Discern: Discourse-Aware Entailment Reasoning Network for Conversational Machine Reading

Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing, pages 2439–2449, November 16–20, 2020. c⃝2020 Association for Computational Linguistics 2439 DISCERN: Discourse-Aware Entailment Reasoning Network for Conversational Machine Reading Yifan Gao†, Chien-Sheng Wu‡, Jingjing Li†, Shafiq Joty‡,§, Steven C.H. Hoi‡, Caiming Xiong‡, Irwin King†, and Michael R. Lyu† † The Chinese University of Hong Kong ‡ Salesforce Research § Nanyang Technological University †{yfgao,lijj,king,lyu}@cse.cuhk.edu.hk ‡{cswu,sjoty,shoi,cxiong}@salesforce.com Abstract Document interpretation and dialog under- standing are the two major challenges for con- versational machine reading. In this work, we propose DISCERN, a discourse-aware en- tailment reasoning network to strengthen the connection and enhance the understanding for both document and dialog. Specifically, we split the document into clause-like elemen- tary discourse units (EDU) using a pre-trained discourse segmentation model, and we train our model in a weakly-supervised manner to predict whether each EDU is entailed by the user feedback in a conversation. Based on the learned EDU and entailment representa- tions, we either reply to the user our final de- cision “yes/no/irrelevant” of the initial ques- tion, or generate a follow-up question to in- quiry more information. Our experiments on the ShARC benchmark (blind, held-out test set) show that DISCERN achieves state-of- the-art results of 78.3% macro-averaged accu- racy on decision making and 64.0 BLEU1 on follow-up question generation. Code and mod- els are released at https://github.com/ Yifan-Gao/Discern. 1 Introduction Conversational Machine Reading (CMR) is chal- lenging because the rule text may not contain the literal answer, but provide a procedure to derive it through interactions (Saeidi et al., 2018). In this case, the machine needs to read the rule text, interpret the user scenario, clarify the unknown user’s background by asking questions, and derive the final answer. Taking Figure 1 as an example, to answer the user whether he is suitable for the loan program, the machine needs to interpret the rule text to know what are the requirements, under- stand he meets “American small business” from the user scenario, ask follow-up clarification questions about “for-profit business” and “not get financing Rule Text: 7(a) loans are the most basic and most used type loan of the Small Business Administration’s (SBA) busi- ness loan programs. It’s name comes from section 7(a) of the Small Business Act, which authorizes the agency to pro- vide business loans to American small businesses. The loan program is designed to assist for-profit businesses that are not able to get other financing from other re- sources. User Scenario: I am a 34 year old man from the United States who owns their own business. We are an American small business. User Question: Is the 7(a) loan program for me? Follow-up Q1: Are you a for-profit business? Follow-up A1: Yes. Follow-up Q2: Are you able to get financing from other resources? Follow-up A2: No. Final Answer: Yes. (You can apply the loan.) Figure 1: An example dialog from the ShARC (Saeidi et al., 2018) dataset. The machine answers the user question by reading the rule text, interpreting the user …

Content truncated to 500 words