3つの要点 ✔️ 文章生成における新たな評価基準BERTScore ✔️ BERTの埋め込み(分散表現)を利用することで文章の類似性を評価 ✔️ 既存手法に比べ人間の判断と高い相関を示す評価基準 BERTScore:Evaluating Text Generation with BERT written by Tianyi Zhang, Varsha Kishore, Felix Wu, Kilian Q. Weinberger, Yoav Artzi (Submitted on 21 Apr 2019 (v1), last revised 24 Feb 2020 (this version, v3)) Comments: Published by ICLR2020 Subjects: Computation and Language (cs.CL) 機械翻訳や文章要約のような文章生成のタ
![従来のBLEUscoreでは正しく評価できない! 自然言語に最適な人間に近い評価基準BERTScore登場!](https://cdn-ak-scissors.b.st-hatena.com/image/square/d1f5177249fef011af4395cd5af420d7b977ba33/height=288;version=1;width=512/https%3A%2F%2Faisholar.s3.ap-northeast-1.amazonaws.com%2Fmedia%2FApril2020%2FBERT%25E3%2582%2592%25E8%25A9%2595%25E4%25BE%25A1%25E3%2581%25AB_%25E7%2594%25A8%25E3%2581%2584%25E3%2582%258B%25E3%2581%25A8%25E4%25BA%25BA%25E9%2596%2593%25E3%2581%25AB%25E8%25BF%2591%25E3%2581%2584%25EF%25BC%259F.png)