Building Interpretable Interaction Trees for Deep NLP Models

Die Zhang, Hao Zhang, Huilin Zhou, Xiaoyi Bao, Da Huo, Ruizhao Chen, Xu Cheng, Mengyue Wu, Quanshi Zhang

AAAI 2021 pp. 14328-14337

doi:10.1609/AAAI.V35I16.17685 /aaai/2021/zhang2021aaai-building/

Abstract

This paper proposes a method to disentangle and quantify interactions among words that are encoded inside a DNN for natural language processing. We construct a tree to encode salient interactions extracted by the DNN. Six metrics are proposed to analyze properties of interactions between constituents in a sentence. The interaction is defined based on Shapley values of words, which are considered as an unbiased estimation of word contributions to the network prediction. Our method is used to quantify word interactions encoded inside the BERT, ELMo, LSTM, CNN, and Transformer networks. Experimental results have provided a new perspective to understand these DNNs, and have demonstrated the effectiveness of our method.

PDF AAAI Semantic Scholar

Cite

Text

Zhang et al. "Building Interpretable Interaction Trees for Deep NLP Models." AAAI Conference on Artificial Intelligence, 2021. doi:10.1609/AAAI.V35I16.17685

Markdown

[Zhang et al. "Building Interpretable Interaction Trees for Deep NLP Models." AAAI Conference on Artificial Intelligence, 2021.](https://mlanthology.org/aaai/2021/zhang2021aaai-building/) doi:10.1609/AAAI.V35I16.17685

BibTeX

@inproceedings{zhang2021aaai-building,
  title     = {{Building Interpretable Interaction Trees for Deep NLP Models}},
  author    = {Zhang, Die and Zhang, Hao and Zhou, Huilin and Bao, Xiaoyi and Huo, Da and Chen, Ruizhao and Cheng, Xu and Wu, Mengyue and Zhang, Quanshi},
  booktitle = {AAAI Conference on Artificial Intelligence},
  year      = {2021},
  pages     = {14328-14337},
  doi       = {10.1609/AAAI.V35I16.17685},
  url       = {https://mlanthology.org/aaai/2021/zhang2021aaai-building/}
}