The Amazon reviews dataset consists of reviews from amazon.
The data span a period of 18 years, including ~35 million reviews up to March 2013.
Reviews include product and user information, ratings, and a plaintext review.
We provide an Amazon product reviews dataset for multilingual text classification. The dataset contains reviews in English, Japanese, German, French, Chinese and Spanish, collected between November 1, 2015 and November 1, 2019. Each record in the dataset contains the review text, the review title, the star rating, an anonymized reviewer ID, an anonymized product ID and the coarse-grained product category (e.g. ‘books’, ‘appliances’, etc.) The corpus is balanced across stars, so each star rating constitutes 20% of the reviews in each language.
For each language, there are 200,000, 5,000 and 5,000 reviews in the training, development and test sets respectively. The maximum number of reviews per reviewer is 20 and the maximum number of reviews per product is 20. All reviews are truncated after 2,000 characters, and all reviews are at least 20 characters long.
Note that the language of a review does not necessarily match the language of its marketplace (e.g. reviews from amazon.de are primarily written in German, but could also be written in English, etc.). For this reason, we applied a language detection algorithm based on the work in Bojanowski et al. (2017) to determine the language of the review text and we removed reviews that were not written in the expected language.
ArCOV-19 is an Arabic COVID-19 Twitter dataset that covers the period from 27th of January till 30th of April 2020. ArCOV-19 is designed to enable research under several domains including natural language processing, information retrieval, and social computing, among others
Abu El-Khair Corpus is an Arabic text corpus, that includes more than five million newspaper articles.
It contains over a billion and a half words in total, out of which, there are about three million unique words.
The corpus is encoded with two types of encoding, namely: UTF-8, and Windows CP-1256.
Also it was marked with two mark-up languages, namely: SGML, and XML.
A dataset of 1.7 million arXiv articles for applications like trend analysis, paper recommender engines, category prediction, co-citation networks, knowledge graph construction and semantic search interfaces.
A parallel news corpus in Turkish, Kurdish and English.
Bianet collects 3,214 Turkish articles with their sentence-aligned Kurdish or English translations from the Bianet online newspaper.
3 languages, 3 bitexts
total number of files: 6
total number of tokens: 2.25M
total number of sentence fragments: 0.14M
BIGPATENT, consisting of 1.3 million records of U.S. patent documents
along with human written abstractive summaries.
Each US patent application is filed under a Cooperative Patent Classification
(CPC) code. There are nine such classification categories:
A (Human Necessities), B (Performing Operations; Transporting),
C (Chemistry; Metallurgy), D (Textiles; Paper), E (Fixed Constructions),
F (Mechanical Engineering; Lightning; Heating; Weapons; Blasting),
G (Physics), H (Electricity), and
Y (General tagging of new or cross-sectional technology)
There are two features:
- description: detailed description of patent.
- abstract: Patent abastract.
The BrWaC (Brazilian Portuguese Web as Corpus) is a large corpus constructed following the Wacky framework,
which was made public for research purposes. The current corpus version, released in January 2017, is composed by
3.53 million documents, 2.68 billion tokens and 5.79 million types. Please note that this resource is available
solely for academic research purposes, and you agreed not to use it for any commercial applications.
Manually download at https://www.inf.ufrgs.br/pln/wiki/index.php?title=BrWaC
The Bosnian web corpus bsWaC was built by crawling the .ba top-level domain in 2014. The corpus was near-deduplicated on paragraph level, normalised via diacritic restoration, morphosyntactically annotated and lemmatised. The corpus is shuffled by paragraphs. Each paragraph contains metadata on the URL, domain and language identification (Bosnian vs. Croatian vs. Serbian).
Version 1.0 of this corpus is described in http://www.aclweb.org/anthology/W14-0405. Version 1.1 contains newer and better linguistic annotations.
In this paper, we introduce Chinese AI and Law challenge dataset (CAIL2018),
the first large-scale Chinese legal dataset for judgment prediction. CAIL contains more than 2.6 million
criminal cases published by the Supreme People's Court of China, which are several times larger than other
datasets in existing works on judgment prediction. Moreover, the annotations of judgment results are more
detailed and rich. It consists of applicable law articles, charges, and prison terms, which are expected
to be inferred according to the fact descriptions of cases. For comparison, we implement several conventional
text classification baselines for judgment prediction and experimental results show that it is still a
challenge for current models to predict the judgment results of legal cases, especially on prison terms.
To help the researchers make improvements on legal judgment prediction.
A parallel corpus of theses and dissertations abstracts in English and Portuguese were collected from the CAPES website (Coordenação de Aperfeiçoamento de Pessoal de Nível Superior) - Brazil. The corpus is sentence aligned for all language pairs. Approximately 240,000 documents were collected and aligned using the Hunalign algorithm.
caWaC is a 780-million-token web corpus of Catalan built from the .cat top-level-domain in late 2013.
This corpus is an attempt to recreate the dataset used for training XLM-R. This corpus comprises of monolingual data for 100+ languages and also includes data for romanized languages (indicated by *_rom). This was constructed using the urls and paragraph indices provided by the CC-Net repository by processing January-December 2018 Commoncrawl snapshots. Each file comprises of documents separated by double-newlines and paragraphs within the same document separated by a newline. The data is generated using the open source CC-Net repository. No claims of intellectual property are made on the work of preparation of the corpus.
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
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
omcsnet_sentences_free has 898,161 lines. omcsnet_sentences_more has
Original downloads are available here
https://github.com/commonsense/conceptnet5/wiki/Downloads. For more
information, see: https://github.com/commonsense/conceptnet5/wiki
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.
This is a parallel corpus made out of PDF documents from the European Medicines Agency. All files are automatically converted from PDF to plain text using pdftotext with the command line arguments -layout -nopgbrk -eol unix. There are some known problems with tables and multi-column layouts - some of them are fixed in the current version.
22 languages, 231 bitexts
total number of files: 41,957
total number of tokens: 311.65M
total number of sentence fragments: 26.51M
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.
Giga-word corpus for French-English from WMT2010 collected by Chris Callison-Burch
2 languages, total number of files: 452
total number of tokens: 1.43G
total number of sentence fragments: 47.55M
The Croatian web corpus hrWaC was built by crawling the .hr top-level domain in 2011 and again in 2014. The corpus was near-deduplicated on paragraph level, normalised via diacritic restoration, morphosyntactically annotated and lemmatised. The corpus is shuffled by paragraphs. Each paragraph contains metadata on the URL, domain and language identification (Croatian vs. Serbian).
Version 2.0 of this corpus is described in http://www.aclweb.org/anthology/W14-0405. Version 2.1 contains newer and better linguistic annotations.
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
from the online news website
Data-To-Text Generation involves converting knowledge graph (KG) triples of the form (subject, relation, object) into
a natural language sentence(s). This dataset consists of English KG data converted into paired natural language text.
The generated corpus consists of ∼18M sentences spanning ∼45M triples with ∼1500 distinct relations.
KILT tasks training and evaluation data.
- [FEVER](https://fever.ai) | Fact Checking | fever
- [AIDA CoNLL-YAGO](https://www.mpi-inf.mpg.de/departments/databases-and-information-systems/research/ambiverse-nlu/aida/downloads) | Entity Linking | aidayago2
- [WNED-WIKI](https://github.com/U-Alberta/wned) | Entity Linking | wned
- [WNED-CWEB](https://github.com/U-Alberta/wned) | Entity Linking | cweb
- [T-REx](https://hadyelsahar.github.io/t-rex) | Slot Filling | trex
- [Zero-Shot RE](http://nlp.cs.washington.edu/zeroshot) | Slot Filling | structured_zeroshot
- [Natural Questions](https://ai.google.com/research/NaturalQuestions) | Open Domain QA | nq
- [HotpotQA](https://hotpotqa.github.io) | Open Domain QA | hotpotqa
- [TriviaQA](http://nlp.cs.washington.edu/triviaqa) | Open Domain QA | triviaqa
- [ELI5](https://facebookresearch.github.io/ELI5/explore.html) | Open Domain QA | eli5
- [Wizard of Wikipedia](https://parl.ai/projects/wizard_of_wikipedia) | Dialogue | wow
To finish linking TriviaQA questions to the IDs provided, follow the instructions [here](http://github.com/huggingface/datasets/datasets/kilt_tasks/README.md).
LAMA is a dataset used to probe and analyze the factual and commonsense knowledge contained in pretrained language models. See https://github.com/facebookresearch/LAMA.
The Large Spanish Corpus is a compilation of 15 unlabelled Spanish corpora spanning Wikipedia to European parliament notes. Each config contains the data corresponding to a different corpus. For example, "all_wiki" only includes examples from Spanish Wikipedia. By default, the config is set to "combined" which loads all the corpora; with this setting you can also specify the number of samples to return per corpus by configuring the "split" argument.
Machine learning models are trained to find patterns in data.
NLP models can inadvertently learn socially undesirable patterns when training on gender biased text.
In this work, we propose a general framework that decomposes gender bias in text along several pragmatic and semantic dimensions:
bias from the gender of the person being spoken about, bias from the gender of the person being spoken to, and bias from the gender of the speaker.
Using this fine-grained framework, we automatically annotate eight large scale datasets with gender information.
In addition, we collect a novel, crowdsourced evaluation benchmark of utterance-level gender rewrites.
Distinguishing between gender bias along multiple dimensions is important, as it enables us to train finer-grained gender bias classifiers.
We show our classifiers prove valuable for a variety of important applications, such as controlling for gender bias in generative models,
detecting gender bias in arbitrary text, and shed light on offensive language in terms of genderedness.
MultiReQA contains the sentence boundary annotation from eight publicly available QA datasets including SearchQA, TriviaQA, HotpotQA, NaturalQuestions, SQuAD, BioASQ, RelationExtraction, and TextbookQA. Five of these datasets, including SearchQA, TriviaQA, HotpotQA, NaturalQuestions, SQuAD, contain both training and test data, and three, including BioASQ, RelationExtraction, TextbookQA, contain only the test data
This dataset provides version 1115 of the belief
extracted by CMU's Never Ending Language Learner (NELL) and version
1110 of the candidate belief extracted by NELL. See
http://rtw.ml.cmu.edu/rtw/overview. NELL is an open information
extraction system that attempts to read the Clueweb09 of 500 million
web pages (http://boston.lti.cs.cmu.edu/Data/clueweb09/) and general
The dataset has 4 configurations: nell_belief, nell_candidate,
nell_belief_sentences, and nell_candidate_sentences. nell_belief is
certainties of belief are lower. The two sentences config extracts the
CPL sentence patterns filled with the applicable 'best' literal string
for the entities filled into the sentence patterns. And also provides
sentences found using web searches containing the entities and
There are roughly 21M entries for nell_belief_sentences, and 100M
sentences for nell_candidate_sentences.
Neural-Code-Search-Evaluation-Dataset presents an evaluation dataset consisting of natural language query and code snippet pairs and a search corpus consisting of code snippets collected from the most popular Android repositories on GitHub.
Large-scale dataset of Filipino news articles. Sourced for the NewsPH-NLI Project (Cruz et al., 2020).
The Ollie dataset includes two configs for the data
used to train the Ollie informatation extraction algorithm, for 18M
sentences and 3M sentences respectively.
This data is for academic use only. From the authors:
Ollie is a program that automatically identifies and extracts binary
relationships from English sentences. Ollie is designed for Web-scale
information extraction, where target relations are not specified in
Ollie is our second-generation information extraction system . Whereas
ReVerb operates on flat sequences of tokens, Ollie works with the
tree-like (graph with only small cycles) representation using
Stanford's compression of the dependencies. This allows Ollie to
capture expression that ReVerb misses, such as long-range relations.
Ollie also captures context that modifies a binary relation. Presently
Ollie handles attribution (He said/she believes) and enabling
conditions (if X then).
More information is available at the Ollie homepage:
This is a new collection of translated movie subtitles from http://www.opensubtitles.org/.
IMPORTANT: If you use the OpenSubtitle corpus: Please, add a link to http://www.opensubtitles.org/ to your website and to your reports and publications produced with the data!
This is a slightly cleaner version of the subtitle collection using improved sentence alignment and better language checking.
62 languages, 1,782 bitexts
total number of files: 3,735,070
total number of tokens: 22.10G
total number of sentence fragments: 3.35G
This is a collection of documents from the Official Journal of the Government of Catalonia, in Catalan and Spanish languages, provided by Antoni Oliver Gonzalez from the Universitat Oberta de Catalunya.
The dataset contains 2231142 cooking recipes (>2 millions). It's processed in more careful way and provides more samples than any other dataset in the area.
scb-mt-en-th-2020: A Large English-Thai Parallel Corpus
The primary objective of our work is to build a large-scale English-Thai dataset for machine translation.
We construct an English-Thai machine translation dataset with over 1 million segment pairs, curated from various sources,
namely news, Wikipedia articles, SMS messages, task-based dialogs, web-crawled data and government documents.
Methodology for gathering data, building parallel texts and removing noisy sentence pairs are presented in a reproducible manner.
We train machine translation models based on this dataset. Our models' performance are comparable to that of
Google Translation API (as of May 2020) for Thai-English and outperform Google when the Open Parallel Corpus (OPUS) is
included in the training data for both Thai-English and English-Thai translation.
The dataset, pre-trained models, and source code to reproduce our work are available for public use.
Dataset with the text of 10% of questions and answers from the Stack Overflow programming Q&A website.
This is organized as three tables:
Questions contains the title, body, creation date, closed date (if applicable), score, and owner ID for all non-deleted Stack Overflow questions whose Id is a multiple of 10.
Answers contains the body, creation date, score, and owner ID for each of the answers to these questions. The ParentId column links back to the Questions table.
Tags contains the tags on each of these questions.
An unannotated Spanish corpus of nearly 1.5 billion words, compiled from different resources from the web.
This resources include the spanish portions of SenSem, the Ancora Corpus, some OPUS Project Corpora and the Europarl,
the Tibidabo Treebank, the IULA Spanish LSP Treebank, and dumps from the Spanish Wikipedia, Wikisource and Wikibooks.
This corpus is a compilation of 100 text files. Each line of these files represents one of the 50 million sentences from the corpus.
The Serbian web corpus srWaC was built by crawling the .rs top-level domain in 2014. The corpus was near-deduplicated on paragraph level, normalised via diacritic restoration, morphosyntactically annotated and lemmatised. The corpus is shuffled by paragraphs. Each paragraph contains metadata on the URL, domain and language identification (Serbian vs. Croatian).
Version 1.0 of this corpus is described in http://www.aclweb.org/anthology/W14-0405. Version 1.1 contains newer and better linguistic annotations.
A freely available paraphrase corpus for 73 languages extracted from the Tatoeba database. Tatoeba is a crowdsourcing project mainly geared towards language learners. Its aim is to provide example sentences and translations for particular linguistic constructions and words. The paraphrase corpus is created by populating a graph with Tatoeba sentences and equivalence links between sentences “meaning the same thing”. This graph is then traversed to extract sets of paraphrases. Several language-independent filters and pruning steps are applied to remove uninteresting sentences. A manual evaluation performed on three languages shows that between half and three quarters of inferred paraphrases are correct and that most remaining ones are either correct but trivial, or near-paraphrases that neutralize a morphological distinction. The corpus contains a total of 1.9 million sentences, with 200 – 250 000 sentences per language. It covers a range of languages for which, to our knowledge,no other paraphrase dataset exists.
Arabic vocalized texts.
it contains 75 million of fully vocalized words mainly97 books from classical and modern Arabic language.
This news dataset is a persistent historical archive of noteable events in the Indian subcontinent from start-2001 to mid-2020, recorded in realtime by the journalists of India. It contains approximately 3.3 million events published by Times of India. Times Group as a news agency, reaches out a very wide audience across Asia and drawfs every other agency in the quantity of english articles published per day. Due to the heavy daily volume over multiple years, this data offers a deep insight into Indian society, its priorities, events, issues and talking points and how they have unfolded over time. It is possible to chop this dataset into a smaller piece for a more focused analysis, based on one or more facets.
This parallel corpus consists of manually translated UN documents from the last 25 years (1990 to 2014) for the six official UN languages, Arabic, Chinese, English, French, Russian, and Spanish.
A dataset of atomic wikipedia edits containing insertions and deletions of a contiguous chunk of text in a sentence. This dataset contains ~43 million edits across 8 languages.
An atomic edit is defined as an edit e applied to a natural language expression S as the insertion, deletion, or substitution of a sub-expression P such that both the original expression S and the resulting expression e(S) are well-formed semantic constituents (MacCartney, 2009). In this corpus, we release such atomic insertions and deletions made to sentences in wikipedia.
The Wikicorpus is a trilingual corpus (Catalan, Spanish, English) that contains large portions of the Wikipedia (based on a 2006 dump) and has been automatically enriched with linguistic information. In its present version, it contains over 750 million words.
Large scale, unlabeled text dataset with 39 Million tokens in the training set. Inspired by the original WikiText Long Term Dependency dataset (Merity et al., 2016). TL means "Tagalog." Originally published in Cruz & Cheng (2019).
Yahoo! Answers Topic Classification is text classification dataset. The dataset is the Yahoo! Answers corpus as of 10/25/2007. The Yahoo! Answers topic classification dataset is constructed using 10 largest main categories. From all the answers and other meta-information, this dataset only used the best answer content and the main category information.