About Parallel Text
Parallel Text is an independent English-language publication that explains natural language processing (NLP): how computers analyse, translate and generate human language, how that work is evaluated, and the questions it raises about openness, fairness and data. It is written for students, researchers from neighbouring fields, journalists, policy staff and curious readers.
What we publish
- Methods and evaluation: tokenization, word sense disambiguation, how models are evaluated and machine translation metrics.
- Open research: open models, dataset licences, reproducibility and documentation.
- Multilingual and low-resource NLP: low-resource languages, parallel corpora, multilingual knowledge bases, language technology in Europe and a practical guide to machine translation for tenants moving abroad.
- Ethics and data: bias, personal data in text and web-scraped corpora and consent.
How articles are researched
- Primary sources first. Peer-reviewed papers (we cite the ACL Anthology, the open archive of the Association for Computational Linguistics, wherever possible), official texts of EU law on EUR-Lex, and documents from public institutions and standards bodies. Every article ends with its sources.
- Dated facts. Laws, tools and datasets change. We give the year of each finding and the date on which we consulted web pages, so you can tell how current a statement is.
- No invented details. If a figure, a quotation or a result cannot be followed back to where it came from, it does not run, and illustrative examples are never dressed up as real cases. Shaky sources get flagged as shaky in the text, or the point is dropped. As for authorship, there are no made-up writers: the byline is Parallel Text itself.
- Not legal advice. Articles on copyright, licences and data protection explain the general framework. For a specific project, consult your institution's legal or data protection officer.
- Versioned like a paper. Each article has a publication date, and an "Updated" date records every later correction or revision.
Independence
No lab, model vendor or funder shapes what Parallel Text writes. It has no affiliation with any university, research project, company, funding body or public institution, sells nothing, uses no affiliate links and takes no sponsored articles or paid placements. The legal notice identifies the company responsible for it.
About this domain
The domain mousse-project.org previously hosted the website of MOUSSE ("Multilingual, Open-text Unified Syntax-independent SEmantics"), a research project in natural language processing funded by a European Research Council Consolidator Grant (grant agreement 726487) and hosted by Sapienza University of Rome. According to its CORDIS fact sheet, the project ran from 1 June 2017 to 31 May 2023. Archived copies of the website show it online from at least August 2018 until June 2026. Parallel Text is a new, separate publication: it has no connection with the MOUSSE project, its researchers, Sapienza University of Rome, the European Research Council or the European Commission, and it has not reused any of the former site's text, logo, images or other material. Because other websites still link to this domain, a few old addresses now lead to the closest articles in this publication.
Contact and corrections
Spotted a misattributed result or a rule that has since changed? Report it through the contact page.