📰 Venues

A guide · about 11 minutes to read

A venue is where a scientific paper first appears in public: a journal, the proceedings of a conference, or the volume of a workshop. The venue decides who reviews the paper, when it comes out, which readers find it and, to a large degree, how the work is remembered. In computational linguistics and artificial intelligence the venue is unusually important, because these fields publish their main results at conferences rather than in journals, and because a few series of meetings have set the agenda of the whole discipline for decades.

This page gathers the venues where the works of this library were published. Most of them belong to the family of meetings run by the Association for Computational Linguistics: the Annual Meeting of the Association for Computational Linguistics (ACL), the Conference on Empirical Methods in Natural Language Processing (EMNLP), the North American chapter conference (NAACL) and the journal Computational Linguistics. Around them stand resource conferences such as LREC, evaluation series such as SemEval and WMT, and the large machine learning and computer vision meetings whose papers the language fields now read as closely as their own.

The page is a companion to the indexes of the institutions behind the works — universities, labs and research institutes, companies and all institutions — and to the geographic indexes of countries and agglomerations. Those pages answer who wrote a work and where; this one answers where the work was put before its readers.

What a venue is

The word covers several kinds of outlet that differ in rhythm, review and prestige.

Journals

A journal publishes articles continuously or in issues, usually after one or more rounds of review in which the authors revise and resubmit. There is no fixed deadline and no meeting; the article is the end product. The field's own journal is Computational Linguistics, which has carried long, carefully reviewed articles since the 1970s. The Transactions of the Association for Computational Linguistics (TACL) is a newer hybrid: a journal with fast review whose accepted papers are also presented in person at the association's conferences. General science journals such as Nature, Science and the Proceedings of the National Academy of Sciences appear here only rarely, for results that reach beyond the field. Survey journals such as ACM Computing Surveys collect overviews of a whole area.

Conferences and their proceedings

A conference series meets on a regular schedule, sets a submission deadline, reviews the submissions in a few weeks and publishes the accepted papers as proceedings. In most sciences proceedings are minor; in computational linguistics and machine learning they are the main record. A paper at ACL, EMNLP or Neural Information Processing Systems counts for as much as, or more than, a journal article, and many careers are built on such papers alone. The reasons are speed — a fast-moving field cannot wait a year for a journal — and the meeting itself, where results are discussed, contested and taken up within months.

Workshops and shared tasks

Workshops are smaller meetings, usually one or two days attached to a main conference, devoted to one problem or community: the Workshop on Abusive Language Online, the Workshop on Noisy User-generated Text, the Workshop on Gender Bias in Natural Language Processing or the BlackboxNLP workshop on analysing neural networks. Some workshops grew into permanent series with their own identity, such as SemEval, the Workshop on Statistical Machine Translation and its successor the Conference on Machine Translation (WMT), or the Workshop on Innovative Use of NLP for Building Educational Applications (BEA). Many of them organise shared tasks, where teams solve the same problem on the same data and report in short system papers.

Peer review and the ACL Anthology

Almost all of these venues use peer review: submissions are read by other researchers, usually without seeing the authors' names, and accepted or rejected on their advice. The association's conferences have long used double-blind review and, since the early 2020s, a shared rolling review cycle that feeds several conferences at once. The Findings of the Association for Computational Linguistics, introduced in 2020, publishes sound papers that did not get a slot in the main programme.

Nearly every paper of these meetings, together with the journal, is freely available in the ACL Anthology, an open archive maintained by the association and its volunteers since the early 2000s. It is one reason why the field's literature is unusually easy to read, cite and mine, and why the works of this library lean so heavily on its venues.

A short history of the venues

The oldest venues here are tied to machine translation. The Association for Computational Linguistics was founded in 1962 as an association for machine translation and computational linguistics, and took its present name in 1968; its annual meeting has run ever since. The International Conference on Computational Linguistics (COLING) began in the mid-1960s as a biennial international meeting and remains one of the field's oldest series. The journal now called Computational Linguistics started in the 1970s and took its current name in the 1980s. The European chapter (EACL) held its first conference in 1983, and the Conference on Applied Natural Language Processing (ANLP) gave practical systems a venue of their own from the late 1980s to 2000. Translation kept specialised outlets too, such as the TMI conference series, the Association for Machine Translation in the Americas (AMTA) and the European Association for Machine Translation (EAMT).

In the late 1980s and the 1990s American funding agencies shaped a second family. The Message Understanding Conferences (MUC) measured information extraction systems against common tasks, and the DARPA Workshop on Speech and Natural Language and the Human Language Technology workshops brought speech and text researchers together. Their successor, the Human Language Technology Conference (HLT), later merged with NAACL.

The statistical turn of the 1990s produced its own meetings. Workshops on very large corpora led to the Joint SIGDAT Conference on Empirical Methods and Very Large Corpora and then to EMNLP, first held in 1996. The Conference on Computational Natural Language Learning (CoNLL) followed in 1997, and LREC, first held in Granada in 1998, gave corpora, lexicons and annotation a home of their own. Word sense evaluations of the late 1990s grew into SemEval. The North American chapter (NAACL) held its first conference in 2000.

The 2010s added journal-speed publishing with TACL in 2013, regional chapters such as the Asia-Pacific chapter (AACL), and many new workshops on ethics, social media and interpretability. They also pulled the language fields towards general machine learning: neural models made Neural Information Processing Systems, the International Conference on Machine Learning and the AAAI conference natural outlets for language research, and vision–language work brought in CVPR and ICCV.

How venues shaped the fields

Shared tasks and common evaluation

Shared tasks are perhaps the venues' largest contribution. MUC defined named entities as a task; CoNLL's yearly challenges fixed standard settings for named entity recognition and parsing, as in the CoNLL-X Shared Task on Multilingual Dependency Parsing. The translation workshops ran yearly competitions whose overview papers, such as Findings of the 2014 Workshop on Statistical Machine Translation and Findings of the 2019 Conference on Machine Translation (WMT19), became the reference record of progress. SemEval did the same for meaning and opinion, with tasks such as SemEval-2014 Task 4: Aspect Based Sentiment Analysis and SemEval-2019 Task 5 on hate speech.

Metrics

Because shared tasks need scores, the venues also became the place where metrics were proposed and argued over. Bleu: a Method for Automatic Evaluation of Machine Translation appeared at ACL; METEOR at a dedicated ACL workshop on evaluation measures; chrF and chrF++ at the translation workshops; Translation Edit Rate at AMTA. Critical papers such as A Call for Clarity in Reporting BLEU Scores were published at the WMT research track.

Benchmarks and resources

Benchmarks travel through venues too. SQuAD came out at EMNLP, GLUE at a workshop, and Natural Questions in TACL. Resource papers have their own natural home at LREC, which published SentiWordNet 3.0 and the conversion of phrase structure into typed dependency parses. The journal carried The Proposition Bank, while toolkits such as Stanford CoreNLP, fairseq and Transformers were presented in system demonstration tracks.

The best-paper canon

Finally, the venues decide which papers the field treats as landmarks. Each conference awards best papers and, more lastingly, test-of-time prizes to older papers whose influence proved durable. The resulting canon is visible in the works here: BERT at NAACL, BART at ACL, Convolutional Neural Networks for Sentence Classification at EMNLP, or Word Association Norms, Mutual Information, and Lexicography in the journal.

The leading venues

The cards above this guide and the figure below show the venues in the order of the site's ranking, which weighs each work by its citations.

Annual Meeting of the Association for Computational Linguistics (ACL). The flagship of the field and the leading venue here, with works stretching from the late 1970s to today. Its most cited works here include BLEU, BART and Stanford CoreNLP, and its topics run from language models and pretraining and machine translation to coreference resolution and sentiment.

EMNLP. Born from the statistical movement of the 1990s, EMNLP is now as large as ACL itself. Its works here include Thumbs up? Sentiment Classification using Machine Learning Techniques, SQuAD, the Transformers library and Kim's convolutional sentence classifier.

NAACL. The North American chapter's conference published BERT, Hierarchical Attention Networks for Document Classification and earlier statistical work such as feature-rich part-of-speech tagging.

Computational Linguistics. The field's journal favours long articles that consolidate a method or resource: A Maximum Entropy Approach to Natural Language Processing, Lexicon-Based Methods for Sentiment Analysis and PropBank are among its most cited works here.

CVPR and ICCV. The two main computer vision conferences enter the list through a small number of highly cited works that language research builds on, such as Momentum Contrast, the Swin Transformer and Sigmoid Loss for Language Image Pre-Training.

Nature. A general science journal with a handful of works here, including AlphaFold, NumPy and DeepSeek-R1.

LREC. The conference of the European Language Resources Association is the home of corpora, lexicons and evaluation data, and one of the largest venues here by the number of works. It leans strongly on Semitic languages, multilingual NLP and evaluation.

TACL. The association's open-access journal published Natural Questions, SpanBERT and Lost in the Middle, a study of how language models use long contexts.

Findings. The youngest of the association's series, focused on language models and reasoning; its works here include CodeBERT and RealToxicityPrompts.

COLING. The international biennial conference, strong on discourse and coreference, with works such as Determining the Sentiment of Opinions and Contextual String Embeddings for Sequence Labeling.

Neural Information Processing Systems. The main machine learning conference appears here through the recent wave of language model research: Large language models are zero-shot reasoners, Toolformer and wav2vec 2.0.

SemEval, EACL and CoNLL. Three long-running series of the association's family: SemEval for shared evaluation tasks, above all on sentiment; EACL for European work, including Exploiting Cloze-Questions for Few-Shot Text Classification and the brat annotation tool; CoNLL for learning-based methods, Universal Dependencies and dependency parsing, with works such as Generating Sentences from a Continuous Space.

Outside the association's family, journals such as the Journal of Artificial Intelligence Research, Machine Learning and IEEE Transactions on Pattern Analysis and Machine Intelligence appear through a few classic papers such as Bagging Predictors and SMOTE. Speech research comes in through Interspeech and the IEEE/ACM Transactions on Audio, Speech and Language Processing, and retrieval through SIGIR.

Topics and fields

What the venues publish here reflects both the field and the shape of this library. Taken together, their works lean most on Semitic languages and Hebrew and Aramaic, much of it from resource venues such as LREC, then on machine translation, language models and pretraining, coreference and evaluation. Further down come sentiment and emotion, discourse, dialogue systems, morphology, question answering, vision and language, speech and bias, fairness and safety.

The venues also divide clearly by field. The association's conferences and journal sit mostly on the natural language processing shelf, with a solid share on linguistics and languages and a smaller one on machine translation. LREC, by contrast, falls almost entirely on the linguistics shelf, as befits a venue of corpora and lexicons. The vision conferences and the general science journals lie on the AI and machine learning, mathematics and statistics and vision shelves, or on none of the language shelves at all. The full list of fields is on the shelves page, and the full list of themes on the topics page.

Specialised venues show the same pattern on a smaller scale. Translation series such as WMT, the WMT shared task papers and AMTA publish almost only on translation and its evaluation; the SIGDIAL meeting and Dialogue & Discourse on conversation; the Linguistic Annotation Workshop on annotation; VarDial and the Arabic NLP workshop (WANLP) on dialects and particular languages.

Trends over time

The yearly figure follows the first venues of the ranking on one common scale.

The works here from ACL, EMNLP and NAACL concentrate in the late 2010s and the early 2020s, the years when neural models and pretraining transformed the field and the conferences grew to many times their earlier size. ACL's tallest year here falls around 2020; EMNLP peaks at about the same time. NAACL shows gaps in some years, because the chapter conference is not held every year and in some years merges with other meetings. After the early 2020s, output here from these conferences becomes smaller year by year, partly because part of the field's work moved to Findings, TACL and the machine learning conferences, and partly because the most recent years of any library are still filling in. The journal Computational Linguistics contributes a small, steady stream across the decade. The vision conferences and Nature appear only in scattered years, each time with individual papers of wide reach rather than a steady flow.

How to use this page

• The cards at the top show the ten leading venues; each card opens the venue's own page.
• The figures show the leading venues, the topics of their works, the shares of each venue's works by field, and their yearly output. Each figure has a «The numbers» section with the exact figures behind it.
• The slices by field below divide the venues by shelf, so that, for instance, the venues of machine translation or of natural language processing can be compared on their own.
• The ranked list shows the venues in the site's order; its full version continues well past the first entries, into specialised series such as the Workshop on Machine Reading for Question Answering, the International Natural Language Generation Conference, the Nordic Conference of Computational Linguistics, RANLP or the Italian Conference on Computational Linguistics.
• Each venue's page gathers its works here, its most cited papers and its topics, with links to the papers and on to their authors and institutions.

The publishers behind these journals and proceedings have an index of their own. For the institutions that wrote the works, see the universities, labs, companies, nonprofits, government research bodies and hospitals pages; for individual researchers, the Paper Stars page.