ファインチューニングを用いた畳み込みニューラルネットワークによる皮肉検出

書誌事項

タイトル別名
  • SARCASM DETECTION USING CONVOLUTIONAL NEURAL NETWORK WITH FINE-TUNING

この論文をさがす

説明

Sarcasm detection is an important task in correctly understanding a sentence. This study proposes a method of sarcasm detection using a Convolutional Neural Network (CNN) with fine-tuning. Assuming sarcasm is caused by sentiment, we attempt to pre-train the CNN on a sentiment analysis task to determine whether a sentence is negative or positive. We fine-tune the pre-trained CNN on the sarcasm detection task to determine whether a sentence is sarcastic or non-sarcastic. We found that the proposed method could detect sarcasm more accurately than CNN, Recurrent Neural Network (RNN), and Long Short-Term Memory (LSTM) with randomly initialized parameters.

収録刊行物

詳細情報 詳細情報について

  • CRID
    1390578050264143872
  • DOI
    10.15002/00026390
  • HANDLE
    10114/00026390
  • ISSN
    24368083
  • 本文言語コード
    ja
  • 資料種別
    departmental bulletin paper
  • データソース種別
    • JaLC
    • IRDB
  • 抄録ライセンスフラグ
    使用可

問題の指摘

ページトップへ