Towards Explainable NLP: A Generative Explanation Framework for Text Classification

Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 5570–5581 Florence, Italy, July 28 - August 2, 2019. c⃝2019 Association for Computational Linguistics 5570 Towards Explainable NLP: A Generative Explanation Framework for Text Classification Hui Liu1, Qingyu Yin2, William Yang Wang3 1 Peking University, China 2 Harbin Institute of Technology, China 3 University of California, Santa Barbara, USA layneliuhui@gmail.com qyyin@ir.hit.edu.cn william@cs.ucsb.edu Abstract Building explainable systems is a critical prob- lem in the field of Natural Language Process- ing (NLP), since most machine learning mod- els provide no explanations for the predictions. Existing approaches for explainable machine learning systems tend to focus on interpret- ing the outputs or the connections between in- puts and outputs. However, the fine-grained information (e.g. textual explanations for the labels) is often ignored, and the systems do not explicitly generate the human-readable ex- planations. To solve this problem, we pro- pose a novel generative explanation frame- work that learns to make classification deci- sions and generate fine-grained explanations at the same time. More specifically, we intro- duce the explainable factor and the minimum risk training approach that learn to generate more reasonable explanations. We construct two new datasets that contain summaries, rat- ing scores, and fine-grained reasons. We con- duct experiments on both datasets, compar- ing with several strong neural network base- line systems. Experimental results show that our method surpasses all baselines on both datasets, and is able to generate concise expla- nations at the same time. 1 Introduction Deep learning methods have produced state-of- the-art results in many natural language process- ing (NLP) tasks (Vaswani et al., 2017; Yin et al., 2018; Peters et al., 2018; Wang et al., 2018; Han- cock et al., 2018; Ma et al., 2018). Though these deep neural network models achieve impressive performance, it is relatively difficult to convince people to trust the predictions of such neural net- works since they are actually black boxes for hu- man beings (Samek et al., 2018). For instance, if an essay scoring system only tells the scores of a given essay without providing explicit rea- sons, the users can hardly be convinced of the judgment. Therefore, the ability to explain the rationale is essential for a NLP system, a need which requires traditional NLP models to provide human-readable explanations. In recent years, lots of works have been done to solve text classification problems, but just a few of them have explored the explainability of their systems (Camburu et al., 2018; Ouyang et al., 2018). Ribeiro et al. (2016) try to identify an inter- pretable model over the interpretable representa- tion that is locally faithful to the classifier. Samek et al. (2018) use heatmap to visualize how much each hidden element contributes to the predicted results. Although these systems are somewhat promising, they typically do not consider fine- grained information that may contain information for interpreting the behavior of models. However, if a human being wants to rate a product, s/he may first write down some reviews, and then …

Content truncated to 500 words