From December 2024 to April 2026, the State Digital Solution Agency (SDSA) implemented the project ‘Creation of the General Lithuanian Language Corpus and Vectorised Models’ (project code: 02-103-P-0001) in which researchers from the VDU Institute of Digital Resources and Interdisciplinary Research, together with their business partners – UAB ‘Neurotechnology’, UAB ‘Tilde IT’, and MB ‘Krilas’ – developed artificial intelligence solutions of national significance.
The SDSA project manager is A. Rakauskas, and the supplier group leader is Assoc. Prof. Dr Andrius Utka.
Project outcomes:
● The General Lithuanian Language Corpus (BLKT) consists of original human-authored texts, the corpus size is 3.9 billion words, and the text types include Media texts (52.28%), Documents (38.29%), Non-fiction (7.08%), Speech (2.04%), and Fiction (0.30%).
● The Small Lithuanian Language Vectorised Model (LT-MLKM-modernBERT) is a Lithuanian masked language model (MLM) built on the ModernBERT architecture and pre-trained on the BLKT Lithuanian Text Corpus Stage 3, comprising over 1.87 billion words. This model is available on the Hugging Face platform at: https://huggingface.co/VSSA-SDSA/LT-MLKM-modernBERT.
● The Large Lithuanian Language Vectorised Model (DLKVM) was built using Llama 3 architecture principles and trained on 3.9 billion words from the BLKT corpus. This model is available on the Hugging Face platform at: https://huggingface.co/datasets/VSSA-SDSA/LT_AI_DLKVM.
More information can be found in this text: ‘A consortium led by VDU has developed a Lithuanian language corpus and two artificial intelligence models’. The article explains the project’s purpose as well as its importance for Lithuania, offers insights into the model development process, and presents the contributions of the partners involved in the project.
Further information about the project is available:
On the CLARIN-LT repository https://clarin-repo.lt/server/api/core/bitstreams/a2073e11-9c22-4aca-bfee-b2c60954f2a6/content
On the VDU SITTI Institute’s website https://sitti.vdu.lt/bendrasis-lietuviu-kalbos-tekstynas-ir-vektorizuoti-modeliai/
On the VSSA website https://vssa.lrv.lt/lt/apie-vssa/projektai/bendrojo-lietuviu-kalbos-tekstyno-ir-vektorizuotu-modeliu-sukurimas/
Follow CLARIN-LT news on our Facebook account and website.
Leave a Reply
You must be logged in to post a comment.