Back to home
All Datasets 🏷
All datasets from our datasets repository and community bucket.
Also check out the list of supported Metrics 📉.
51 results
The ALT project aims to advance the state-of-the-art Asian natural language processing (NLP) techniques through the open collaboration for developing and using ALT. It was first conducted by NICT and UCSY as described in Ye Kyaw Thu, Win Pa Pa, Masao Utiyama, Andrew Finch and Eiichiro Sumita (2016). Then, it was developed under ASEAN IVO as described in this Web page. The process of building ALT began with sampling about 20,000 sentences from English Wikinews, and then these sentences were translated into the other languages. ALT now has 13 languages: Bengali, English, Filipino, Hindi, Bahasa Indonesia, Japanese, Khmer, Lao, Malay, Myanmar (Burmese), Thai, Vietnamese, Chinese (Simplified Chinese).
This dataset provides the template sentences and relationships defined in the ATOMIC common sense dataset. There are three splits - train, test, and dev. From the authors. Disclaimer/Content warning: the events in atomic have been automatically extracted from blogs, stories and books written at various times. The events might depict violent or problematic actions, which we left in the corpus for the sake of learning the (probably negative but still important) commonsense implications associated with the events. We removed a small set of truly out-dated events, but might have missed some so please email us ( if you have any concerns.
The Circa (meaning ‘approximately’) dataset aims to help machine learning systems to solve the problem of interpreting indirect answers to polar questions. The dataset contains pairs of yes/no questions and indirect answers, together with annotations for the interpretation of the answer. The data is collected in 10 different social conversational situations (eg. food preferences of a friend). NOTE: There might be missing labels in the dataset and we have replaced them with -1. The original dataset contains no train/dev/test splits.
This dataset is designed to provide training data for common sense relationships pulls together from various sources. The dataset is multi-lingual. See langauge codes and language info here: This dataset provides an interface for the conceptnet5 csv file, and some (but not all) of the raw text data used to build conceptnet5: omcsnet_sentences_free.txt, and omcsnet_sentences_more.txt. One use of this dataset would be to learn to extract the conceptnet relationship from the omcsnet sentences. Conceptnet5 has 34,074,917 relationships. Of those relationships, there are 2,176,099 surface text sentences related to those 2M entries. omcsnet_sentences_free has 898,161 lines. omcsnet_sentences_more has 2,001,736 lines. Original downloads are available here For more information, see: The omcsnet data comes with the following warning from the authors of the above site: Remember: this data comes from various forms of crowdsourcing. Sentences in these files are not necessarily true, useful, or appropriate.
The COrpus of Urdu News TExt Reuse (COUNTER) corpus contains 1200 documents with real examples of text reuse from the field of journalism. It has been manually annotated at document level with three levels of reuse: wholly derived, partially derived and non derived.
In October 2012, the European Union's (EU) Directorate General for Education and Culture ( DG EAC) released a translation memory (TM), i.e. a collection of sentences and their professionally produced translations, in twenty-six languages. This resource bears the name EAC Translation Memory, short EAC-TM. EAC-TM covers up to 26 languages: 22 official languages of the EU (all except Irish) plus Icelandic, Croatian, Norwegian and Turkish. EAC-TM thus contains translations from English into the following 25 languages: Bulgarian, Czech, Danish, Dutch, Estonian, German, Greek, Finnish, French, Croatian, Hungarian, Icelandic, Italian, Latvian, Lithuanian, Maltese, Norwegian, Polish, Portuguese, Romanian, Slovak, Slovenian, Spanish, Swedish and Turkish. All documents and sentences were originally written in English (source language is English) and then translated into the other languages. The texts were translated by staff of the National Agencies of the Lifelong Learning and Youth in Action programmes. They are typically professionals in the field of education/youth and EU programmes. They are thus not professional translators, but they are normally native speakers of the target language.
The GenericsKB contains 3.4M+ generic sentences about the world, i.e., sentences expressing general truths such as "Dogs bark," and "Trees remove carbon dioxide from the atmosphere." Generics are potentially useful as a knowledge source for AI systems requiring general world knowledge. The GenericsKB is the first large-scale resource containing naturally occurring generic sentences (as opposed to extracted or crowdsourced triples), and is rich in high-quality, general, semantically complete statements. Generics were primarily extracted from three large text sources, namely the Waterloo Corpus, selected parts of Simple Wikipedia, and the ARC Corpus. A filtered, high-quality subset is also available in GenericsKB-Best, containing 1,020,868 sentences. We recommend you start with GenericsKB-Best.
When humans read or listen, they make implicit commonsense inferences that frame their understanding of what happened and why. As a step toward AI systems that can build similar mental models, we introduce GLUCOSE, a large-scale dataset of implicit commonsense causal knowledge, encoded as causal mini-theories about the world, each grounded in a narrative context.
Hatexplain is the first benchmark hate speech dataset covering multiple aspects of the issue. Each post in the dataset is annotated from three different perspectives: the basic, commonly used 3-class classification (i.e., hate, offensive or normal), the target community (i.e., the community that has been the victim of hate speech/offensive speech in the post), and the rationales, i.e., the portions of the post on which their labelling decision (as hate, offensive or normal) is based.
The Hausa VOA NER dataset is a labeled dataset for named entity recognition in Hausa. The texts were obtained from Hausa Voice of America News articles . We concentrate on four types of named entities: persons [PER], locations [LOC], organizations [ORG], and dates & time [DATE]. The Hausa VOA NER data files contain 2 columns separated by a tab ('\t'). Each word has been put on a separate line and there is an empty line after each sentences i.e the CoNLL format. The first item on each line is a word, the second is the named entity tag. The named entity tags have the format I-TYPE which means that the word is inside a phrase of type TYPE. For every multi-word expression like 'New York', the first word gets a tag B-TYPE and the subsequent words have tags I-TYPE, a word with tag O is not part of a phrase. The dataset is in the BIO tagging scheme. For more details, see
The Hong Kong Cantonese Corpus (HKCanCor) comprise transcribed conversations recorded between March 1997 and August 1998. It contains recordings of spontaneous speech (51 texts) and radio programmes (42 texts), which involve 2 to 4 speakers, with 1 text of monologue. In total, the corpus contains around 230,000 Chinese words. The text is word-segmented, annotated with part-of-speech (POS) tags and romanised Cantonese pronunciation. Romanisation scheme - Linguistic Society of Hong Kong (LSHK) POS scheme - Peita-Fujitsu-Renmin Ribao (PRF) corpus (Duan et al., 2000), with extended tags for Cantonese-specific phenomena added by Luke and Wang (see original paper for details).
A Hope Speech dataset for Equality, Diversity and Inclusion (HopeEDI) containing user-generated comments from the social media platform YouTube with 28,451, 20,198 and 10,705 comments in English, Tamil and Malayalam, respectively, manually labelled as containing hope speech or not.
The CLICK-ID dataset is a collection of Indonesian news headlines that was collected from 12 local online news publishers; detikNews, Fimela, Kapanlagi, Kompas, Liputan6, Okezone, Posmetro-Medan, Republika, Sindonews, Tempo, Tribunnews, and Wowkeren. This dataset is comprised of mainly two parts; (i) 46,119 raw article data, and (ii) 15,000 clickbait annotated sample headlines. Annotation was conducted with 3 annotator examining each headline. Judgment were based only on the headline. The majority then is considered as the ground truth. In the annotated sample, our annotation shows 6,290 clickbait and 8,710 non-clickbait.
The dataset contains around 500K articles (136M of words) from 7 Indonesian newspapers: Detik, Kompas, Tempo, CNN Indonesia, Sindo, Republika and Poskota. The articles are dated between 1st January 2018 and 20th August 2018 (with few exceptions dated earlier). The size of uncompressed 500K json files (newspapers-json.tgz) is around 2.2GB, and the cleaned uncompressed in a big text file (newspapers.txt.gz) is about 1GB. The original source in Google Drive contains also a dataset in html format which include raw data (pictures, css, javascript, ...) from the online news website
This dataset is designed to identify speaker intention based on real-life spoken utterance in Korean into one of 7 categories: fragment, description, question, command, rhetorical question, rhetorical command, utterances.
The LAMBADA evaluates the capabilities of computational models for text understanding by means of a word prediction task. LAMBADA is a collection of narrative passages sharing the characteristic that human subjects are able to guess their last word if they are exposed to the whole passage, but not if they only see the last sentence preceding the target word. To succeed on LAMBADA, computational models cannot simply rely on local context, but must be able to keep track of information in the broader discourse. The LAMBADA dataset is extracted from BookCorpus and consists of 10'022 passages, divided into 4'869 development and 5'153 test passages. The training data for language models to be tested on LAMBADA include the full text of 2'662 novels (disjoint from those in dev+test), comprising 203 million words.
Mac-Morpho is a corpus of Brazilian Portuguese texts annotated with part-of-speech tags. Its first version was released in 2003 [1], and since then, two revisions have been made in order to improve the quality of the resource [2, 3]. The corpus is available for download split into train, development and test sections. These are 76%, 4% and 20% of the corpus total, respectively (the reason for the unusual numbers is that the corpus was first split into 80%/20% train/test, and then 5% of the train section was set aside for development). This split was used in [3], and new POS tagging research with Mac-Morpho is encouraged to follow it in order to make consistent comparisons possible. [1] Aluísio, S., Pelizzoni, J., Marchi, A.R., de Oliveira, L., Manenti, R., Marquiafável, V. 2003. An account of the challenge of tagging a reference corpus for brazilian portuguese. In: Proceedings of the 6th International Conference on Computational Processing of the Portuguese Language. PROPOR 2003 [2] Fonseca, E.R., Rosa, J.L.G. 2013. Mac-morpho revisited: Towards robust part-of-speech. In: Proceedings of the 9th Brazilian Symposium in Information and Human Language Technology – STIL [3] Fonseca, E.R., Aluísio, Sandra Maria, Rosa, J.L.G. 2015. Evaluating word embeddings and a revised corpus for part-of-speech tagging in Portuguese. Journal of the Brazilian Computer Society.
MENYO-20k is a multi-domain parallel dataset with texts obtained from news articles, ted talks, movie transcripts, radio transcripts, science and technology texts, and other short articles curated from the web and professional translators. The dataset has 20,100 parallel sentences split into 10,070 training sentences, 3,397 development sentences, and 6,633 test sentences (3,419 multi-domain, 1,714 news domain, and 1,500 ted talks speech transcript domain). The development and test sets are available upon request.
Offensive language identification in dravidian lanaguages dataset. The goal of this task is to identify offensive language content of the code-mixed dataset of comments/posts in Dravidian Languages ( (Tamil-English, Malayalam-English, and Kannada-English)) collected from social media.
ParaPat: The Multi-Million Sentences Parallel Corpus of Patents Abstracts This dataset contains the developed parallel corpus from the open access Google Patents dataset in 74 language pairs, comprising more than 68 million sentences and 800 million tokens. Sentences were automatically aligned using the Hunalign algorithm for the largest 22 language pairs, while the others were abstract (i.e. paragraph) aligned.
This new dataset is the large scale sentence aligned corpus in 11 Indian languages, viz. CVIT-PIB corpus that is the largest multilingual corpus available for Indian languages.
This dataset is for studying computational models trained to reason about prototypical situations. Using deterministic filtering a sampling from a larger set of all transcriptions was built. It contains 9789 instances where each instance represents a survey question from Family Feud game. Each instance exactly is a question, a set of answers, and a count associated with each answer. Each line is a json dictionary, in which: 1. question - contains the question (in original and a normalized form) 2. answerstrings - contains the original answers provided by survey respondents (when available), along with the counts for each string. Because the FamilyFeud data has only cluster names rather than strings, those cluster names are included with 0 weight. 3. answer-clusters - lists clusters, with the count of each cluster and the strings included in that cluster. Each cluster is given a unique ID that can be linked to in the assessment files.
ReDial (Recommendation Dialogues) is an annotated dataset of dialogues, where users recommend movies to each other. The dataset was collected by a team of researchers working at Polytechnique Montréal, MILA – Quebec AI Institute, Microsoft Research Montréal, HEC Montreal, and Element AI. The dataset allows research at the intersection of goal-directed dialogue systems (such as restaurant recommendation) and free-form (also called “chit-chat”) dialogue systems.
ROPES (Reasoning Over Paragraph Effects in Situations) is a QA dataset which tests a system's ability to apply knowledge from a passage of text to a new situation. A system is presented a background passage containing a causal or qualitative relation(s) (e.g., "animal pollinators increase efficiency of fertilization in flowers"), a novel situation that uses this background, and questions that require reasoning about effects of the relationships in the background passage in the background of the situation.
The SemEval-2014 Task 1 focuses on Evaluation of Compositional Distributional Semantic Models on Full Sentences through Semantic Relatedness and Entailment. The task was designed to predict the degree of relatedness between two sentences and to detect the entailment relation holding between them.
DFKI SmartData Corpus is a dataset of 2598 German-language documents which has been annotated with fine-grained geo-entities, such as streets, stops and routes, as well as standard named entity types. It has also been annotated with a set of 15 traffic- and industry-related n-ary relations and events, such as Accidents, Traffic jams, Acquisitions, and Strikes. The corpus consists of newswire texts, Twitter messages, and traffic reports from radio stations, police and railway companies. It allows for training and evaluating both named entity recognition algorithms that aim for fine-grained typing of geo-entities, as well as n-ary relation extraction systems.
The SNLI corpus (version 1.0) is a collection of 570k human-written English sentence pairs manually labeled for balanced classification with the labels entailment, contradiction, and neutral, supporting the task of natural language inference (NLI), also known as recognizing textual entailment (RTE).
About SNOW T15: The simplified corpus for the Japanese language. The corpus has 50,000 manually simplified and aligned sentences. This corpus contains the original sentences, simplified sentences and English translation of the original sentences. It can be used for automatic text simplification as well as translating simple Japanese into English and vice-versa. The core vocabulary is restricted to 2,000 words where it is selected by accounting for several factors such as meaning preservation, variation, simplicity and the UniDic word segmentation criterion. For details, refer to the explanation page of Japanese simplification ( The original texts are from "small_parallel_enja: 50k En/Ja Parallel Corpus for Testing SMT Methods", which is a bilingual corpus for machine translation. About SNOW T23: An expansion corpus of 35,000 sentences rewritten in easy Japanese (simple Japanese vocabulary) based on SNOW T15. The original texts are from "Tanaka Corpus" (
Social Bias Frames is a new way of representing the biases and offensiveness that are implied in language. For example, these frames are meant to distill the implication that "women (candidates) are less qualified" behind the statement "we shouldn’t lower our standards to hire more women."
The SOFC-Exp corpus consists of 45 open-access scholarly articles annotated by domain experts. A corpus and an inter-annotator agreement study demonstrate the complexity of the suggested named entity recognition and slot filling tasks as well as high annotation quality is presented in the accompanying paper.
The Swahili dataset developed specifically for language modeling task. The dataset contains 28,000 unique words with 6.84M, 970k, and 2M words for the train, valid and test partitions respectively which represent the ratio 80:10:10. The entire dataset is lowercased, has no punctuation marks and, the start and end of sentence markers have been incorporated to facilitate easy tokenization during language modeling.
Swahili is spoken by 100-150 million people across East Africa. In Tanzania, it is one of two national languages (the other is English) and it is the official language of instruction in all schools. News in Swahili is an important part of the media sphere in Tanzania. News contributes to education, technology, and the economic growth of a country, and news in local languages plays an important cultural role in many Africa countries. In the modern age, African languages in news and other spheres are at risk of being lost as English becomes the dominant language in online spaces. The Swahili news dataset was created to reduce the gap of using the Swahili language to create NLP technologies and help AI practitioners in Tanzania and across Africa continent to practice their NLP skills to solve different problems in organizations or societies related to Swahili language. Swahili News were collected from different websites that provide news in the Swahili language. I was able to find some websites that provide news in Swahili only and others in different languages including Swahili. The dataset was created for a specific task of text classification, this means each news content can be categorized into six different topics (Local news, International news , Finance news, Health news, Sports news, and Entertainment news). The dataset comes with a specified train/test split. The train set contains 75% of the dataset and test set contains 25% of the dataset.
Webbnyheter 2012 from Spraakbanken, semi-manually annotated and adapted for CoreNLP Swedish NER. Semi-manually defined in this case as: Bootstrapped from Swedish Gazetters then manually correcte/reviewed by two independent native speaking swedish annotators. No annotator agreement calculated.
The problem of verifying whether a textual hypothesis holds the truth based on the given evidence, also known as fact verification, plays an important role in the study of natural language understanding and semantic representation. However, existing studies are restricted to dealing with unstructured textual evidence (e.g., sentences and passages, a pool of passages), while verification using structured forms of evidence, such as tables, graphs, and databases, remains unexplored. TABFACT is large scale dataset with 16k Wikipedia tables as evidence for 118k human annotated statements designed for fact verification with semi-structured evidence. The statements are labeled as either ENTAILED or REFUTED. TABFACT is challenging since it involves both soft linguistic reasoning and hard symbolic reasoning.
Taskmaster-1 is a goal-oriented conversational dataset. It includes 13,215 task-based dialogs comprising six domains. Two procedures were used to create this collection, each with unique advantages. The first involves a two-person, spoken "Wizard of Oz" (WOz) approach in which trained agents and crowdsourced workers interact to complete the task while the second is "self-dialog" in which crowdsourced workers write the entire dialog themselves.
Taskmaster is dataset for goal oriented conversations. The Taskmaster-2 dataset consists of 17,289 dialogs in the seven domains which include restaurants, food ordering, movies, hotels, flights, music and sports. Unlike Taskmaster-1, which includes both written "self-dialogs" and spoken two-person dialogs, Taskmaster-2 consists entirely of spoken two-person dialogs. In addition, while Taskmaster-1 is almost exclusively task-based, Taskmaster-2 contains a good number of search- and recommendation-oriented dialogs. All dialogs in this release were created using a Wizard of Oz (WOz) methodology in which crowdsourced workers played the role of a 'user' and trained call center operators played the role of the 'assistant'. In this way, users were led to believe they were interacting with an automated system that “spoke” using text-to-speech (TTS) even though it was in fact a human behind the scenes. As a result, users could express themselves however they chose in the context of an automated interface.
Taskmaster is dataset for goal oriented conversations. The Taskmaster-3 dataset consists of 23,757 movie ticketing dialogs. By "movie ticketing" we mean conversations where the customer's goal is to purchase tickets after deciding on theater, time, movie name, number of tickets, and date, or opt out of the transaction. This collection was created using the "self-dialog" method. This means a single, crowd-sourced worker is paid to create a conversation writing turns for both speakers, i.e. the customer and the ticketing agent.
Turkish Wikipedia Named-Entity Recognition and Text Categorization (TWNERTC) dataset is a collection of automatically categorized and annotated sentences obtained from Wikipedia. The authors constructed large-scale gazetteers by using a graph crawler algorithm to extract relevant entity and domain information from a semantic knowledge base, Freebase. The constructed gazetteers contains approximately 300K entities with thousands of fine-grained entity types under 77 different domains.
Shrinked version (48 entity type) of the turkish_ner. Original turkish_ner dataset: Automatically annotated Turkish corpus for named entity recognition and text categorization using large-scale gazetteers. The constructed gazetteers contains approximately 300K entities with thousands of fine-grained entity types under 25 different domains. Shrinked entity types are: academic, academic_person, aircraft, album_person, anatomy, animal, architect_person, capital, chemical, clothes, country, culture, currency, date, food, genre, government, government_person, language, location, material, measure, medical, military, military_person, nation, newspaper, organization, organization_person, person, production_art_music, production_art_music_person, quantity, religion, science, shape, ship, software, space, space_person, sport, sport_name, sport_person, structure, subject, tech, train, vehicle
A Winograd schema is a pair of sentences that differ in only one or two words and that contain an ambiguity that is resolved in opposite ways in the two sentences and requires the use of world knowledge and reasoning for its resolution. The schema takes its name from a well-known example by Terry Winograd: > The city councilmen refused the demonstrators a permit because they [feared/advocated] violence. If the word is ``feared'', then ``they'' presumably refers to the city council; if it is ``advocated'' then ``they'' presumably refers to the demonstrators.
A multilingual fine-grained emotion dataset. The dataset consists of human annotated Finnish (25k) and English sentences (30k). Plutchik’s core emotions are used to annotate the dataset with the addition of neutral to create a multilabel multiclass dataset. The dataset is carefully evaluated using language-specific BERT models and SVMs to show that XED performs on par with other similar datasets and is therefore a useful tool for sentiment analysis and emotion detection.
Neural abstractive summarization models are highly prone to hallucinate content that is unfaithful to the input document. The popular metric such as ROUGE fails to show the severity of the problem. The dataset consists of faithfulness and factuality annotations of abstractive summaries for the XSum dataset. We have crowdsourced 3 judgements for each of 500 x 5 document-system pairs. This will be a valuable resource to the abstractive summarization community.
ZEST tests whether NLP systems can perform unseen tasks in a zero-shot way, given a natural language description of the task. It is an instantiation of our proposed framework "learning from task descriptions". The tasks include classification, typed entity extraction and relationship extraction, and each task is paired with 20 different annotated (input, output) examples. ZEST's structure allows us to systematically test whether models can generalize in five different ways.