Datasets:

feradauto

/

NLP4SGPapers

like 2

cases-etal-2019-recursive

https://aclanthology.org/N19-1365

Recursive Routing Networks: Learning to Compose Modules for Language Understanding

We introduce Recursive Routing Networks (RRNs), which are modular, adaptable models that learn effectively in diverse environments. RRNs consist of a set of functions, typically organized into a grid, and a meta-learner decision-making component called the router. The model jointly optimizes the parameters of the functions and the meta-learner's policy for routing inputs through those functions. RRNs can be incorporated into existing architectures in a number of ways; we explore adding them to word representation layers, recurrent network hidden layers, and classifier layers. Our evaluation task is natural language inference (NLI). Using the MULTINLI corpus, we show that an RRN's routing decisions reflect the high-level genre structure of that corpus. To show that RRNs can learn to specialize to more fine-grained semantic distinctions, we introduce a new corpus of NLI examples involving implicative predicates, and show that the model components become fine-tuned to the inferential signatures that are characteristic of these predicates. x Routing across examples Weight sharing Possible distribution Orthogonalized Knowledge

We thank George Supaniratisai, Arun Chaganty, Kenny Xu and Abi See for valuable discussions, and the anonymous reviewers for their useful suggestions. Clemens Rosenbaum was a recipient of an IBM PhD Fellowship while working on this publication. We acknowledge the Office of the Vice Provost for Undergraduate Education at Stanford for the summer internships for Atticus Geiger, Olivia Li and Sandhini Agarwal. This research is based in part upon work supported by the Stanford Data Science Initiative, by the NSF under Grant No. BCS-1456077, by the NSF Award IIS-1514268, and by the Air Force Research Laboratory and DARPA under agreement number FA8750-18-2-0126. The U.S. Government is authorized to reproduce and distribute reprints for Governmental purposes notwithstanding any copyright notation thereon. The views and conclusions contained herein are those of the authors and should not be interpreted as necessarily representing the official policies or endorsements, either expressed or implied, of the Air Force Research Laboratory and DARPA or the U.S. Government.

2019

rapaport-shapiro-1984-quasi

https://aclanthology.org/P84-1016

Quasi-Indexical Reference in Propositional Semantic Networks

We discuss how a deductive question-answering system can represent the beliefs or other cognitive states of users, of other (interacting) systems, and of itself.

1984

ryu-1996-argument

https://aclanthology.org/Y96-1034

Argument Structure and Unaccusativity in the Constraint-based Lexicon

This paper addresses the issue of Split Intransitivity (si) and Unaccusative Mismatches (uMs), proposing a constraint-based approach to si and ums within a recent framework of Head-driven Phrase Structure Grammar. I argue against the widely accepted dichotomous distinction of intransitive verbs, which has been advanced by the Unaccusative Hypothesis [Perlmutter (1978)]. I then propose a quadripartitive distinction of intransitive verbs on the basis of the distribution of subject argument in the semantically motivated argument structure, and show that this quadripartitive distinction allows a better understanding of si and ums. The main idea of this proposal will be summarized as the Quadripartitive Split Intransitivity Hypothesis (Qsm).

1996

liu-1995-preferred

https://aclanthology.org/Y95-1029

Preferred Clause Structure in Mandarin Spoken and Written Discourse

This paper studies the preferred clause structure in Mandarin. Tao's [I] pioneering work proposed the following "preferred clause structure in conversational Mandarin":

1995

windhouwer-2012-relcat

http://www.lrec-conf.org/proceedings/lrec2012/pdf/954_Paper.pdf

RELcat: a Relation Registry for ISOcat data categories

The ISOcat Data Category Registry contains basically a flat and easily extensible list of data category specifications. To foster reuse and standardization only very shallow relationships among data categories are stored in the registry. However, to assist crosswalks, possibly based on personal views, between various (application) domains and to overcome possible proliferation of data categories more types of ontological relationships need to be specified. RELcat is a first prototype of a Relation Registry, which allows storing arbitrary relationships. These relationships can reflect the personal view of one linguist or a larger community. The basis of the registry is a relation type taxonomy that can easily be extended. This allows on one hand to load existing sets o f relations specified in, for example, an OWL (2) ontology or SKOS taxonomy. And on the other hand allows algorithms that query the registry to traverse the stored semantic network to remain ignorant of the original source vocabulary. This paper describes first experiences with RELcat and explains some initial design decisions.

Thanks to early adaptors Matej Durco (SMC4LRT), Irina Nevskaya (RELISH) and Ineke Schuurman (CLARIN-NL/VL) for driving this first version of RELcat forward.

2012

strobelt-etal-2021-lmdiff

https://aclanthology.org/2021.emnlp-demo.12

LMdiff: A Visual Diff Tool to Compare Language Models

While different language models are ubiquitous in NLP, it is hard to contrast their outputs and identify which contexts one can handle better than the other. To address this question, we introduce LMDIFF, a tool that visually compares probability distributions of two models that differ, e.g., through finetuning, distillation, or simply training with different parameter sizes. LMDIFF allows the generation of hypotheses about model behavior by investigating text instances token by token and further assists in choosing these interesting text instances by identifying the most interesting phrases from large corpora. We showcase the applicability of LMDIFF for hypothesis generation across multiple case studies. A demo is available at http://lmdiff.net.

We thank Ankur Parikh and Ian Tenney for helpful comments on an earlier draft of this paper. This work was supported by the MIT-IBM Watson AI Lab. This work has been developed in part during the BigScience Summer of Language Models 2021.

2021

taghipour-ng-2016-neural

https://aclanthology.org/D16-1193

A Neural Approach to Automated Essay Scoring

Traditional automated essay scoring systems rely on carefully designed features to evaluate and score essays. The performance of such systems is tightly bound to the quality of the underlying features. However, it is laborious to manually design the most informative features for such a system. In this paper, we develop an approach based on recurrent neural networks to learn the relation between an essay and its assigned score, without any feature engineering. We explore several neural network models for the task of automated essay scoring and perform some analysis to get some insights of the models. The results show that our best system, which is based on long short-term memory networks, outperforms a strong baseline by 5.6% in terms of quadratic weighted Kappa, without requiring any feature engineering.

Quality Education

This research is supported by Singapore Ministry of Education Academic Research Fund Tier 2 grant MOE2013-T2-1-150. We are also grateful to the anonymous reviewers for their helpful comments.

2016

papageorgiou-etal-2000-unified

http://www.lrec-conf.org/proceedings/lrec2000/pdf/181.pdf

A Unified POS Tagging Architecture and its Application to Greek

This paper proposes a flexible and unified tagging architecture that could be incorporated into a number of applications like information extraction, cross-language information retrieval, term extraction, or summarization, while providing an essential component for subsequent syntactic processing or lexicographical work. A feature-based multi-tiered approach (FBT tagger) is introduced to part-of-speech tagging. FBT is a variant of the well-known transformation based learning paradigm aiming at improving the quality of tagging highly inflective languages such as Greek. Additionally, a large experiment concerning the Greek language is conducted and results are presented for a variety of text genres, including financial reports, newswires, press releases and technical manuals. Finally, the adopted evaluation methodology is discussed.

2000

reiter-etal-2008-resource

https://aclanthology.org/W08-2231

A Resource-Poor Approach for Linking Ontology Classes to Wikipedia Articles

The applicability of ontologies for natural language processing depends on the ability to link ontological concepts and relations to their realisations in texts. We present a general, resource-poor account to create such a linking automatically by extracting Wikipedia articles corresponding to ontology classes. We evaluate our approach in an experiment with the Music Ontology. We consider linking as a promising starting point for subsequent steps of information extraction.

Acknowledgements. We kindly thank our annotators for their effort and Rüdiger Wolf for technical support.

2008

saleh-etal-2014-study

https://aclanthology.org/C14-1020

A Study of using Syntactic and Semantic Structures for Concept Segmentation and Labeling

This paper presents an empirical study on using syntactic and semantic information for Concept Segmentation and Labeling (CSL), a well-known component in spoken language understanding. Our approach is based on reranking N-best outputs from a state-of-the-art CSL parser. We perform extensive experimentation by comparing different tree-based kernels with a variety of representations of the available linguistic information, including semantic concepts, words, POS tags, shallow and full syntax, and discourse trees. The results show that the structured representation with the semantic concepts yields significant improvement over the base CSL parser, much larger compared to learning with an explicit feature vector representation. We also show that shallow syntax helps improve the results and that discourse relations can be partially beneficial.

This research is developed by the Arabic Language Technologies (ALT) group at Qatar Computing Research Institute (QCRI) within the Qatar Foundation in collaboration with MIT. It is part of the Interactive sYstems for Answer Search (Iyas) project.

2014

malmasi-etal-2015-norwegian

https://aclanthology.org/R15-1053

Norwegian Native Language Identification

We present a study of Native Language Identification (NLI) using data from learners of Norwegian, a language not yet used for this task. NLI is the task of predicting a writer's first language using only their writings in a learned language. We find that three feature types, function words, part-of-speech n-grams and a hybrid part-of-speech/function word mixture n-gram model are useful here. Our system achieves an accuracy of 79% against a baseline of 13% for predicting an author's L1. The same features can distinguish non-native writing with 99% accuracy. We also find that part-of-speech n-gram performance on this data deviates from previous NLI results, possibly due to the use of manually post-corrected tags.

We would like to thank Kari Tenfjord and Paul Meurer for providing access to the ASK corpus and their assistance in using the data.

2015

siblini-etal-2021-towards

https://aclanthology.org/2021.acl-short.130

Towards a more Robust Evaluation for Conversational Question Answering

With the explosion of chatbot applications, Conversational Question Answering (CQA) has generated a lot of interest in recent years. Among proposals, reading comprehension models which take advantage of the conversation history (previous QA) seem to answer better than those which only consider the current question. Nevertheless, we note that the CQA evaluation protocol has a major limitation. In particular, models are allowed, at each turn of the conversation, to access the ground truth answers of the previous turns. Not only does this severely prevent their applications in fully autonomous chatbots, it also leads to unsuspected biases in their behavior. In this paper, we highlight this effect and propose new tools for evaluation and training in order to guard against the noted issues. The new results that we bring come to reinforce methods of the current state of the art.

2021

dudy-etal-2018-multi

https://aclanthology.org/W18-1210

A Multi-Context Character Prediction Model for a Brain-Computer Interface

Brain-computer interfaces and other augmentative and alternative communication devices introduce language-modeing challenges distinct from other character-entry methods. In particular, the acquired signal of the EEG (electroencephalogram) signal is noisier, which, in turn, makes the user intent harder to decipher. In order to adapt to this condition, we propose to maintain ambiguous history for every time step, and to employ, apart from the character language model, word information to produce a more robust prediction system. We present preliminary results that compare this proposed Online-Context Language Model (OCLM) to current algorithms that are used in this type of setting. Evaluations on both perplexity and predictive accuracy demonstrate promising results when dealing with ambiguous histories in order to provide to the front end a distribution of the next character the user might type.

We would like to thank the reviewers of the SCLeM workshop for their insightful comments and feedback. We also would like to thank Brian Roark for his helpful advice, as well as our clinical team in the Institute on Development & Disability at OHSU. Research reported in this paper was supported the National Institute on Deafness and Other Communication Disorders of the NIH under award number 5R01DC009834-09. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

2018

danieli-etal-2004-evaluation

http://www.lrec-conf.org/proceedings/lrec2004/pdf/371.pdf

Evaluation of Consensus on the Annotation of Prosodic Breaks in the Romance Corpus of Spontaneous Speech ``C-ORAL-ROM''

CORAL -ROM, Integrated Reference Corpora For Spoken Romance Languages, is a multilingual corpus of spontaneous speech delivered within the IST Program. Corpora are tagged with respect to terminal and non terminal prosodic breaks. Terminal breaks are considered the most perceptively relevant cues to determine the utterance boundaries in spontaneous speech resources. The paper presents the evaluation of the inter-annotator agreement accomplished by an institution external to the consortium and shows the level of reliability of the tagging delivered and the annotation scheme adopted. The data show, at cross-linguistic level, a very high K coefficient (between 7.7 and 9.2, according to the language resource). A strong level of agreement specifically for terminal breaks has also been recorded. The data thus show that the annotation of the utterances identified in terms of their prosodic breaks is able to capture relevant perceptual facts, and it appears that the proposed coding scheme can be applied in a highly replicable way.

2004

quirk-etal-2015-language

https://aclanthology.org/P15-1085

Language to Code: Learning Semantic Parsers for If-This-Then-That Recipes

Using natural language to write programs is a touchstone problem for computational linguistics. We present an approach that learns to map natural-language descriptions of simple "if-then" rules to executable code. By training and testing on a large corpus of naturally-occurring programs (called "recipes") and their natural language descriptions, we demonstrate the ability to effectively map language to code. We compare a number of semantic parsing approaches on the highly noisy training data collected from ordinary users, and find that loosely synchronous systems perform best.

The authors would like to thank William Dolan and the anonymous reviewers for their helpful advice and suggestions.

2015

sakaji-etal-2019-financial

https://aclanthology.org/W19-5507

Financial Text Data Analytics Framework for Business Confidence Indices and Inter-Industry Relations

In this paper, we propose a novel framework for analyzing inter-industry relations using the contact histories of local banks. Contact histories are data recorded when employees communicate with customers. By analyzing contact histories, we can determine business confidence levels in the local region and analyze inter-industry relations using industrial data that is attached to the contact history. However, it is often difficult for bankers to create analysis programs. Therefore, we propose a banker-friendly inter-industry relations analysis framework. In this study, we generated regional business confidence indices and used them to analyze inter-industry relations.

2019

wu-etal-2021-newsbert-distilling

https://aclanthology.org/2021.findings-emnlp.280

NewsBERT: Distilling Pre-trained Language Model for Intelligent News Application

Pre-trained language models (PLMs) like BERT have made great progress in NLP. News articles usually contain rich textual information, and PLMs have the potentials to enhance news text modeling for various intelligent news applications like news recommendation and retrieval. However, most existing PLMs are in huge size with hundreds of millions of parameters. Many online news applications need to serve millions of users with low latency, which poses great challenge to incorporating PLMs in these scenarios. Knowledge distillation techniques can compress a large PLM into a much smaller one and meanwhile keep good performance. However, existing language models are pre-trained and distilled on general corpus like Wikipedia, which have gaps with the news domain and may be suboptimal for news intelligence. In this paper, we propose NewsBERT, which can distill PLMs for efficient and effective news intelligence. In our approach, we design a teacher-student joint learning and distillation framework to collaboratively learn both teacher and student models, where the student model can learn from the learning experience of the teacher model. In addition, we propose a momentum distillation method by incorporating the gradients of teacher model into the update of student model to better transfer the knowledge learned by the teacher model. Thorough experiments on two real-world datasets with three tasks show that NewsBERT can empower various intelligent news applications with much smaller models.

Peace, Justice and Strong Institutions

This work was supported by the National Natural Science Foundation of China under Grant numbers 82090053, 61862002, and Tsinghua-Toyota Research Funds 20213930033.

2021

bengler-2000-automotive

http://www.lrec-conf.org/proceedings/lrec2000/pdf/312.pdf

Automotive Speech-Recognition - Success Conditions Beyond Recognition Rates

From a car-manufacturer's point of view it is very important to integrate evaluation procedures into the MMI development process. Focusing the usability evaluation of speech-input and speech-output systems aspects beyond recognition rates must be fulfilled. Two of these conditions will be discussed based upon user studies conducted in 1999: • Mental-workload and distraction • Learnability

2000

lin-2002-web

http://www.lrec-conf.org/proceedings/lrec2002/pdf/85.pdf

The Web as a Resource for Question Answering: Perspectives and Challenges

The vast amounts of information readily available on the World Wide Web can be effectively used for question answering in two fundamentally different ways. In the federated approach, techniques for handling semistructured data are applied to access Web sources as if they were databases, allowing large classes of common questions to be answered uniformly. In the distributed approach, largescale text-processing techniques are used to extract answers directly from unstructured Web documents. Because the Web is orders of magnitude larger than any human-collected corpus, question answering systems can capitalize on its unparalleled-levels of data redundancy. Analysis of real-world user questions reveals that the federated and distributed approaches complement each other nicely, suggesting a hybrid approach in future question answering systems.

I'd like to thank Boris Katz, Greg Marton, and Vineet Sinha for their helpful comments on earlier drafts.

2002

schulte-im-walde-2006-experiments

https://aclanthology.org/J06-2001

Experiments on the Automatic Induction of German Semantic Verb Classes

This article presents clustering experiments on German verbs: A statistical grammar model for German serves as the source for a distributional verb description at the lexical syntax-semantics interface, and the unsupervised clustering algorithm k-means uses the empirical verb properties to perform an automatic induction of verb classes. Various evaluation measures are applied to compare the clustering results to gold standard German semantic verb classes under different criteria. The primary goals of the experiments are (1) to empirically utilize and investigate the well-established relationship between verb meaning and verb behavior within a cluster analysis and (2) to investigate the required technical parameters of a cluster analysis with respect to this specific linguistic task. The clustering methodology is developed on a small-scale verb set and then applied to a larger-scale verb set including 883 German verbs.

The work reported here was performed while the author was a member of the DFG-funded PhD program "Graduiertenkolleg" Sprachliche Repräsentationen und ihre Interpretation at the Institute for Natural Language Processing (IMS), University of Stuttgart, Germany. Many thanks to Helmut Schmid, Stefan Evert, Frank Keller, Scott McDonald, Alissa Melinger, Chris Brew, Hinrich Schütze, Jonas Kuhn, and the two anonymous reviewers for their valuable comments on previous versions of this article.

2006

turchi-etal-2014-adaptive

https://aclanthology.org/P14-1067

Adaptive Quality Estimation for Machine Translation

The automatic estimation of machine translation (MT) output quality is a hard task in which the selection of the appropriate algorithm and the most predictive features over reasonably sized training sets plays a crucial role. When moving from controlled lab evaluations to real-life scenarios the task becomes even harder. For current MT quality estimation (QE) systems, additional complexity comes from the difficulty to model user and domain changes. Indeed, the instability of the systems with respect to data coming from different distributions calls for adaptive solutions that react to new operating conditions. To tackle this issue we propose an online framework for adaptive QE that targets reactivity and robustness to user and domain changes. Contrastive experiments in different testing conditions involving user and domain changes demonstrate the effectiveness of our approach.

This work has been partially supported by the ECfunded project MateCat (ICT-2011.4.2-287688).

2014

kongthon-etal-2011-semantic

https://aclanthology.org/W11-3106

A Semantic Based Question Answering System for Thailand Tourism Information

This paper reports our ongoing research work to create a semantic based question answering system for Thailand tourism information. Our proposed system focuses on mapping expressions in Thai natural language into ontology query language (SPARQL).

2011

schatzmann-etal-2007-agenda

https://aclanthology.org/N07-2038

Agenda-Based User Simulation for Bootstrapping a POMDP Dialogue System

This paper investigates the problem of bootstrapping a statistical dialogue manager without access to training data and proposes a new probabilistic agenda-based method for simulating user behaviour. In experiments with a statistical POMDP dialogue system, the simulator was realistic enough to successfully test the prototype system and train a dialogue policy. An extensive study with human subjects showed that the learned policy was highly competitive, with task completion rates above 90%.

2007

hazem-hernandez-2019-meta

https://aclanthology.org/R19-1055

Meta-Embedding Sentence Representation for Textual Similarity

Word embedding models are now widely used in most NLP applications. Despite their effectiveness, there is no clear evidence about the choice of the most appropriate model. It often depends on the nature of the task and on the quality and size of the used data sets. This remains true for bottom-up sentence embedding models. However, no straightforward investigation has been conducted so far. In this paper, we propose a systematic study of the impact of the main word embedding models on sentence representation. By contrasting in-domain and pre-trained embedding models, we show under which conditions they can be jointly used for bottom-up sentence embeddings. Finally, we propose the first bottom-up meta-embedding representation at the sentence level for textual similarity. Significant improvements are observed in several tasks including question-to-question similarity, paraphrasing and next utterance ranking.

2019

carlberger-etal-2001-improving

https://aclanthology.org/W01-1703

Improving Precision in Information Retrieval for Swedish using Stemming

We will in this paper present an evaluation 1 of how much stemming improves precision in information retrieval for Swedish texts. To perform this, we built an information retrieval tool with optional stemming and created a tagged corpus in Swedish. We know that stemming in information retrieval for English, Dutch and Slovenian gives better precision the more inflecting the language is, but precision depends also on query length and document length. Our final results were that stemming improved both precision and recall with 15 respectively 18 percent for Swedish texts having an average length of 181 words.

We would like to thank the search engine team and specifically Jesper Ekhall at Euroseek AB for their support with the integration of our stemming algorithms in their search engine and allowing us to use their search engine in our experiments.

2001

dredze-crammer-2008-active

https://aclanthology.org/P08-2059

Active Learning with Confidence

Active learning is a machine learning approach to achieving high-accuracy with a small amount of labels by letting the learning algorithm choose instances to be labeled. Most of previous approaches based on discriminative learning use the margin for choosing instances. We present a method for incorporating confidence into the margin by using a newly introduced online learning algorithm and show empirically that confidence improves active learning.

2008

ji-etal-2020-span

https://aclanthology.org/2020.coling-main.8

Span-based Joint Entity and Relation Extraction with Attention-based Span-specific and Contextual Semantic Representations

Span-based joint extraction models have shown their efficiency on entity recognition and relation extraction. These models regard text spans as candidate entities and span tuples as candidate relation tuples. Span semantic representations are shared in both entity recognition and relation extraction, while existing models cannot well capture semantics of these candidate entities and relations. To address these problems, we introduce a span-based joint extraction framework with attention-based semantic representations. Specially, attentions are utilized to calculate semantic representations, including span-specific and contextual ones. We further investigate effects of four attention variants in generating contextual semantic representations. Experiments show that our model outperforms previous systems and achieves state-of-the-art results on ACE2005, CoNLL2004 and ADE.

The work is supported by the National Key Research and Development Program of China (2018YFB1004502) and the National Natural Science Foundation of China (61532001).

2020

poelitz-bartz-2014-enhancing

https://aclanthology.org/W14-0606

Enhancing the possibilities of corpus-based investigations: Word sense disambiguation on query results of large text corpora

Common large digital text corpora do not distinguish between different meanings of word forms, intense manual effort has to be done for disambiguation tasks when querying for homonyms or polysemes. To improve this situation, we ran experiments with automatic word sense disambiguation methods operating directly on the output of the corpus query. In this paper, we present experiments with topic models to cluster search result snippets in order to separate occurrences of homonymous or polysemous queried words by their meanings.

2014

rondeau-hazen-2018-systematic

https://aclanthology.org/W18-2602

Systematic Error Analysis of the Stanford Question Answering Dataset

We analyzed the outputs of multiple question answering (QA) models applied to the Stanford Question Answering Dataset (SQuAD) to identify the core challenges for QA systems on this data set. Through an iterative process, challenging aspects were hypothesized through qualitative analysis of the common error cases. A classifier was then constructed to predict whether SQuAD test examples were likely to be difficult for systems to answer based on features associated with the hypothesized aspects. The classifier's performance was used to accept or reject each aspect as an indicator of difficulty. With this approach, we ensured that our hypotheses were systematically tested and not simply accepted based on our pre-existing biases. Our explanations are not accepted based on human evaluation of individual examples. This process also enabled us to identify the primary QA strategy learned by the models, i.e., systems determined the acceptable answer type for a question and then selected the acceptable answer span of that type containing the highest density of words present in the question within its local vicinity in the passage.

We would like to thank Eric Lin, Peter Potash, Yadollah Yaghoobzadeh, and Kaheer Suleman for their feedback and helpful comments. We also thanks the anonymous reviewers for their comments.

2018

wang-etal-2021-fine-grained

https://aclanthology.org/2021.findings-emnlp.9

Fine-grained Semantic Alignment Network for Weakly Supervised Temporal Language Grounding

Temporal language grounding (TLG) aims to localize a video segment in an untrimmed video based on a natural language description. To alleviate the expensive cost of manual annotations for temporal boundary labels, we are dedicated to the weakly supervised setting, where only video-level descriptions are provided for training. Most of the existing weakly supervised methods generate a candidate segment set and learn cross-modal alignment through a MIL-based framework. However, the temporal structure of the video as well as the complicated semantics in the sentence are lost during the learning. In this work, we propose a novel candidatefree framework: Fine-grained Semantic Alignment Network (FSAN), for weakly supervised TLG. Instead of view the sentence and candidate moments as a whole, FSAN learns token-by-clip cross-modal semantic alignment by an iterative cross-modal interaction module, generates a fine-grained cross-modal semantic alignment map, and performs grounding directly on top of the map. Extensive experiments are conducted on two widelyused benchmarks: ActivityNet-Captions, and DiDeMo, where our FSAN achieves state-ofthe-art performance.

This work was supported by the National Natural Science Foundation of China under Contract 61632019.

2021

zeng-etal-2021-gene

https://aclanthology.org/2021.textgraphs-1.5

GENE: Global Event Network Embedding

Current methods for event representation ignore related events in a corpus-level global context. For a deep and comprehensive understanding of complex events, we introduce a new task, Event Network Embedding, which aims to represent events by capturing the connections among events. We propose a novel framework, Global Event Network Embedding (GENE), that encodes the event network with a multi-view graph encoder while preserving the graph topology and node semantics. The graph encoder is trained by minimizing both structural and semantic losses. We develop a new series of structured probing tasks, and show that our approach effectively outperforms baseline models on node typing, argument role classification, and event coreference resolution. 1

This research is based upon work supported in part by U.S. DARPA KAIROS Program No. FA8750-19-2-1004, U.S. DARPA AIDA Program No. FA8750-18-2-0014, Air Force No. FA8650-17-C-7715. The views and conclusions contained herein are those of the authors and should not be interpreted as necessarily representing the official policies, either expressed or implied, of DARPA, or the U.S. Government. The U.S. Government is authorized to reproduce and distribute reprints for governmental purposes notwithstanding any copyright annotation therein.

2021

bagherbeygi-shamsfard-2012-corpus

http://www.lrec-conf.org/proceedings/lrec2012/pdf/1013_Paper.pdf

Corpus based Semi-Automatic Extraction of Persian Compound Verbs and their Relations

Nowadays, Wordnet is used in natural language processing as one of the major linguistic resources. Having such a resource for Persian language helps researchers in computational linguistics and natural language processing fields to develop more accurate systems with higher performances. In this research, we propose a model for semi-automatic construction of Persian wordnet of verbs. Compound verbs are a very productive structure in Persian and number of compound verbs is much greater than simple verbs in this language This research is aimed at finding the structure of Persian compound verbs and the relations between verb components. The main idea behind developing this system is using the wordnet of other POS categories (here means noun and adjective) to extract Persian compound verbs, their synsets and their relations. This paper focuses on three main tasks: 1.extracting compound verbs 2.extracting verbal synsets and 3.extracting the relations among verbal synsets such as hypernymy, antonymy and cause.

2012

geertzen-etal-2007-multidimensional

https://aclanthology.org/2007.sigdial-1.26

A Multidimensional Approach to Utterance Segmentation and Dialogue Act Classification

In this paper we present a multidimensional approach to utterance segmentation and automatic dialogue act classif cation. We show that the use of multiple dimensions in distinguishing and annotating units not only supports a more accurate analysis of human communication, but can also help to solve some notorious problems concerning the segmentation of dialogue into functional units. We introduce the use of per-dimension segmentation for dialogue act taxonomies that feature multi-functionality and show that better classif cation results are obtained when using a separate segmentation for each dimension than when using one segmentation that f ts all dimensions. Three machine learning techniques are applied and compared on the task of automatic classifcation of multiple communicative functions of utterances. The results are encouraging and indicate that communicative functions in important dimensions are easy machinelearnable.

2007

zanzotto-etal-2006-discovering

https://aclanthology.org/P06-1107

Discovering Asymmetric Entailment Relations between Verbs Using Selectional Preferences

In this paper we investigate a novel method to detect asymmetric entailment relations between verbs. Our starting point is the idea that some point-wise verb selectional preferences carry relevant semantic information. Experiments using Word-Net as a gold standard show promising results. Where applicable, our method, used in combination with other approaches, significantly increases the performance of entailment detection. A combined approach including our model improves the AROC of 5% absolute points with respect to standard models.

2006

guthrie-etal-2008-unsupervised

http://www.lrec-conf.org/proceedings/lrec2008/pdf/866_paper.pdf

An Unsupervised Probabilistic Approach for the Detection of Outliers in Corpora

Many applications of computational linguistics are greatly influenced by the quality of corpora available and as automatically generated corpora continue to play an increasingly common role, it is essential that we not overlook the importance of well-constructed and homogeneous corpora. This paper describes an automatic approach to improving the homogeneity of corpora using an unsupervised method of statistical outlier detection to find documents and segments that do not belong in a corpus. We consider collections of corpora that are homogeneous with respect to topic (i.e. about the same subject), or genre (written for the same audience or from the same source) and use a combination of stylistic and lexical features of the texts to automatically identify pieces of text in these collections that break the homogeneity. These pieces of text that are significantly different from the rest of the corpus are likely to be errors that are out of place and should be removed from the corpus before it is used for other tasks. We evaluate our techniques by running extensive experiments over large artificially constructed corpora that each contain single pieces of text from a different topic, author, or genre than the rest of the collection and measure the accuracy of identifying these pieces of text without the use of training data. We show that when these pieces of text are reasonably large (1,000 words) we can reliably identify them in a corpus.

2008

al-natsheh-etal-2017-udl

https://aclanthology.org/S17-2013

UdL at SemEval-2017 Task 1: Semantic Textual Similarity Estimation of English Sentence Pairs Using Regression Model over Pairwise Features

This paper describes the model UdL we proposed to solve the semantic textual similarity task of SemEval 2017 workshop. The track we participated in was estimating the semantics relatedness of a given set of sentence pairs in English. The best run out of three submitted runs of our model achieved a Pearson correlation score of 0.8004 compared to a hidden human annotation of 250 pairs. We used random forest ensemble learning to map an expandable set of extracted pairwise features into a semantic similarity estimated value bounded between 0 and 5. Most of these features were calculated using word embedding vectors similarity to align Part of Speech (PoS) and Name Entities (NE) tagged tokens of each sentence pair. Among other pairwise features, we experimented a classical tf-idf weighted Bag of Words (BoW) vector model but with character-based range of n-grams instead of words. This sentence vector BoW-based feature gave a relatively high importance value percentage in the feature importances analysis of the ensemble learning.

We would like to thank ARC6 Auvergne-Rhône-Alpes that funds the current PhD studies of the first author and the program "Investissements d'Avenir" ISTEX for funding the post-doctoral position of the second author.

2017

braschler-etal-2000-evaluation

http://www.lrec-conf.org/proceedings/lrec2000/pdf/70.pdf

The Evaluation of Systems for Cross-language Information Retrieval

We describe the creation of an infrastructure for the testing of cross-language text retrieval systems within the context of the Text REtrieval Conferences (TREC) organised by the US National Institute of Standards and Technology (NIST). The approach adopted and the issues that had to be taken into consideration when building a multilingual test suite and developing appropriate evaluation procedures to test cross-language systems are described. From 2000 on, a cross-language evaluation activity for European languages known as CLEF (Cross-Language Evaluation Forum) will be coordinated in Europe, while TREC will focus on Asian languages. The implications of the move to Europe and the intentions for the future are discussed.

We gratefully acknowledge the support of all the data providers and copyright holders, and in particular: Newswires: Associated Press, USA; SDA -Schweizerische Depeschenagentur, Switzerland.

2000

oneill-mctear-1999-object

https://aclanthology.org/E99-1004

An Object-Oriented Approach to the Design of Dialogue Management Functionality

Dialogues may be seen as comprising commonplace routines on the one hand and specialized, task-specific interactions on the other. Object-orientation is an established means of separating the generic from the specialized. The system under discussion combines this objectoriented approach with a self-organizing, mixed-initiative dialogue strategy, raising the possibility of dialogue systems that can be assembled from ready-made components and tailored, specialized components.

1999

roemmele-etal-2021-answerquest

https://aclanthology.org/2021.eacl-demos.6

AnswerQuest: A System for Generating Question-Answer Items from Multi-Paragraph Documents

One strategy for facilitating reading comprehension is to present information in a questionand-answer format. We demo a system that integrates the tasks of question answering (QA) and question generation (QG) in order to produce Q&A items that convey the content of multi-paragraph documents. We report some experiments for QA and QG that yield improvements on both tasks, and assess how they interact to produce a list of Q&A items for a text. The demo is accessible at qna.sdl.com.

2021

ballard-tinkham-1984-phrase

https://aclanthology.org/J84-2001

A Phrase-Structured Grammatical Framework for Transportable Natural Language Processing

We present methods of dealing with the syntactic problems that arise in the construction of natural language processors that seek to allow users, as opposed to computational linguists, to customize an interface to operate with a new domain of data. In particular, we describe a grammatical formalism, based on augmented phrase-structure rules, which allows a parser to perform many important domain-specific disambiguations by reference to a pre-defined grammar and a collection of auxiliary files produced during an initial knowledge acquisition session with the user. We illustrate the workings of this formalism with examples from the grammar developed for our Layered Domain Class (LDC) system, though similarly motivated systems ought also to benefit from our formalisms. In addition to showing the theoretical advantage of providing many of the fine-tuning capabilities of so-called semantic grammars within the context of a domain-independent grammar, we demonstrate several practical benefits to our approach. The results of three experiments with our grammar and parser are also given.

1984

martschat-etal-2015-analyzing

https://aclanthology.org/N15-3002

Analyzing and Visualizing Coreference Resolution Errors

We present a toolkit for coreference resolution error analysis. It implements a recently proposed analysis framework and contains rich components for analyzing and visualizing recall and precision errors.

This work has been funded by the Klaus Tschira Foundation, Heidelberg, Germany. The first author has been supported by a HITS PhD scholarship.

2015

hoffman-etal-1963-application

https://aclanthology.org/1963.earlymt-1.16

Application of decision tables to syntactic analysis

1963

beltagy-etal-2019-scibert

https://aclanthology.org/D19-1371

SciBERT: A Pretrained Language Model for Scientific Text

Obtaining large-scale annotated data for NLP tasks in the scientific domain is challenging and expensive. We release SCIBERT, a pretrained language model based on BERT (Devlin et al., 2019) to address the lack of highquality, large-scale labeled scientific data. SCIBERT leverages unsupervised pretraining on a large multi-domain corpus of scientific publications to improve performance on downstream scientific NLP tasks. We evaluate on a suite of tasks including sequence tagging, sentence classification and dependency parsing, with datasets from a variety of scientific domains. We demonstrate statistically significant improvements over BERT and achieve new state-of-the-art results on several of these tasks. The code and pretrained models are available at https://github. com/allenai/scibert/.

Industry, Innovation and Infrastructure

We thank the anonymous reviewers for their comments and suggestions. We also thank Waleed Ammar, Noah Smith, Yoav Goldberg, Daniel King, Doug Downey, and Dan Weld for their helpful discussions and feedback. All experiments were performed on beaker.org and supported in part by credits from Google Cloud.

2019

fillwock-traum-2018-identification

https://aclanthology.org/L18-1629

Identification of Personal Information Shared in Chat-Oriented Dialogue

We present an analysis of how personal information is shared in chat-oriented dialogue. We develop an annotation scheme, including entity-types, attributes, and values, that can be used to annotate the presence and type of personal information in these dialogues. A collection of attribute types is identified from the annotation of three chat-oriented dialogue corpora and a taxonomy of personal information pertinent to chat-oriented dialogue is presented. We examine similarities and differences in the frequency of specific attributes in the three corpora and observe that there is much overlap between the attribute types which are shared between dialogue participants in these different settings. The work presented here suggests that there is a common set of attribute types that frequently occur within chat-oriented dialogue in general. This resource can be used in the development of chat-oriented dialogue systems by providing common topics that a dialogue system should be able to talk about.

2018

keesing-etal-2020-convolutional

https://aclanthology.org/2020.alta-1.13

Convolutional and Recurrent Neural Networks for Spoken Emotion Recognition

We test four models proposed in the speech emotion recognition (SER) literature on 15 public and academic licensed datasets in speaker-independent cross-validation. Results indicate differences in the performance of the models which is partly dependent on the dataset and features used. We also show that a standard utterance-level feature set still performs competitively with neural models on some datasets. This work serves as a starting point for future model comparisons, in addition to open-sourcing the testing code.

The authors would like to thank the University of Auckland for funding this research through a PhD scholarship. We would like to thank in particular the School of Computer Science for providing the computer hardware to train and test these models. We would also like to thank the anonymous reviewers who submitted helpful feedback on this paper.

2020

kunilovskaya-etal-2021-fiction

https://aclanthology.org/2021.ranlp-1.84

Fiction in Russian Translation: A Translationese Study

This paper presents a translationese study based on the parallel data from the Russian National Corpus (RNC). We explored differences between literary texts originally authored in Russian and fiction translated into Russian from 11 languages. The texts are represented with frequency-based features that capture structural and lexical properties of language. Binary classification results indicate that literary translations can be distinguished from non-translations with an accuracy ranging from 82 to 92% depending on the source language and feature set. Multiclass classification confirms that translations from distant languages are more distinct from non-translations than translations from languages that are typologically close to Russian. It also demonstrates that translations from same-family source languages share translationese properties. Structural features return more consistent results than features relying on external resources and capturing lexical properties of texts in both translationese detection and source language identification tasks.

2021

kim-etal-2021-mostly

https://aclanthology.org/2021.naloma-1.9

A (Mostly) Symbolic System for Monotonic Inference with Unscoped Episodic Logical Forms

We implement the formalization of natural logic-like monotonic inference using Unscoped Episodic Logical Forms (ULFs) by Kim et al. (2020). We demonstrate this system's capacity to handle a variety of challenging semantic phenomena using the FraCaS dataset (Cooper et al., 1996).These results give empirical evidence for prior claims that ULF is an appropriate representation to mediate natural logic-like inferences. 1

This work was supported by NSF EAGER grant NSF IIS-1908595, DARPA CwC subcontract W911NF-15-1-0542, and a Sproull Graduate Fellowship from the University of Rochester. We are grateful to the anonymous reviewers for their helpful feedback.

2021

litman-etal-2006-characterizing

https://aclanthology.org/J06-3004

Characterizing and Predicting Corrections in Spoken Dialogue Systems

This article focuses on the analysis and prediction of corrections, defined as turns where a user tries to correct a prior error made by a spoken dialogue system. We describe our labeling procedure of various corrections types and statistical analyses of their features in a corpus collected from a train information spoken dialogue system. We then present results of machinelearning experiments designed to identify user corrections of speech recognition errors. We investigate the predictive power of features automatically computable from the prosody of the turn, the speech recognition process, experimental conditions, and the dialogue history. Our best-performing features reduce classification error from baselines of 25.70-28.99% to 15.72%.

Marc Swerts is also affiliated with the University of Antwerp. His research is sponsored by the Netherlands Organisation for Scientific Research (NWO). This work was performed when the authors were at AT&T Labs-Research.

2006

belz-2005-statistical

https://aclanthology.org/W05-1601

Statistical Generation: Three Methods Compared and Evaluated

Statistical NLG has largely meant n-gram modelling which has the considerable advantages of lending robustness to NLG systems, and of making automatic adaptation to new domains from raw corpora possible. On the downside, n-gram models are expensive to use as selection mechanisms and have a built-in bias towards shorter realisations. This paper looks at treebank-training of generators, an alternative method for building statistical models for NLG from raw corpora, and two different ways of using treebank-trained models during generation. Results show that the treebank-trained generators achieve improvements similar to a 2-gram generator over a baseline of random selection. However, the treebank-trained generators achieve this at a much lower cost than the 2-gram generator, and without its strong preference for shorter realisations.

The research reported in this paper is part of the CoGenT project, an ongoing research project supported under UK EP-SRC Grant GR/S24480/01. Many thanks to John Carroll, Roger Evans and Richard Power, as well as to the anonymous reviewers, for very helpful comments.

2005

bergmair-2009-proposal

https://aclanthology.org/W09-2502

A Proposal on Evaluation Measures for RTE

We outline problems with the interpretation of accuracy in the presence of bias, arguing that the issue is a particularly pressing concern for RTE evaluation. Furthermore, we argue that average precision scores are unsuitable for RTE, and should not be reported. We advocate mutual information as a new evaluation measure that should be reported in addition to accuracy and confidence-weighted score.

I would like to thank the anonymous reviewers and my colleague Ekaterina Shutova for providing many helpful comments and my supervisor Ann Copestake for reading multiple drafts of this paper and providing a great number of suggestions within a very short timeframe. All errors and omissions are, of course, entirely my own. I gratefully acknowledge financial support by the Austrian Academy of Sciences.

2009

hellwig-etal-2018-multi

https://aclanthology.org/L18-1011

Multi-layer Annotation of the Rigveda

The paper introduces a multi-level annotation of the R. GVEDA, a fundamental Sanskrit text composed in the 2. millenium BCE that is important for South-Asian and Indo-European linguistics, as well as Cultural Studies. We describe the individual annotation levels, including phonetics, morphology, lexicon, and syntax, and show how these different levels of annotation are merged to create a novel annotated corpus of Vedic Sanskrit. Vedic Sanskrit is a complex, but computationally under-resourced language. Therefore, creating this resource required considerable domain adaptation of existing computational tools, which is discussed in this paper. Because parts of the annotations are selective, we propose a bi-directional LSTM based sequential model to supplement missing verb-argument links.

Research for this project was partially funded by the Cluster of Excellence "Multimodal Computing and Interaction" of German Science Foundation (DFG). We thank the Akademie der Wissenschaften und der Literatur Mainz for hosting the annotated corpus.

2018

ciobanu-etal-2015-readability

https://aclanthology.org/R15-1014

Readability Assessment of Translated Texts

In this paper we investigate how readability varies between texts originally written in English and texts translated into English. For quantification, we analyze several factors that are relevant in assessing readability-shallow, lexical and morpho-syntactic features-and we employ the widely used Flesch-Kincaid formula to measure the variation of the readability level between original English texts and texts translated into English. Finally, we analyze whether the readability features have enough discriminative power to distinguish between originals and translations.

We thank the anonymous reviewers for their helpful and constructive comments. The contribution of the authors to this paper is equal. Liviu P. Dinu was supported by UEFISCDI, PNII-ID-PCE-2011-3-0959.

2015

miller-etal-2014-employing

https://aclanthology.org/W14-5308

Employing Phonetic Speech Recognition for Language and Dialect Specific Search

We discuss the notion of language and dialect-specific search in the context of audio indexing. A system is described where users can find dialect or language-specific pronunciations of Afghan placenames in Dari and Pashto. We explore the efficacy of a phonetic speech recognition system employed in this task.

2014

pecar-2018-towards

https://aclanthology.org/P18-3001

Towards Opinion Summarization of Customer Reviews

In recent years, the number of texts has grown rapidly. For example, most reviewbased portals, like Yelp or Amazon, contain thousands of user-generated reviews. It is impossible for any human reader to process even the most relevant of these documents. The most promising tool to solve this task is a text summarization. Most existing approaches, however, work on small, homogeneous, English datasets, and do not account to multi-linguality, opinion shift, and domain effects. In this paper, we introduce our research plan to use neural networks on user-generated travel reviews to generate summaries that take into account shifting opinions over time. We outline future directions in summarization to address all of these issues. By resolving the existing problems, we will make it easier for users of review-sites to make more informed decisions.

I would like to thank my supervisors Marian Simko and Maria Bielikova. This work has been partially supported by the STU Grant scheme for Support of Young Researchers and grants No. VG 1/0646/15 and No. KEGA 028STU-4/2017.

2018

montgomery-1997-fulcrum

https://aclanthology.org/1997.mtsummit-plenaries.5

The Fulcrum Approach to Machine Translation

In a paper from a distinguished collection of papers prepared for a 1959 course entitled "Computer Programming and Artificial Intelligence," Paul Garvin described two types of machine translation problems "in terms of the two components of the term: machine problems, and translation problems." While the machine problems made us crazy, the translation problems made us think differently about language than we might otherwise have done, which has had some advantages and some disadvantages in the long run. I will save anecdotes about the former and comments about the latter for the discussion. In this paper I will focus on the translation problems and, in particular, the translation approach that was developed by Paul Garvin, with whom I was associated, initially at Georgetown University, and later in the Synthetic Intelligence Department of the Ramo-Wooldridge Corporation and successor corporations: Thompson Ramo Wooldridge and Bunker-Ramo.

1997

corpas-pastor-etal-2008-translation

https://aclanthology.org/2008.amta-papers.5

Translation universals: do they exist? A corpus-based NLP study of convergence and simplification

Convergence and simplification are two of the so-called universals in translation studies. The first one postulates that translated texts tend to be more similar than nontranslated texts. The second one postulates that translated texts are simpler, easier-tounderstand than non-translated ones. This paper discusses the results of a project which applies NLP techniques over comparable corpora of translated and nontranslated texts in Spanish seeking to establish whether these two universals hold Corpas Pastor (2008).

2008

brun-2012-learning

https://aclanthology.org/C12-2017

Learning Opinionated Patterns for Contextual Opinion Detection

This paper tackles the problem of polar vocabulary ambiguity. While some opinionated words keep their polarity in any context and/or across any domain (except for the ironic style that goes beyond the present article), some other have an ambiguous polarity which is highly dependent of the context or the domain: in this case, the opinion is generally carried by complex expressions ("patterns") rather than single words. In this paper, we propose and evaluate an original hybrid method, based on syntactic information extraction and clustering techniques, to learn automatically such patterns and integrate them into an opinion detection system.

2012

turton-etal-2021-deriving

https://aclanthology.org/2021.repl4nlp-1.26

Deriving Contextualised Semantic Features from BERT (and Other Transformer Model) Embeddings

Models based on the transformer architecture, such as BERT, have marked a crucial step forward in the field of Natural Language Processing. Importantly, they allow the creation of word embeddings that capture important semantic information about words in context. However, as single entities, these embeddings are difficult to interpret and the models used to create them have been described as opaque. Binder and colleagues proposed an intuitive embedding space where each dimension is based on one of 65 core semantic features. Unfortunately, the space only exists for a small data-set of 535 words, limiting its uses. Previous work (Utsumi, 2018, 2020; Turton et al., 2020) has shown that Binder features can be derived from static embeddings and successfully extrapolated to a large new vocabulary. Taking the next step, this paper demonstrates that Binder features can be derived from the BERT embedding space. This provides two things; (1) semantic feature values derived from contextualised word embeddings and (2) insights into how semantic features are represented across the different layers of the BERT model.

2021

gemes-recski-2021-tuw

https://aclanthology.org/2021.germeval-1.10

TUW-Inf at GermEval2021: Rule-based and Hybrid Methods for Detecting Toxic, Engaging, and Fact-Claiming Comments

This paper describes our methods submitted for the GermEval 2021 shared task on identifying toxic, engaging and factclaiming comments in social media texts (Risch et al., 2021). We explore simple strategies for semi-automatic generation of rule-based systems with high precision and low recall, and use them to achieve slight overall improvements over a standard BERT-based classifier.

Peace, Justice and Strong Institutions

Research conducted in collaboration with Botium GmbH.

2021

temnikova-cohen-2013-recognizing

https://aclanthology.org/W13-1909

Recognizing Sublanguages in Scientific Journal Articles through Closure Properties

It has long been realized that sublanguages are relevant to natural language processing and text mining. However, practical methods for recognizing or characterizing them have been lacking. This paper describes a publicly available set of tools for sublanguage recognition. Closure properties are used to assess the goodness of fit of two biomedical corpora to the sublanguage model. Scientific journal articles are compared to general English text, and it is shown that the journal articles fit the sublanguage model, while the general English text does not. A number of examples of implications of the sublanguage characteristics for natural language processing are pointed out. The software is made publicly available at [edited for anonymization].

Industry, Innovation and Infrastructure

Irina Temnikova's work on the research reported in this paper was supported by the project AComIn "Advanced Computing for Innovation", grant 316087, funded by the FP7 Capacity Programme (Research Potential of Convergence Re-gions). Kevin Bretonnel Cohen's work was supported by grants NIH 5R01 LM009254-07 and NIH 5R01 LM008111-08 to Lawrence E. Hunter, NIH 1R01MH096906-01A1 to Tal Yarkoni, NIH R01 LM011124 to John Pestian, and NSF IIS-1207592 to Lawrence E. Hunter and Barbara Grimpe. The authors thank Tony McEnery and Andrew Wilson for advice on dealing with the tag sets.

2013

krishnamurthy-mitchell-2012-weakly

https://aclanthology.org/D12-1069

Weakly Supervised Training of Semantic Parsers

We present a method for training a semantic parser using only a knowledge base and an unlabeled text corpus, without any individually annotated sentences. Our key observation is that multiple forms of weak supervision can be combined to train an accurate semantic parser: semantic supervision from a knowledge base, and syntactic supervision from dependencyparsed sentences. We apply our approach to train a semantic parser that uses 77 relations from Freebase in its knowledge representation. This semantic parser extracts instances of binary relations with state-of-theart accuracy, while simultaneously recovering much richer semantic structures, such as conjunctions of multiple relations with partially shared arguments. We demonstrate recovery of this richer structure by extracting logical forms from natural language queries against Freebase. On this task, the trained semantic parser achieves 80% precision and 56% recall, despite never having seen an annotated logical form.

This research has been supported in part by DARPA under contract number FA8750-09-C-0179, and by a grant from Google. Additionally, we thank Yahoo! for use of their M45 cluster. We also gratefully acknowledge the contributions of our colleagues on the NELL project, Justin Betteridge for collecting the Freebase relations, Jamie Callan and colleagues for the web crawl, and Thomas Kollar and Matt Gardner for helpful comments on earlier drafts of this paper.

2012

benotti-blackburn-2021-recipe

https://aclanthology.org/2021.naacl-main.320

A recipe for annotating grounded clarifications

In order to interpret the communicative intents of an utterance, it needs to be grounded in something that is outside of language; that is, grounded in world modalities. In this paper we argue that dialogue clarification mechanisms make explicit the process of interpreting the communicative intents of the speaker's utterances by grounding them in the various modalities in which the dialogue is situated. This paper frames dialogue clarification mechanisms as an understudied research problem and a key missing piece in the giant jigsaw puzzle of natural language understanding. We discuss both the theoretical background and practical challenges posed by this problem, and propose a recipe for obtaining grounding annotations. We conclude by highlighting ethical issues that need to be addressed in future work. 1 We are suspicious of the common assumption that requests for information regarding references that are grounded in vision (e.g. the red or the blue jacket?) are clarifications, whereas requests for information grounded in other modalities are not (e.g. do I take the stairs up or down?). 2 See also the supplement on ethical considerations.

We thank the anonymous reviewers for their detailed reviews and insightful comments.

2021

chen-etal-2022-discrete

https://aclanthology.org/2022.acl-long.145

Discrete Opinion Tree Induction for Aspect-based Sentiment Analysis

Dependency trees have been intensively used with graph neural networks for aspect-based sentiment classification. Though being effective, such methods rely on external dependency parsers, which can be unavailable for low-resource languages or perform worse in low-resource domains. In addition, dependency trees are also not optimized for aspect-based sentiment classification. In this paper, we propose an aspect-specific and language-agnostic discrete latent opinion tree model as an alternative structure to explicit dependency trees. To ease the learning of complicated structured latent variables, we build a connection between aspect-to-context attention scores and syntactic distances, inducing trees from the attention scores. Results on six English benchmarks, one Chinese dataset and one Korean dataset show that our model can achieve competitive performance and interpretability.

Zhiyang Teng and Yue Zhang are the corresponding authors. Our thanks to anonymous reviewers for their insightful comments and suggestions. We appreciate Prof. Pengyuan Liu sharing the Chinese Hotel dataset, Prof. Jingjing Wang sharing the reinforcement learning code of Wang et al. 2019 Wu et al. (2020) upon our request. We thank Dr. Xuebin Wang for providing us with 2 V100 GPU cards for use. This publication is conducted with the financial support of "Pioneer" and "Leading Goose" R&D Program of Zhejiang under Grant Number 2022SDXHDX0003.

2022

espla-gomis-etal-2016-ualacant

https://aclanthology.org/W16-2383

UAlacant word-level and phrase-level machine translation quality estimation systems at WMT 2016

This paper describes the Universitat d'Alacant submissions (labeled as UAlacant) to the machine translation quality estimation (MTQE) shared task at WMT 2016, where we have participated in the word-level and phrase-level MTQE subtasks. Our systems use external sources of bilingual information as a black box to spot sub-segment correspondences between the source segment and the translation hypothesis. For our submissions, two sources of bilingual information have been used: machine translation (Lucy LT KWIK Translator and Google Translate) and the bilingual concordancer Reverso Context. Building upon the word-level approach implemented for WMT 2015, a method for phrase-based MTQE is proposed which builds on the probabilities obtained for word-level MTQE. For each sub-task we have submitted two systems: one using the features produced exclusively based on online sources of bilingual information, and one combining them with the baseline features provided by the organisers of the task.

Work partially funded by the European Commission through project PIAP-GA-2012-324414 (Abu-MaTran) and by the Spanish government through project TIN2015-69632-R (Effortune). We specially thank Reverso-Softissimo and Prompsit Language Engineering for providing the access to the Reverso Context concordancer, the University Research Program for Google Translate that granted us access to the Google Translate service, and Anna Civil from Lucy Software for providing access to the Lucy LT machine translation system.

2016

ceska-fox-2009-influence

https://aclanthology.org/R09-1011

The Influence of Text Pre-processing on Plagiarism Detection

This paper explores the influence of text preprocessing techniques on plagiarism detection. We examine stop-word removal, lemmatization, number replacement, synonymy recognition, and word generalization. We also look into the influence of punctuation and word-order within N-grams. All these techniques are evaluated according to their impact on F1-measure and speed of execution. Our experiments were performed on a Czech corpus of plagiarized documents about politics. At the end of this paper, we propose what we consider to be the best combination of text pre-processing techniques.

Industry, Innovation and Infrastructure

Peace, Justice and Strong Institutions

This research was supported in part by National Research Programme II, project 2C06009 (COT-SEWing). Special thanks go to Michal Toman who helped us to employ the disambiguation process.

2009

lefrancois-gandon-2013-reasoning

https://aclanthology.org/W13-3719

Reasoning with Dependency Structures and Lexicographic Definitions Using Unit Graphs

We are interested in a graph-based Knowledge Representation (KR) formalism that would allow for the representation, manipulation, query, and reasoning over dependency structures, and linguistic knowledge of the lexicon in the Meaning-Text Theory framework. Neither the semantic web formalisms nor the conceptual graphs appear to be suitable for this task, and this led to the introduction of the new Unit Graphs (UG) framework. In this paper we will overview the foundational concepts of this framework: the UGs are defined over a UG-support that contains: i) a hierarchy of unit types which is strongly driven by the actantial structure of unit types, ii) a hierarchy of circumstantial symbols, and iii) a set of unit identifiers. Based on these foundational concepts and on the definition of UGs, this paper justifies the use of a deep semantic representation level to represent meanings of lexical units. Rules over UGs are then introduced, and lexicographic definitions of lexical units are added to the hierarchy of unit types. Finally this paper provides UGs with semantics (in the logical sense), and pose the entailment problem, so as to enable the reasoning in the UGs framework.

2013

gali-etal-2008-aggregating

https://aclanthology.org/I08-5005

Aggregating Machine Learning and Rule Based Heuristics for Named Entity Recognition

This paper, submitted as an entry for the NERSSEAL-2008 shared task, describes a system build for Named Entity Recognition for South and South East Asian Languages. Our paper combines machine learning techniques with language specific heuristics to model the problem of NER for Indian languages. The system has been tested on five languages: Telugu, Hindi, Bengali, Urdu and Oriya. It uses CRF (Conditional Random Fields) based machine learning, followed by post processing which involves using some heuristics or rules. The system is specifically tuned for Hindi and Telugu, we also report the results for the other four languages.

We would like to thank the organizer Mr. Anil Kumar Singh deeply for his continuous support during the shared task.

2008

osborne-2013-distribution

https://aclanthology.org/W13-3730

The Distribution of Floating Quantifiers: A Dependency Grammar Analysis

This contribution provides a dependency grammar analysis of the distribution of floating quantifiers in English and German. Floating quantifiers are deemed to be "base generated", meaning that they are not moved into their surface position by a transformation. Their distribution is similar to that of modal adverbs. The nominal (noun or pronoun) over which they quantify is an argument of the predicate to which they attach. Variation in their placement across English and German is due to independent word order principles associated with each language.

2013

hossain-etal-2021-nlp-cuet

https://aclanthology.org/2021.ltedi-1.25

NLP-CUET@LT-EDI-EACL2021: Multilingual Code-Mixed Hope Speech Detection using Cross-lingual Representation Learner

In recent years, several systems have been developed to regulate the spread of negativity and eliminate aggressive, offensive or abusive contents from the online platforms. Nevertheless, a limited number of researches carried out to identify positive, encouraging and supportive contents. In this work, our goal is to identify whether a social media post/comment contains hope speech or not. We propose three distinct models to identify hope speech in English, Tamil and Malayalam language to serve this purpose. To attain this goal, we employed various machine learning (support vector machine, logistic regression, ensemble), deep learning (convolutional neural network + long short term memory) and transformer (m-BERT, Indic-BERT, XLNet, XLM-Roberta) based methods. Results indicate that XLM-Roberta outdoes all other techniques by gaining a weighted f 1-score of 0.93, 0.60 and 0.85 respectively for English, Tamil and Malayalam language. Our team has achieved 1 st , 2 nd and 1 st rank in these three tasks respectively.

Good Health and Well-Being

2021

morales-etal-2007-multivariate

https://aclanthology.org/W07-2421

Multivariate Cepstral Feature Compensation on Band-limited Data for Robust Speech Recognition

This paper describes a new method for compensating bandwidth mismatch for automatic speech recognition using multivariate linear combinations of feature vector components. It is shown that multivariate compensation is superior to methods based on linear compensations of individual features. Performance is evaluated on a real microphone-telephone mismatch condition (this involves noise compensation and bandwidth extension of real data), as well as on several artificial bandwidth limitations. Speech recognition accuracy using this approach is similar to that of acoustic model compensation methods for small to moderate mismatches, and allows keeping active a single acoustic model set for multiple bandwidth limitations.

This research is supported in part by an MCyT project (TIC 2006-13141-C03).

2007

jansche-2003-parametric

https://aclanthology.org/P03-1037

Parametric Models of Linguistic Count Data

It is well known that occurrence counts of words in documents are often modeled poorly by standard distributions like the binomial or Poisson. Observed counts vary more than simple models predict, prompting the use of overdispersed models like Gamma-Poisson or Beta-binomial mixtures as robust alternatives. Another deficiency of standard models is due to the fact that most words never occur in a given document, resulting in large amounts of zero counts. We propose using zeroinflated models for dealing with this, and evaluate competing models on a Naive Bayes text classification task. Simple zero-inflated models can account for practically relevant variation, and can be easier to work with than overdispersed models.

Thanks to Chris Brew and three anonymous reviewers for valuable feedback. Cue the usual disclaimers.

2003

collier-etal-1998-refining

https://aclanthology.org/W98-1109

Refining the Automatic Identification of Conceptual Relations in Large-scale Corpora

In the ACRONYM Project, we have taken the Firthian view (e.g. Firth 1957) that context is part of the meaning of the word, and measured similarity of meaning between words through second-order collocation. Using large-scale, free text corpora of UK journalism, we have generated collocational data for all words except for highfrequency grammatical words, and have found that semantically related word pairings can be identified, whilst syntactic relations are disfavoured. We have then moved on to refine this system, to deal with multi-word terms and identify changing conceptual relationships across time. The system, conceived in the late 80's and developed in 1994-97, differs from others of the 90's in purpose, scope, methodology and results, and comparisons will be drawn in the course of the paper.

1998

bjerva-augenstein-2018-phonology

https://aclanthology.org/N18-1083

From Phonology to Syntax: Unsupervised Linguistic Typology at Different Levels with Language Embeddings

A core part of linguistic typology is the classification of languages according to linguistic properties, such as those detailed in the World Atlas of Language Structure (WALS). Doing this manually is prohibitively time-consuming, which is in part evidenced by the fact that only 100 out of over 7,000 languages spoken in the world are fully covered in WALS. We learn distributed language representations, which can be used to predict typological properties on a massively multilingual scale. Additionally, quantitative and qualitative analyses of these language embeddings can tell us how language similarities are encoded in NLP models for tasks at different typological levels. The representations are learned in an unsupervised manner alongside tasks at three typological levels: phonology (grapheme-to-phoneme prediction, and phoneme reconstruction), morphology (morphological inflection), and syntax (part-of-speech tagging). We consider more than 800 languages and find significant differences in the language representations encoded, depending on the target task. For instance, although Norwegian Bokmål and Danish are typologically close to one another, they are phonologically distant, which is reflected in their language embeddings growing relatively distant in a phonological task. We are also able to predict typological features in WALS with high accuracies, even for unseen language families.

We would also like to thank RobertÖstling for giving us access to the pre-trained language embeddings. Isabelle Augenstein is supported by Eurostars grant Number E10138. We further gratefully acknowledge the support of NVIDIA Corporation with the donation of the Titan Xp GPU used for this research.

2018

godard-etal-2018-adaptor

https://aclanthology.org/W18-5804

Adaptor Grammars for the Linguist: Word Segmentation Experiments for Very Low-Resource Languages

Computational Language Documentation attempts to make the most recent research in speech and language technologies available to linguists working on language preservation and documentation. In this paper, we pursue two main goals along these lines. The first is to improve upon a strong baseline for the unsupervised word discovery task on two very lowresource Bantu languages, taking advantage of the expertise of linguists on these particular languages. The second consists in exploring the Adaptor Grammar framework as a decision and prediction tool for linguists studying a new language. We experiment 162 grammar configurations for each language and show that using Adaptor Grammars for word segmentation enables us to test hypotheses about a language. Specializing a generic grammar with language specific knowledge leads to great improvements for the word discovery task, ultimately achieving a leap of about 30% token F-score from the results of a strong baseline.

We thank the anonymous reviewers for their insightful comments. We also thank Ramy Eskander for his help in the early stages of this research. This work was partly funded by French ANR and German DFG under grant ANR-14-CE35-0002 (BULB project).

2018

king-2008-osu

https://aclanthology.org/W08-1137

OSU-GP: Attribute Selection Using Genetic Programming

This system's approach to the attribute selection task was to use a genetic programming algorithm to search for a solution to the task. The evolved programs for the furniture and people domain exhibit quite naive behavior, and the DICE and MASI scores on the training sets reflect the poor humanlikeness of the programs.

2008

herbelot-2020-solve

https://aclanthology.org/2020.conll-1.27

Re-solve it: simulating the acquisition of core semantic competences from small data

Many tasks are considered to be 'solved' in the computational linguistics literature, but the corresponding algorithms operate in ways which are radically different from human cognition. I illustrate this by coming back to the notion of semantic competence, which includes basic linguistic skills encompassing both referential phenomena and generic knowledge, in particular a) the ability to denote, b) the mastery of the lexicon, or c) the ability to model one's language use on others. Even though each of those faculties has been extensively tested individually, there is still no computational model that would account for their joint acquisition under the conditions experienced by a human. In this paper, I focus on one particular aspect of this problem: the amount of linguistic data available to the child or machine. I show that given the first competence mentioned above (a denotation function), the other two can in fact be learned from very limited data (2.8M token), reaching state-of-theart performance. I argue that both the nature of the data and the way it is presented to the system matter to acquisition.

I thank Ann Copestake and Katrin Erk for reading an early draft of this paper, as well as the participants to the GeCKo workshop in Barcelona for their helpful comments. I would also like to thank the anonymous reviewers for their helpful suggestions and comments. Finally, I gratefully acknowledge the support of NVIDIA Corporation with the donation of the Titan V GPU used for this research.

2020