🖨️ Publishers

A guide · about 9 minutes to read

A publisher is the body that puts a scholarly work into its public, citable form: it runs the review, fixes the text and the record, assigns the identifier and keeps the work available. In linguistics, natural language processing and machine translation the publisher is often not a commercial house at all, but a learned society that organizes the conferences of the field and issues their proceedings. This page lists the publishers of the works in this library.

The picture here is unusual compared with most sciences. One society, the Association for Computational Linguistics, publishes the larger part of the field's literature through its conferences and its open archive. Around it stand other societies, university presses, the large commercial houses, open-access publishers and a scattering of universities, libraries and workshop series that publish small, specialised collections.

The publisher is a different kind of entry from the authors' institutions. A paper written at a university, a lab or a company is published by a society or a press; the Universities, Labs and research institutes and Companies pages answer who did the work, this page answers who made it public.

What a publisher of scholarly work is

Learned societies

A learned society is a membership association of researchers. In computing and language research, societies publish mainly through conferences: a programme committee reviews submissions, accepted papers appear in proceedings, and the proceedings are the archival record. The Association for Computational Linguistics is the central example. Its neighbours in this library include the European Language Resources Association (ELRA), which organizes the LREC conference on language resources and evaluation; the International Committee on Computational Linguistics (ICCL), the body behind the COLING conferences; the Association for the Advancement of Artificial Intelligence; and the machine translation associations, the Association for Machine Translation in the Americas and the European Association for Machine Translation. Broader engineering and computing societies, the Institute of Electrical and Electronics Engineers, the IEEE Computer Society and the Association for Computing Machinery, publish the language work that appears at their conferences and in their journals. For linguistics proper, the Linguistic Society of America plays the society role.

Regional societies follow the same pattern: the Asian Federation of Natural Language Processing, the Chinese Information Processing Society of China, The Association for Computational Linguistics and Chinese Language Processing (ACLCLP), the Northern European Association for Language Technology, the NLP Association of India (NLPAI), the French ATALA and the Global Wordnet Association.

University presses

University presses publish books and journals on behalf of a university, usually as non-profit departments. Oxford University Press and Cambridge University Press are among the oldest publishers in the world. The MIT Press matters most to this library: it publishes the journal Computational Linguistics on behalf of the ACL, and Transactions of the ACL. Smaller presses appear as well: CSLI Publications at Stanford, Hamburg University Press, Firenze University Press, Presses universitaires de Caen and Linköping University Electronic Press. Several universities appear as publishers directly, often through their libraries or repositories, such as the University of Tartu, the University of Alberta and Cornell University, which operates the arXiv preprint server.

Commercial houses

The large commercial publishers carry journals and book series across all sciences. Here they include Springer Science+Business Media, Springer Nature and Springer International Publishing, Elsevier BV, Wiley, Taylor & Francis, Nature Portfolio, John Benjamins Publishing Company and Brill. For computational work they matter less than in most fields, but they publish much of linguistics, the general science journals and the journals of neighbouring disciplines.

Open-access publishers

Open-access publishers make every work free to read, usually under a Creative Commons licence, and cover their costs through article charges, institutional support or volunteer work. Examples here are the Public Library of Science, Ubiquity Press, Frontiers Media, the Multidisciplinary Digital Publishing Institute, De Gruyter Open and the AI Access Foundation, which publishes the Journal of Artificial Intelligence Research. Workshop series such as CEUR Workshop Proceedings offer free, low-cost publication for computer science workshops.

History of scholarly publishing in these fields

The modern scholarly journal began with the society journals of the seventeenth century, and the pattern of societies publishing their own work continued into the twentieth. Linguistics followed the classic model: the Linguistic Society of America, founded in 1924, has published its journal Language since 1925, and much of the discipline grew around society journals and books from presses in Europe and North America.

Computational linguistics took a different path. The ACL was founded in 1962, first as the Association for Machine Translation and Computational Linguistics, and took its present name in 1968. Its journal, later called Computational Linguistics, became the field's journal of record, while the annual meeting and its printed proceedings carried most new work. COLING, under the ICCL, began in the 1960s as an international conference series. Computer science as a whole developed the same habit: in fast-moving areas, peer-reviewed conference proceedings became the main place to publish, and journals took a slower, archival role.

The 1990s and 2000s added new societies and new formats. The machine translation associations were formed in the early 1990s, ELRA was founded in 1995 and held the first LREC in 1998, and arXiv, started in 1991, gave researchers a way to share preprints before review. Around 2001–2002 the ACL began its Anthology, a free online archive of the papers of its conferences and journals, later extended to the events of related societies. Computational Linguistics became open access in 2009, and the MIT Press began Transactions of the ACL in 2013, a journal whose papers can be presented at the conferences.

In linguistics a parallel shift came through new open-access journals. Glossa, launched in 2016 with Ubiquity Press after the editorial board of a commercial journal left its publisher, is the best-known case and explains why Ubiquity Press is prominent here for theoretical linguistics.

How publishers shaped the fields

The choice of publisher had practical effects on what researchers could read and build on. Because the ACL Anthology is free, almost the whole of the field's conference literature has been open for two decades; papers such as Bleu: a Method for Automatic Evaluation of Machine Translation and SQuAD: 100,000+ Questions for Machine Comprehension of Text reached everyone who wanted to use them, not only readers at subscribing libraries. The open archive also made the literature itself a dataset, used for studies of citation, terminology and the history of the field.

Society publishing kept the field close to its own resources. ELRA, as both a publisher and a distributor of language resources, made the description of corpora, lexicons and annotation schemes a recognised kind of publication; works such as SentiWordNet 3.0: An Enhanced Lexical Resource for Sentiment Analysis and Opinion Mining and Generating Typed Dependency Parses from Phrase Structure Parses are typical of what appears under it. Proceedings from the ICCL, such as The Stanford Typed Dependencies Representation, carried similar standards work.

Licensing followed the same direction. The ACL moved its Anthology to Creative Commons licences, and the open-access houses publish under such licences by default, which allows text mining and redistribution. The commercial houses and the general science journals remain important where language research meets other sciences: Highly accurate protein structure prediction with AlphaFold and Array Programming with NumPy appear here through Nature Portfolio as tools and results that the language fields borrowed.

The leading publishers

Association for Computational Linguistics. The society of the field, based in the United States, and the leading publisher here by a wide margin. Its proceedings and journals cover nearly every topic, with Machine translation, Coreference resolution and Evaluation & benchmarks prominent. Among its most cited works here are Transformers: State-of-the-Art Natural Language Processing and BART: Denoising Sequence-to-Sequence Pre-training for Natural Language Generation, Translation, and Comprehension.

Nature Portfolio. The publisher of Nature and its sister journals, based in the United Kingdom. Few of its works are here, but they are heavily cited across science, including SciPy 1.0 — Fundamental Algorithms for Scientific Computing in Python and Stop Explaining Black Box Machine Learning Models for High Stakes Decisions and Use Interpretable Models Instead.

The MIT Press. The press of the Massachusetts Institute of Technology, publisher of the field's main journals. Its works here range from A Maximum Entropy Approach to Natural Language Processing and Lexicon-Based Methods for Sentiment Analysis to Natural Questions: A Benchmark for Question Answering Research and Latent Dirichlet Allocation.

Institute of Electrical and Electronics Engineers and IEEE Computer Society. The engineering society and its computing branch publish the computer vision and machine learning work that language research draws on, such as Rich feature hierarchies for accurate object detection and semantic segmentation, Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks and Billion-scale similarity search with GPUs.

Springer Science+Business Media. The German commercial house, publisher of machine learning journals and lecture-note series; its works here include Bagging Predictors, Text Categorization with Support Vector Machines: Learning with Many Relevant Features and End-to-End Object Detection with Transformers, with a notable share on Semitic languages.

European Language Resources Association (ELRA). The European association for language resources and the publisher of the LREC proceedings, second only to the ACL in the volume of works here. Its strongest topics are Semitic languages, Language models & pretraining and Multilingual & cross-lingual NLP; Twitter as a Corpus for Sentiment Analysis and Opinion Mining is among its most cited works here.

Association for Computing Machinery. The computing society publishes surveys and web research read across the field, such as Authoritative Sources in a Hyperlinked Environment, Survey of Hallucination in Natural Language Generation and Pre-train, Prompt, and Predict: A Systematic Survey of Prompting Methods in Natural Language Processing.

Elsevier BV. The Dutch commercial house, based in the Netherlands; here it carries The Anatomy of a Large-Scale Hypertextual Web Search Engine and Explanation in Artificial Intelligence: Insights from the Social Sciences.

AI Access Foundation. A non-profit that publishes the Journal of Artificial Intelligence Research, one of the early open-access journals; its works here include SMOTE: Synthetic Minority Over-sampling Technique and A Survey of Paraphrasing and Textual Entailment Methods.

International Committee on Computational Linguistics (ICCL). The committee behind COLING; its works here lean to Discourse analysis and Sentiment & emotion, with Determining the Sentiment of Opinions among the most cited here.

Association for the Advancement of Artificial Intelligence. The AI society, publisher of the AAAI conference proceedings, with works such as Character-Aware Neural Language Models and PIQA: Reasoning about Physical Commonsense in Natural Language.

Further down, the machine translation societies keep their own record: the Association for Machine Translation in the Americas published A Study of Translation Edit Rate with Targeted Human Annotation, and the European Association for Machine Translation published Estimating the Sentence-Level Quality of Machine Translation Systems.

Topics and fields

The topics of the publishers' works follow the shape of the library. Semitic languages leads, carried by ELRA, the commercial houses and the linguistics publishers, then Machine translation and Language models & pretraining. Close behind are Coreference resolution, Evaluation & benchmarks, Sentiment & emotion and Discourse analysis. Hebrew & Aramaic comes mostly from humanities publishers: Ubiquity Press, the University of Alberta, Oxford University Press and the Public Library of Science, which carries work on the Dead Sea Scrolls. The Linguistic Society of America stands out for Implicature and pragmatics.

By field, the Association for Computational Linguistics divides its works between Natural language processing and Linguistics & languages, with a steady part on Machine translation. ELRA sits almost entirely on the linguistics shelf, The MIT Press leans to natural language processing, and the general and engineering publishers spread into AI & machine learning, Computing & GPUs and Mathematics & statistics. Smaller shelves, Law, History & humanities and Education, draw on a wider set of publishers. The Shelves and Topics indexes list them all.

Trends over time

The ACL's output here is far larger than any other publisher's in every year and peaks around 2020, at the height of the expansion of its conferences. It falls back in the following years and rises again in the current year as new proceedings arrive. The MIT Press follows a gentler curve, strongest in the early 2020s. The commercial and engineering publishers appear more thinly, with works that cluster in the years of deep learning and large language models; the IEEE Computer Society, for example, shows a small cluster of works in 2023.

How to use this page

The cards above this guide show the ten leading publishers; each opens a publisher page with its works here, the topics they cover, the authors who published there and the institutions they came from. Each figure has a «The numbers» panel with the exact values it draws. Below the guide, slices by country and by field narrow the list, and the ranked list shows the publishers in order, with a link to its full version. For the conferences and journals behind each publisher, use the venues index; for the authors' side of the same works, see Institutions, all kinds, Countries, Agglomerations, Nonprofits, Government research bodies and Hospitals and medical centres. For the most cited works across the library, see Paper Stars.