@inproceedings{wu-etal-2020-structured,
    title = "Structured Self-{A}ttention{W}eights Encode Semantics in Sentiment Analysis",
    author = "Wu, Zhengxuan  and
      Nguyen, Thanh-Son  and
      Ong, Desmond",
    editor = "Alishahi, Afra  and
      Belinkov, Yonatan  and
      Chrupa{\l}a, Grzegorz  and
      Hupkes, Dieuwke  and
      Pinter, Yuval  and
      Sajjad, Hassan",
    booktitle = "Proceedings of the Third BlackboxNLP Workshop on Analyzing and Interpreting Neural Networks for NLP",
    month = nov,
    year = "2020",
    address = "Online",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/fix-sig-urls/2020.blackboxnlp-1.24/",
    doi = "10.18653/v1/2020.blackboxnlp-1.24",
    pages = "255--264",
    abstract = "Neural attention, especially the self-attention made popular by the Transformer, has become the workhorse of state-of-the-art natural language processing (NLP) models. Very recent work suggests that the self-attention in the Transformer encodes syntactic information; Here, we show that self-attention scores encode semantics by considering sentiment analysis tasks. In contrast to gradient-based feature attribution methods, we propose a simple and effective Layer-wise Attention Tracing (LAT) method to analyze structured attention weights. We apply our method to Transformer models trained on two tasks that have surface dissimilarities, but share common semantics{---}sentiment analysis of movie reviews and time-series valence prediction in life story narratives. Across both tasks, words with high aggregated attention weights were rich in emotional semantics, as quantitatively validated by an emotion lexicon labeled by human annotators. Our results show that structured attention weights encode rich semantics in sentiment analysis, and match human interpretations of semantics."
}