Using Transfer Learning Methods to Enable Reduction of Supervised Data for Conversation Analysis in Collaborative Learning

DOI Web Site

Bibliographic Information

Other Title
  • 協調学習における会話分析用教師データの削減を可能とする転移学習の活用
  • キョウチョウ ガクシュウ ニ オケル カイワ ブンセキヨウ キョウシ データ ノ サクゲン オ カノウ ト スル テンイ ガクシュウ ノ カツヨウ

Search this article

Abstract

<p>　In recent years，interest in learning analytics that analyzes big educational data has increased.，and research using deep learning technology has also been conducted. Our research team has also constructed AI models that automatically and in real time add coding label to collaborative learning data，verified its accuracy and its usefulness. However，it has been a challenge to create supervised data from big data，requiring enormous human power and time. In addition，it will cost more if a new supervised data has to be constructed for each research subject and needs for practical use. In this study，we verified how much human labeling work can be reduced by conducting pre-training as transfer learning before the supervised learning.</p>

Journal

Related Projects

See more

Keywords

Details 詳細情報について

Report a problem

Back to top