--- title: "Managing Growth in a National Corpus: The Hungarian National" description: "The paper sets out the design principles and the curated, domain-by-domain data collection, including filtered Common Crawl… 3 references." authors: ["Noémi Ligeti-Nagy", "Enikő Héja", "Ágnes Bánfi", "Flóra Földesi", "Bence Sárossy", "Boglárka Skrabák", "Tamás Váradi", "Gábor Prószéky"] year: 2026 venue: "Proceedings of the 12th Workshop on Challenges in the Management of Large Corpora" source: "https://aclanthology.org/2026.cmlc-1.14/" doi: "10.63317/46aafswuwv4m" licence: "CC BY-NC 4.0 (ELRA, LREC proceedings)" licence_url: "https://creativecommons.org/licenses/by-nc/4.0/" summary_licence: "Elamur Cosmic License" canonical: "https://research.elamur.ai/free/en/papers/nagy-2026-managing-growth-national-corpus-hungarian/" --- [Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/nagy-2026-managing-growth-national-corpus-hungarian/) [home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2026](https://md.research.elamur.ai/free/en/papers/2026/page/35/) / Managing Growth in a National Corpus: The Hungarian National Corpus 3.0 (MNSZ3) # Managing Growth in a National Corpus: The Hungarian National Corpus 3.0 (MNSZ3) **Authors:** [Noémi Ligeti-Nagy](https://md.research.elamur.ai/free/en/authors/noemi-ligeti-nagy/), Enikő Héja, Ágnes Bánfi, Flóra Földesi, Bence Sárossy, Boglárka Skrabák, Tamás Váradi and Gábor Prószéky · **Year:** 2026 · **Venue:** [Proceedings of the 12th Workshop on Challenges in the Management of Large Corpora](https://md.research.elamur.ai/free/en/venues/cmlc/) ([Workshop on Challenges in the Management of Large Corpora 2026](https://md.research.elamur.ai/free/en/venues/cmlc/2026/)) · **Publisher:** [European Language Resources Association (ELRA)](https://md.research.elamur.ai/free/en/publishers/european-language-resources-association-elra/) · **Licence:** [CC BY-NC 4.0 (ELRA, LREC proceedings)](https://creativecommons.org/licenses/by-nc/4.0/) **Links:** [Publisher's page](https://aclanthology.org/2026.cmlc-1.14/) · [PDF](https://aclanthology.org/2026.cmlc-1.14.pdf) · [DOI 10.63317/46aafswuwv4m](https://doi.org/10.63317/46aafswuwv4m) · [OpenAlex](https://openalex.org/W7166514145) **Citations:** 0 citations (OpenAlex, 2026-10-14) ## Summary MNSZ3 is the Hungarian National Corpus in its third iteration. It aims to enlarge its predecessor MNSZ2, the Hungarian Gigaword Corpus of about 1.04 billion running words, roughly tenfold to around 10 billion tokens while keeping the balanced, domain-based model. MNSZ3 keeps MNSZ2's six domains and their proportions: press, fiction, scientific, official, personal and transcribed spoken. It also adds much more Hungarian text from neighbouring countries, with richer metadata. The paper sets out the design principles and the curated, domain-by-domain data collection, including filtered Common Crawl press pages, targeted portal harvesting and parliamentary/institutional documents. It also covers deduplication, boilerplate removal and domain-specific dating rules. For annotation, the authors compared the HuSpaCy and e-magyar pipelines and found that each does different things well, with no overall winner. As a result, HuSpaCy serves as the main UD-oriented processing pipeline, and e-magyar's emMorph output is kept as a separate layer for morphology and lemmatisation. The project offers a reproducible, versioned model for scaling a curated national corpus for linguists and NLP developers. ## Key points - **Scale and regions:** MNSZ2 totals about 1039.7M words, of which 1013.9M come from Hungary and only small amounts from the other four regions. These are Slovakia (17.3M), Transylvania (3.9M), Transcarpathia (2.5M) and Vojvodina (2.0M). MNSZ3 targets roughly 10 billion tokens, with special focus on region–domain combinations that are missing or marginal. - **Time span:** Temporal coverage is set per domain. Around 1990 is generally the lower limit, but fiction may go back to 1945. Translations are dated by their year of translation. - **Named sources:** These include the Hungarian part of DCEP (~35M words) and additional Hungarian parliamentary documents and minutes (~79M words), which carry sub-labels within the official domain. - **Pipeline comparison:** HuSpaCy was stronger in tokenisation, dependency parsing and NER. e-magyar was better at morphology and lemmatisation and more precise in assigning entity types. Some e-magyar parses lacked a ROOT relation, which causes problems for downstream processing. - **Final annotation design:** HuSpaCy handles tokenisation, POS/UD morphology, dependency parsing and NER, and emMorph analyses are kept as an extra layer. A deeper HuSpaCy–emMorph integration is being developed in parallel. *Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.* ## Institutions - [Eötvös Loránd University](https://md.research.elamur.ai/free/en/universities/eotvos-lorand-university/) — University, Budapest, Hungary · 5 works · 256 citations - [ELTE Research Centre for Linguistics](https://md.research.elamur.ai/free/en/labs/elte-research-centre-for-linguistics/) — Research institute, Budapest, Hungary · 2 works 📚 Shelves (5): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 16,667 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 16,194 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 16,039 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 15,891 works · [📚 Literature and Literary Theory](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/literature-and-literary-theory/) · 519 works ## Cite Cite as: Noémi Ligeti-Nagy, Enikő Héja, Ágnes Bánfi, Flóra Földesi, Bence Sárossy, Boglárka Skrabák, Tamás Váradi and Gábor Prószéky. 2026. Managing Growth in a National Corpus: The Hungarian National Corpus 3.0 (MNSZ3). Proceedings of the 12th Workshop on Challenges in the Management of Large Corpora. The paper itself, not this page. BibTeX from the library's record of the work. ```bibtex @inproceedings{ligetinagy2026managing, title = {{Managing Growth in a National Corpus: The Hungarian National Corpus 3.0 (MNSZ3)}}, author = {Noémi Ligeti-Nagy, Enikő Héja, Ágnes Bánfi, Flóra Földesi, Bence Sárossy, Boglárka Skrabák, Tamás Váradi and Gábor Prószéky}, booktitle = {Proceedings of the 12th Workshop on Challenges in the Management of Large Corpora}, year = {2026}, doi = {10.63317/46aafswuwv4m}, url = {https://doi.org/10.63317/46aafswuwv4m} } ``` - [← previous MameLoshnLM: Yiddish Language Model and Evaluation Benchmark](https://md.research.elamur.ai/free/en/papers/katz-2026-mameloshnlm-yiddish-language-model-evaluation/) - [next → ManCC: A Task-Anchored Benchmark for Manchu–Classical Chinese Cross-Lingual Modeling](https://md.research.elamur.ai/free/en/papers/wang-2026-mancc-task-anchored-benchmark-manchu/) - [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/35/) ## References 3 references, 0 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet. 1. BNC Consortium. 2007. British National Corpus, XML Edition. Available at [http://www.natcorp.ox.ac.uk/](http://www.natcorp.ox.ac.uk/),. [↗](http://www.natcorp.ox.ac.uk/) 2. Named entity recognition. HuSpaCy tended to identify more named-entity tokens, while e-magyar was more conservative but more precise in type assignment on the manually checked sample. 4.2. \#\# Conclusion V. Brezina, A. Hawtin, and T. McEnery. 2021. The written British National Corpus 2014 – design an… [find](https://scholar.google.com/scholar?q=Named+entity+recognition%2E+HuSpaCy+tended+to+identify+more+named%2Dentity+tokens%2C+while+e%2Dmagyar+was+more+conservative+but+more+precise+in+type+assignment+on+the+manually+checked+sample%2E+4%2E2%2E+%5C%23%5C%23+Conclu) 3. Annotation strategy Milena Hnátková, Michal Křen, Pavel Procházka, and Hana Skoumalová. 2014. The SYN-series corpora of written Czech. In Proceedings of the Ninth International Conference on Language Resources and Evaluation (LREC’14), pages 160– 164. European Language Resources Association (ELRA). [find](https://scholar.google.com/scholar?q=Annotation+strategy+Milena+Hn%C3%A1tkov%C3%A1%2C+Michal+K%C5%99en%2C+Pavel+Proch%C3%A1zka%2C+and+Hana+Skoumalov%C3%A1%2E+2014%2E+The+SYN%2Dseries+corpora+of+written+Czech%2E+In+Proceedings+of+the+Ninth+International+Conference+on+Language+) *Updated 1 November 2026*