Skip to content

Using Polarity Lexicons and BERT Model for Supervised Learning With Russian Initially Unlabeled Datasets

Notifications You must be signed in to change notification settings

antongolubev5/Auto-Dataset-For-Transfer-Learning

Repository files navigation

This repository provides a source code and data for the paper "Transfer Learning for Improving Results on Russian Sentiment Datasets" (Anton Golubev and Natalia Loukachevitch, Dialogue 2021).

This repo provides only scripts for creating dataset of thematic contexts from raw corpus of news. The source code for the BERT models described in the article, you can find in https://github.com/antongolubev5/Targeted-SA-for-Russian-Datasets.

Requirements

  • python: 3.7.1
  • numpy: 1.15.4
  • nltk
  • sklearn
  • pandas
  • os
  • tqdm
  • xml
  • time
  • spacy
  • gensim
  • pymorphy2
  • seaborn

References

@misc{golubev2021improvingtf,
    title={Transfer Learning for Improving Results on Russian Sentiment Datasets},
    author={Anton Golubev and Natalia Loukachevitch},
    year={2021},
    eprint={2007.14310},
    archivePrefix={arXiv},
    primaryClass={cs.CL}
}

About

Using Polarity Lexicons and BERT Model for Supervised Learning With Russian Initially Unlabeled Datasets

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages