https://mousse-project.org/Natural language processing, explained with sourceshttps://mousse-project.org/bias-in-nlp/https://mousse-project.org/bias-in-nlp/2026-10-26A sourced introduction to bias in natural language processing: findings from word embeddings, coreference, machine translation and hate speech detection, the difference between representational and allocational harms, the limits of debiasing, and practical steps for teams.https://mousse-project.org/dataset-licences-for-nlp/https://mousse-project.org/dataset-licences-for-nlp/2026-10-26A practical guide to licensing text datasets in NLP research: Creative Commons and CC0, ShareAlike and NonCommercial clauses, Open Data Commons, LDC and ELRA agreements, EU database rights and text and data mining exceptions, and how to document licences.https://mousse-project.org/datasheets-data-statements-model-cards/https://mousse-project.org/datasheets-data-statements-model-cards/2026-10-26What datasheets for datasets, data statements for NLP and model cards are, where they came from, what they ask, how they relate to the EU AI Act's documentation duties, and how to write useful ones without boilerplate.https://mousse-project.org/european-language-technology/https://mousse-project.org/european-language-technology/2026-10-26How Europe approaches language technology for its 24 official languages and many regional ones: the warnings of the META-NET white papers, the European Parliament's call for language equality, the European Language Grid, eTranslation, data collection and how to find EU-funded NLP research on CORDIS.https://mousse-project.org/evaluating-nlp-models/https://mousse-project.org/evaluating-nlp-models/2026-10-26A practical guide to evaluating natural language processing systems: test sets and splits, accuracy, precision, recall and F1, benchmark suites such as GLUE and HELM, significance testing, annotation artefacts and data contamination.https://mousse-project.org/gdpr-and-text-data/https://mousse-project.org/gdpr-and-text-data/2026-10-26How the EU General Data Protection Regulation applies to text corpora and language models: what counts as personal data in text, lawful bases and the research provisions, why anonymising text is hard, what the EDPB said about AI models in 2024, and a practical checklist.https://mousse-project.org/how-tokenization-works/https://mousse-project.org/how-tokenization-works/2026-10-26A clear guide to tokenization in natural language processing: why models split text into subword units, how BPE, WordPiece, unigram and SentencePiece differ, why some languages cost more tokens than others, and what to check in practice.https://mousse-project.org/low-resource-languages/https://mousse-project.org/low-resource-languages/2026-10-26Why most of the world's languages are poorly served by language technology, how researchers classify resource levels, the problems with web-crawled data for small languages, and the methods and community-led projects that are changing the picture.https://mousse-project.org/machine-translation-evaluation/https://mousse-project.org/machine-translation-evaluation/2026-10-26How machine translation quality is measured: BLEU and its pitfalls, chrF, learned metrics such as COMET, direct assessment and MQM human evaluation, quality estimation without references, and how to compare two systems fairly.https://mousse-project.org/machine-translation-for-tenants-moving-abroad/https://mousse-project.org/machine-translation-for-tenants-moving-abroad/2026-10-26A practical, research-based guide for people renting a home in another country with the help of machine translation: what translation tools handle well in listings and leases, where they fail (rental jargon, false friends, numbers, legal terms), how to check key clauses, and how to protect your personal data.https://mousse-project.org/multilingual-lexical-knowledge-bases/https://mousse-project.org/multilingual-lexical-knowledge-bases/2026-10-26How lexical knowledge bases link words and meanings across languages: Princeton WordNet, EuroWordNet and the Open Multilingual Wordnet, BabelNet, Wikidata and ConceptNet, what they are used for in NLP, their licences and their blind spots.https://mousse-project.org/parallel-corpora/https://mousse-project.org/parallel-corpora/2026-10-26What parallel corpora are, how sentences are aligned, the landmark collections from the Canadian Hansard and Europarl to EU translation memories, OPUS and ParaCrawl, how bitext is mined from the web, and the quality and translationese problems to watch for.https://mousse-project.org/reproducibility-in-nlp/https://mousse-project.org/reproducibility-in-nlp/2026-10-26Why natural language processing results are often hard to reproduce, from preprocessing details and random seeds to changing API models, what reproducibility studies have found, the checklists conferences now use, and a practical routine for researchers.https://mousse-project.org/web-scraped-corpora-copyright-consent/https://mousse-project.org/web-scraped-corpora-copyright-consent/2026-10-26How language models are trained on web crawls such as Common Crawl, what the EU text and data mining exceptions and the AI Act require, how robots.txt and opt-outs work, what audits of crawled corpora found, and the ethical questions consent raises.https://mousse-project.org/what-makes-a-language-model-open/https://mousse-project.org/what-makes-a-language-model-open/2026-10-26Open weights, open source and open science are not the same thing. A guide to the components of a language model release, the Open Source AI Definition 1.0, common licences with use restrictions, and how the EU AI Act treats open-source models.https://mousse-project.org/word-sense-disambiguation/https://mousse-project.org/word-sense-disambiguation/2026-10-26What word sense disambiguation is, from Lesk's dictionary overlaps to contextual embeddings, how it is evaluated with WordNet, SemCor and the Senseval/SemEval benchmarks, and why ambiguity still trips up translation and search.