Embeddings

From Clarin K-Centre
Revision as of 12:47, 26 March 2024 by Floyd (talk | contribs) (Created page with "* [https://github.com/coosto/dutch-word-embeddings Github pagina]")
Jump to navigation Jump to search

Voor Large Language Models (LLM), verwijzen wij naar Taalmodellering.

Word2Vec embeddings

Database van de word embeddings die zijn beschreven in het paper 'Evaluating Unsupervised Dutch Word Embeddings as a Linguistic Resource', dat werd gepresenteerd bij LREC in 2016.

FastText embeddings

Woord-vectors in 157 talen, getraind op CommonCrawl en Wikipedia-corpora.

Coosto embeddings

Deze database bevat een Word2Vec-model dat is getraind op een groot Nederlands corpus, bestaande uit social-media berichten en posts van Nederlands nieuws, blogs en fora.

GeenStijl.nl embeddings

GeenStijl.nl embeddings contains over 8M messages from the controversial Dutch websites GeenStijl and Dumpert to train a word embedding model that captures the toxic language representations contained in the dataset. The trained word embeddings (±150MB) are released for free and may be useful for further study on toxic online discourse.

NLPL Word Embeddings Repository

Made by the University of Oslo. Models trained with clearly stated hyperparametes, on clearly described and linguistically pre-processed corpora.

For Dutch, Word2Vec and ELMO embeddings are available.