ymym3412 / acl-papers

paper summary of Association for Computational Linguistics
184 stars 10 forks source link

Exploiting Document Knowledge for Aspect-level Sentiment Classification #289

Open ymym3412 opened 6 years ago

ymym3412 commented 6 years ago

0. 論文

Exploiting Document Knowledge for Aspect-level Sentiment Classification Ruidan He, Wee Sun Lee, Hwee Tou Ng, Daniel Dahlmeier

1. どんなもの?

Aspect-level Sentiment Classificationの学習のためにDocument-levelの感情分類のデータであらかじめ学習を行う

2. 先行研究と比べてどこがすごい?

Document-Levelデータで事前学習することでデータの少なさを補う

3. 技術や手法のキモはどこ?

Aspect-level Sentiment Classificationはアノテーション付きデータが少ないため、Document-levelのデータを使うことでデータ不足を補う。 事前に埋め込み層やLSTM層を学習させたり、Aspet-levelとDocument-levelのマルチタスク学習などを行う。

4. どうやって有効だと検証した?

4種類データセットを用意し、AccuracyとMacro-F1で比較。 事前学習とマルチタスク学習を組み合わせたものが最も良い数値を出した

5. 議論はある?

埋め込み層とLSTM層の転移が有効。 また提案手法はLow-resourceなタスクで有向

6. 次に読むべき論文は?

How Transferable are Neural Networks in NLP Applications? Lili Mou, Zhao Meng, Rui Yan, Ge Li, Yan Xu, Lu Zhang, Zhi Jin

ymym3412 commented 6 years ago

以前まとめたもの How Transferable are Neural Networks in NLP Applications? https://github.com/arXivTimes/arXivTimes/issues/484