Embedding Model Fine-Tuning Dataset

  Autori

Deksne, Daiga

  Vienuma identifikators

http://hdl.handle.net/20.500.12574/136

 Projekta URL

https://www.lzp.gov.lv/lv/projekts/maksliga-intelekta-metozu-piemerotibas-analize-eiropas-savienibas-fondu-projektu-joma

 Izdošanas datums

2025-08-27

 Tips

corpus, text

 Lielums

432895 items

 Valoda(-s)

Latvian

 Apraksts

Dataset for Embedding Model Fine-Tuning has been created within the framework of the National Research Program project "Analysis of the applicability of artificial intelligence methods in the field of EU fund projects". For the purposes of this project, we fine-tuned the bge-m3 model developed by BAAI (Chen et al., 2024). For fine-tuning, we collected DOCX procurement documents from the Electronic Procurement System (https://www.eis.gov.lv/EKEIS/Supplier/), allocating 7,083 files to the training set and 50 files to the validation set. The text from these documents was extracted and segmented. For each text segment, we used the OpenAI gpt-4o model to generate statements or questions whose correctness can be verified against that specific segment. License and Attribution The dataset is distributed under the CC-BY-NC-SA license: https://creativecommons.org/licenses/by-nc-sa/4.0/. When using this dataset, please cite as: Project "Analysis of the Applicability of Artificial Intelligence Methods in the Field of European Union Fund Projects" (VPP-CFLA-Mākslīgais intelekts-2024/1-0003). Dataset for Embedding Model Fine-Tuning. Licensed under CC BY-NC-SA 4.0.

 Izdevējs

University of Latvia

 Atbalsts

Latvian Council of Science

Projekta kods: VPP-CFLA-Mākslīgais intelekts-2024/1-0003

Projekta nosaukums: Mākslīgā intelekta metožu piemērotības analīze Eiropas Savienības fondu projektu jomā

 Tēma(-s)

Embedding Model Fine-Tuning

 Kolekcija (s)

Language resources of the Faculty of Science and Technology, University of Latvia

Rādīt pilnu ierakstu