Enhancing Multiple-choice Machine Reading Comprehension by Punishing Illogical Interpretations

Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, pages 3641–3652 November 7–11, 2021. c⃝2021 Association for Computational Linguistics 3641 Enhancing Multiple-choice Machine Reading Comprehension by Punishing Illogical Interpretations Yiming Ju1,2, Yuanzhe Zhang1,2, Zhixing Tian1,2, Kang Liu1,2, Xiaohuan Cao3, Wenting Zhao3, Jinlong li3, Jun Zhao1,2 1 National Laboratory of Pattern Recognition, Institute of Automation, CAS, Beijing, China 2 School of Artificial Intelligence, University of Chinese Academy of Sciences, Beijing, China 3 AI Lab, China Merchant Bank, ShenZhen, 518057, China {yiming.ju, yzzhang, zhixing.tian, kliu}@nlpr.ia.ac.cn {xhcao, wtzhao, lucida}@cmbchina.com, jzhao@nlpr.ia.ac.cn Abstract Machine Reading Comprehension (MRC), which requires a machine to answer questions given the relevant documents, is an impor- tant way to test machines’ ability to under- stand human language. Multiple-choice MRC is one of the most studied tasks in MRC due to the convenience of evaluation and the flex- ibility of answer format. Post-hoc interpreta- tion aims to explain a trained model and reveal how the model arrives at the prediction. One of the most important interpretation forms is to attribute model decisions to input features. Based on post-hoc interpretation methods, we assess attributions of paragraphs in multiple- choice MRC and improve the model by punish- ing the illogical attributions. Our method can improve model performance without any ex- ternal information and model structure change. Furthermore, we also analyze how and why such a self-training method works. 1 Introduction Machine reading comprehension (MRC), which re- quires a machine to answer questions according to given documents, is an important way to test the ability of intelligence systems to understand human language (Hermann et al., 2015; Chen, 2018). As with other tasks in Natural Language Processing (NLP), deep models have achieved great success on MRC. At the same time, deep models’ opaqueness grows in tandem with their power (Doshi-Velez and Kim, 2017), which has motivated efforts to inter- pret how these black-box models work. Post-hoc interpretation aims to explain a trained model and reveal how the model arrives at the prediction (Ja- covi and Goldberg, 2020; Molnar, 2020), as shown in Figure 1. This goal is usually approached with attribution method, which assesses attributions of inputs to model predictions (Bach et al., 2015; Sun- dararajan et al., 2017; Shrikumar et al., 2017). In NLP, interpretations are usually given by assess- ing attributions of words, phrases, sentences and paragraphs (Ribeiro et al., 2016; Lundberg and Lee, 2017; Plumb et al., 2018; Chen et al., 2020; De Cao et al., 2020; Jacovi and Goldberg, 2020), in which positive attributions mean support to the prediction and negative ones mean opposition. Figure 1: Post-hoc interpretation aims to explain a trained model and reveal how the model arrives at the prediction. It is well known that the strong fit ability of deep models can cause incredibly high performance on the training set. The correct prediction of an MRC model on the training set can’t reflect the model has understood the sample and used a suitable way to predict. Since post-hoc interpretations can provide insights into how …

Content truncated to 500 words