2023-SOFT-Kyushu-SSLsenti

This repository is a model for paper.
If you use STEP.2 and STEP.3 models like a paper, change branch of name.

@ARTICLE{Kondo2023-va,
  title    = "Evaluation of the {BERT} model which is fine-tuned by the
              datasets including {Pseudo-Label} which is generated by
              {Self-Training} for sentimental classification",
  author   = "Kondo, Aoi and Oda, Mikio",
  abstract = "Recently, NLP tasks need many datasets because of the use of
              transformer architecture. However, the amount of datasets is
              limited since it is created by humans. So we strive to improve it
              by using the Pseudo-Labeling approach in Semi-Supervised
              learning. This paper creates the model of sentimental
              classification by Fine-Tuning using the Pseudo-Label which is
              generated by Self-Training, and evaluates it.",
  journal  = "第25回日本知能情報ファジィ学会九州支部学術講演会",
  month    =  dec,
  year     =  2023
}