task to endow empathetic dialogue systems with personas and present the first empirical study on the impacts of persona on empathetic responding.", "before_sent_with_intent": " To this end, we propose a new task to endow empathetic dialogue systems with personas and present the first empirical study on the impacts of persona on empathetic responding.", "after_sent": " To this end, we propose a new task towards persona-based empathetic conversations and present the first empirical study on the impacts of persona on empathetic responding.", "labels": "clarity", "doc_id": "2004.12316", "revision_depth": 2} {"before_sent": " To this end, we propose a new task to endow empathetic dialogue systems with personas and present the first empirical study on the impacts of persona on empathetic responding.", "before_sent_with_intent": " To this end, we propose a new task to endow empathetic dialogue systems with personas and present the first empirical study on the impacts of persona on empathetic responding.", "after_sent": " To this end, we propose a new task to endow empathetic dialogue systems with personas and present the first empirical study on the impact of persona on empathetic responding.", "labels": "fluency", "doc_id": "2004.12316", "revision_depth": 2} {"before_sent": " Specifically, we first present a novel large-scale multi-domain dataset for empathetic dialogues with personas .", "before_sent_with_intent": " Specifically, we first present a novel large-scale multi-domain dataset for empathetic dialogues with personas .", "after_sent": " Specifically, we first present a novel large-scale multi-domain dataset for persona-based empathetic conversations .", "labels": "clarity", "doc_id": "2004.12316", "revision_depth": 2} {"before_sent": " Finally, we conduct extensive experiments to investigate the impacts of persona on empathetic responding.", "before_sent_with_intent": " Finally, we conduct extensive experiments to investigate the impacts of persona on empathetic responding.", "after_sent": " Finally, we conduct extensive experiments to investigate the impact of persona on empathetic responding.", "labels": "fluency", "doc_id": "2004.12316", "revision_depth": 2} {"before_sent": " Notably, our results show that persona improves empathetic responding more when CoBERT is trained on empathetic dialogues than non-empathetic ones, establishing an empirical link between persona and empathy in human dialogues .", "before_sent_with_intent": " Notably, our results show that persona improves empathetic responding more when CoBERT is trained on empathetic dialogues than non-empathetic ones, establishing an empirical link between persona and empathy in human dialogues .", "after_sent": " Notably, our results show that persona improves empathetic responding more when CoBERT is trained on empathetic conversations than non-empathetic ones, establishing an empirical link between persona and empathy in human dialogues .", "labels": "clarity", "doc_id": "2004.12316", "revision_depth": 2} {"before_sent": " Notably, our results show that persona improves empathetic responding more when CoBERT is trained on empathetic dialogues than non-empathetic ones, establishing an empirical link between persona and empathy in human dialogues .", "before_sent_with_intent": " Notably, our results show that persona improves empathetic responding more when CoBERT is trained on empathetic dialogues than non-empathetic ones, establishing an empirical link between persona and empathy in human dialogues .", "after_sent": " Notably, our results show that persona improves empathetic responding more when CoBERT is trained on empathetic dialogues than non-empathetic ones, establishing an empirical link between persona and empathy in human conversations .", "labels": "clarity", "doc_id": "2004.12316", "revision_depth": 2} {"before_sent": "Automatic humor detection has interesting use cases in modern technologies, such as chatbots and personal assistants. In this paper, we describe a novel approach for detecting humor in short texts using BERT sentence embedding.", "before_sent_with_intent": " Automatic humor detection has interesting use cases in modern technologies, such as chatbots and personal assistants. In this paper, we describe a novel approach for detecting humor in short texts using BERT sentence embedding.", "after_sent": "Automatic humor detection has interesting use cases in modern technologies, such as chatbots and virtual assistants. Based on the general linguistic structure of humor, in this paper, we describe a novel approach for detecting humor in short texts using BERT sentence embedding.", "labels": "meaning-changed", "doc_id": "2004.12765", "revision_depth": 1} {"before_sent": " In this paper, we describe a novel approach for detecting humor in short texts using BERT sentence embedding.", "before_sent_with_intent": " In this paper, we describe a novel approach for detecting humor in short texts using BERT sentence embedding.", "after_sent": " In this paper, we propose a novel approach for detecting humor in short texts using BERT sentence embedding.", "labels": "clarity", "doc_id": "2004.12765", "revision_depth": 1} {"before_sent": " In this paper, we describe a novel approach for detecting humor in short texts using BERT sentence embedding.", "before_sent_with_intent": " In this paper, we describe a novel approach for detecting humor in short texts using BERT sentence embedding.", "after_sent": " In this paper, we describe a novel approach for detecting humor in short texts by using BERT sentence embedding.", "labels": "fluency", "doc_id": "2004.12765", "revision_depth": 1} {"before_sent": " Our proposed model uses BERT to generate tokens and sentence embedding for texts.", "before_sent_with_intent": " Our proposed model uses BERT to generate tokens and sentence embedding for texts.", "after_sent": " Our proposed method uses BERT to generate tokens and sentence embedding for texts.", "labels": "clarity", "doc_id": "2004.12765", "revision_depth": 1} {"before_sent": " Our proposed model uses BERT to generate tokens and sentence embedding for texts. It sends embedding outputs as input to a two-layered neural networkthat predicts the target value.", "before_sent_with_intent": " Our proposed model uses BERT to generate tokens and sentence embedding for texts. It sends embedding outputs as input to a two-layered neural networkthat predicts the target value.", "after_sent": " Our proposed model uses BERT to generate embeddings for sentences of a given text and uses these embeddings as inputs for parallel lines of hidden layers in a neural network. These lines are finally concatenated to predict the target value.", "labels": "clarity", "doc_id": "2004.12765", "revision_depth": 1} {"before_sent": " For evaluation , we created a new dataset for humor detection consisting of 200k formal short texts (100k positive , 100k negative).", "before_sent_with_intent": " For evaluation , we created a new dataset for humor detection consisting of 200k formal short texts (100k positive , 100k negative).", "after_sent": " For evaluation purposes , we created a new dataset for humor detection consisting of 200k formal short texts (100k positive , 100k negative).", "labels": "clarity", "doc_id": "2004.12765", "revision_depth": 1} {"before_sent": " For evaluation , we created a new dataset for humor detection consisting of 200k formal short texts (100k positive , 100k negative).", "before_sent_with_intent": " For evaluation , we created a new dataset for humor detection consisting of 200k formal short texts (100k positive , 100k negative).", "after_sent": " For evaluation , we created a new dataset for humor detection consisting of 200k formal short texts (100k positive and 100k negative).", "labels": "fluency", "doc_id": "2004.12765", "revision_depth": 1} {"before_sent": " Experimental results show an accuracy of 98.1 percent for the proposed method , 2.1 percent improvement compared to the best CNN and RNN models and 1.1 percentbetter than a fine-tuned BERT model. In addition, the combination of RNN-CNN was not successful in this task compared to the CNN model .", "before_sent_with_intent": " Experimental results show an accuracy of 98.1 percent for the proposed method , 2.1 percent improvement compared to the best CNN and RNN models and 1.1 percentbetter than a fine-tuned BERT model. In addition, the combination of RNN-CNN was not successful in this task compared to the CNN model .", "after_sent": " Experimental results show that our proposed method can determine humor in short texts with accuracy and an F1-score of 98.2 percent. Our 8-layer model with 110M parameters outperforms all baseline models with a large margin, showing the importance of utilizing linguistic structure in machine learning models .", "labels": "clarity", "doc_id": "2004.12765", "revision_depth": 1} {"before_sent": "Arabic is a morphological rich language, posing many challenges for information extraction (IE) tasks, including Named Entity Recognition (NER) , Part-of-Speech tagging (POS), Argument Role Labeling (ARL), and Relation Extraction (RE). A few multilingual pre-trained models have been proposed and show good performance for Arabic, however, most experiment results are reported on language understanding tasks, such as natural language inference, question answering and sentiment analysis.", "before_sent_with_intent": " Arabic is a morphological rich language, posing many challenges for information extraction (IE) tasks, including Named Entity Recognition (NER) , Part-of-Speech tagging (POS), Argument Role Labeling (ARL), and Relation Extraction (RE). A few multilingual pre-trained models have been proposed and show good performance for Arabic, however, most experiment results are reported on language understanding tasks, such as natural language inference, question answering and sentiment analysis.", "after_sent": "Multilingual pre-trained models have been proposed and show good performance for Arabic, however, most experiment results are reported on language understanding tasks, such as natural language inference, question answering and sentiment analysis.", "labels": "clarity", "doc_id": "2004.14519", "revision_depth": 2} {"before_sent": " A few multilingual pre-trained models have been proposed and show good performance for Arabic, however, most experiment results are reported on language understanding tasks, such as natural language inference, question answering and sentiment analysis. Their performance on the IEtasks is less known, in particular, the cross-lingual transfer capability from English to Arabic .", "before_sent_with_intent": " A few multilingual pre-trained models have been proposed and show good performance for Arabic, however, most experiment results are reported on language understanding tasks, such as natural language inference, question answering and sentiment analysis. Their performance on the IEtasks is less known, in particular, the cross-lingual transfer capability from English to Arabic .", "after_sent": " A few multilingual pre-trained Transformers, such as mBERT (Devlin et al., 2019) and XLM-RoBERTa (Conneau et al., 2020a), have been shown to enable the effective cross-lingual transfer capability from English to Arabic .", "labels": "meaning-changed", "doc_id": "2004.14519", "revision_depth": 2} {"before_sent": " Their performance on the IEtasks is less known, in particular, the cross-lingual transfer capability from English to Arabic .", "before_sent_with_intent": " Their performance on the IEtasks is less known, in particular, the cross-lingual transfer capability from English to Arabic .", "after_sent": " Their performance on the IEtasks is less known, in particular, the cross-lingual zero-shot transfer. However, their performance on Arabic information extraction (IE) tasks is not very well studied .", "labels": "meaning-changed", "doc_id": "2004.14519", "revision_depth": 2} {"before_sent": " In this work , we pre-train a Gigaword-based bilingual language model (GigaBERT) to study these two distant languages as well as zero-short transfer learningon various IE tasks.", "before_sent_with_intent": " In this work , we pre-train a Gigaword-based bilingual language model (GigaBERT) to study these two distant languages as well as zero-short transfer learningon various IE tasks.", "after_sent": " In this paper , we pre-train a Gigaword-based bilingual language model (GigaBERT) to study these two distant languages as well as zero-short transfer learningon various IE tasks.", "labels": "clarity", "doc_id": "2004.14519", "revision_depth": 2} {"before_sent": " In this work , we pre-train a Gigaword-based bilingual language model (GigaBERT) to study these two distant languages as well as zero-short transfer learningon various IE tasks. Our GigaBERT outperforms multilingual BERT and and monolingual AraBERT on these tasks, in both supervised and zero-shot learning settings. footnote We have made our pre-trained models publicly available at URL", "before_sent_with_intent": " In this work , we pre-train a Gigaword-based bilingual language model (GigaBERT) to study these two distant languages as well as zero-short transfer learningon various IE tasks. Our GigaBERT outperforms multilingual BERT and and monolingual AraBERT on these tasks, in both supervised and zero-shot learning settings. footnote We have made our pre-trained models publicly available at URL", "after_sent": " In this work , we pre-train a customized bilingual BERT, dubbed GigaBERT, that is designed specifically for Arabic NLP and English-to-Arabic zero-shot transfer learning. We study GigaBERT's effectiveness on zero-short transfer across four IE tasks: named entity recognition, part-of-speech tagging, argument role labeling, and relation extraction. Our best model significantly outperforms mBERT, XLM-RoBERTa, and AraBERT (Antoun et al., 2020) in both the supervised and zero-shot learning settings. footnote We have made our pre-trained models publicly available at URL", "labels": "meaning-changed", "doc_id": "2004.14519", "revision_depth": 2} {"before_sent": " Our GigaBERT outperforms multilingual BERT and and monolingual AraBERT on these tasks, in both supervised and zero-shot learning settings. footnote We have made our pre-trained models publicly available at URL", "before_sent_with_intent": " Our GigaBERT outperforms multilingual BERT and and monolingual AraBERT on these tasks, in both supervised and zero-shot learning settings. footnote We have made our pre-trained models publicly available at URL", "after_sent": " Our GigaBERT outperforms multilingual BERT and and monolingual AraBERT on these tasks, in both supervised and zero-shot transfer settings. We have made our pre-trained models publicly available at URL", "labels": "clarity", "doc_id": "2004.14519", "revision_depth": 2} {"before_sent": "We propose a novel methodology for analyzing the encoding of grammatical structure in neural language models through transfer learning.", "before_sent_with_intent": " We propose a novel methodology for analyzing the encoding of grammatical structure in neural language models through transfer learning.", "after_sent": "We propose transfer learning as a method for analyzing the encoding of grammatical structure in neural language models through transfer learning.", "labels": "clarity", "doc_id": "2004.14601", "revision_depth": 1} {"before_sent": "We propose a novel methodology for analyzing the encoding of grammatical structure in neural language models through transfer learning. We test how a language model can leverage its internal representations to transfer knowledge across languages and symbol systems. We train LSTMs on non-linguistic , structured data and test their performance on human language to assess which kinds of data induce generalizable encodings that LSTMs can use for natural language.", "before_sent_with_intent": " We propose a novel methodology for analyzing the encoding of grammatical structure in neural language models through transfer learning. We test how a language model can leverage its internal representations to transfer knowledge across languages and symbol systems. We train LSTMs on non-linguistic , structured data and test their performance on human language to assess which kinds of data induce generalizable encodings that LSTMs can use for natural language.", "after_sent": "We propose a novel methodology for analyzing the encoding of grammatical structure in neural language models . We train LSTMs on non-linguistic , structured data and test their performance on human language to assess which kinds of data induce generalizable encodings that LSTMs can use for natural language.", "labels": "clarity", "doc_id": "2004.14601", "revision_depth": 1} {"before_sent": " We train LSTMs on non-linguistic , structured data and test their performance on human language to assess which kinds of data induce generalizable encodings that LSTMs can use for natural language.", "before_sent_with_intent": " We train LSTMs on non-linguistic , structured data and test their performance on human language to assess which kinds of data induce generalizable encodings that LSTMs can use for natural language.", "after_sent": " We train LSTMs on non-linguistic data and evaluate their performance on human language to assess which kinds of data induce generalizable encodings that LSTMs can use for natural language.", "labels": "clarity", "doc_id": "2004.14601", "revision_depth": 1} {"before_sent": " We train LSTMs on non-linguistic , structured data and test their performance on human language to assess which kinds of data induce generalizable encodings that LSTMs can use for natural language.", "before_sent_with_intent": " We train LSTMs on non-linguistic , structured data and test their performance on human language to assess which kinds of data induce generalizable encodings that LSTMs can use for natural language.", "after_sent": " We train LSTMs on non-linguistic , structured data and test their performance on natural language to assess which kinds of data induce generalizable encodings that LSTMs can use for natural language.", "labels": "clarity", "doc_id": "2004.14601", "revision_depth": 1} {"before_sent": " We train LSTMs on non-linguistic , structured data and test their performance on human language to assess which kinds of data induce generalizable encodings that LSTMs can use for natural language.", "before_sent_with_intent": " We train LSTMs on non-linguistic , structured data and test their performance on human language to assess which kinds of data induce generalizable encodings that LSTMs can use for natural language.", "after_sent": " We train LSTMs on non-linguistic , structured data and test their performance on human language to assess which kinds of data induce generalizable structural features that LSTMs can use for natural language.", "labels": "clarity", "doc_id": "2004.14601", "revision_depth": 1} {"before_sent": " We find that models trained on structured data such as music and Java codehave internal representations that help in modelling human language, and that, surprisingly, adding minimal amounts of structure to the training data makes a large difference in transfer to natural language . Further experiments on transfer between human languages show that zero-shot performance on a test language is highly correlated with syntactic similarity to the training language, even after removing any vocabulary overlap.", "before_sent_with_intent": " We find that models trained on structured data such as music and Java codehave internal representations that help in modelling human language, and that, surprisingly, adding minimal amounts of structure to the training data makes a large difference in transfer to natural language . Further experiments on transfer between human languages show that zero-shot performance on a test language is highly correlated with syntactic similarity to the training language, even after removing any vocabulary overlap.", "after_sent": " We find that training on non-linguistic data with latent structure (MIDI music or Java code) improves test performance on natural language, despite no overlap in surface form or vocabulary. Training on artificial languages containing recursion (hierarchical structure) also improves performance on natural language, again with no vocabulary overlap. Surprisingly, training on artificial languages consisting of sets of separated pairs of words, but with no recursion, improves performance on natural language as well as recursive languages do. Experiments on transfer between human languages show that zero-shot performance on a test language is highly correlated with syntactic similarity to the training language, even after removing any vocabulary overlap.", "labels": "meaning-changed", "doc_id": "2004.14601", "revision_depth": 1} {"before_sent": " Further experiments on transfer between human languages show that zero-shot performance on a test language is highly correlated with syntactic similarity to the training language, even after removing any vocabulary overlap.", "before_sent_with_intent": " Further experiments on transfer between human languages show that zero-shot performance on a test language is highly correlated with syntactic similarity to the training language, even after removing any vocabulary overlap.", "after_sent": " Further experiments on transfer between natural languages show that zero-shot performance on a test language is highly correlated with syntactic similarity to the training language, even after removing any vocabulary overlap.", "labels": "clarity", "doc_id": "2004.14601", "revision_depth": 1} {"before_sent": " Further experiments on transfer between human languages show that zero-shot performance on a test language is highly correlated with syntactic similarity to the training language, even after removing any vocabulary overlap.", "before_sent_with_intent": " Further experiments on transfer between human languages show that zero-shot performance on a test language is highly correlated with syntactic similarity to the training language, even after removing any vocabulary overlap.", "after_sent": " Further experiments on transfer between human languages show that zero-shot performance on a test language is highly correlated with typological syntactic similarity to the training language, even after removing any vocabulary overlap.", "labels": "meaning-changed", "doc_id": "2004.14601", "revision_depth": 1} {"before_sent": " Further experiments on transfer between human languages show that zero-shot performance on a test language is highly correlated with syntactic similarity to the training language, even after removing any vocabulary overlap. This suggests that the internal representations induced from natural languages are typologically coherent: they encode the features and differences outlined in typological studies .", "before_sent_with_intent": " Further experiments on transfer between human languages show that zero-shot performance on a test language is highly correlated with syntactic similarity to the training language, even after removing any vocabulary overlap. This suggests that the internal representations induced from natural languages are typologically coherent: they encode the features and differences outlined in typological studies .", "after_sent": " Further experiments on transfer between human languages show that zero-shot performance on a test language is highly correlated with syntactic similarity to the training language, suggesting that representations induced from natural languages are typologically coherent: they encode the features and differences outlined in typological studies .", "labels": "clarity", "doc_id": "2004.14601", "revision_depth": 1} {"before_sent": " This suggests that the internal representations induced from natural languages are typologically coherent: they encode the features and differences outlined in typological studies .", "before_sent_with_intent": "