Do It Once: An Embarrassingly Simple Joint Matching Approach to Response Selection

Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021, pages 4872–4877 August 1–6, 2021. ©2021 Association for Computational Linguistics 4872 Do It Once: An Embarrassingly Simple Joint Matching Approach to Response Selection Linhao Zhang1, Dehong Ma2, Sujian Li1, Houfeng Wang1 1MOE Key Lab of Computational Linguistics, Peking University, Beijing, 100871, China 2Baidu Inc., China {zhanglinhao, lisujian, wanghf}@pku.edu.cn madehong@baidu.com Abstract Existing matching models for response selec- tion adopt the independent matching (IM) ap- proach. To complete a prediction, they have to perform N independent matches, where N is the number of response options. In this paper, we explore a joint matching (JM) approach which performs matching only once regardless of the number of options. The JM approach does not change the structure of matching component but only modifies its input and output format. It also enables a cheap but effective data augmen- tation method. Extensive experiments on the MuTual dataset demonstrate that, even with the simplest formulation, JM outperforms IM ap- proach by a large margin and reduces training time by over half. 1 Introduction The availability of large-scale datasets has driven the development of neural dialogue systems. One important task in dialogue systems is response se- lection, which plays an essential role in retrieval- based chatbots (Ji et al., 2014). It aims to select the best-matched response from a set of response options for a dialogue. As shown in Figure 1, given a dialogue context and four response options, we need to choose the only logically correct one. Previous work in response selection follows an independent matching (IM) approach and computes a matching score for each of the N response options independently. Various matching models follow- ing this approach have been proposed (Zhou et al., 2016; Wu et al., 2017; Zhou et al., 2018; Chaudhuri et al., 2018; Tao et al., 2019; Yuan et al., 2019). De- spite its success in pre-BERT era, we argue that the IM approach does not make full use of the ability of pretrained encoders (such as BERT and RoBERTa) to encode multiple sentences, hence may hinder both efficiency and effectiveness. Specifically, to Options: ✘A: Sorry. I won't smoke in the hospital again. ✓B: OK. I won't smoke. Could you please give me a menu? ✘C: Could you please tell the customer over there not to smoke? We can't stand the smell. ✘D: Sorry. I will smoke when I get off the bus. M: Excuse me, sir. This is a non smoking area F: Oh, sorry. I will move to the smoking area M: I’m afraid no table in the smoking area is available now Dialogue: Figure 1: Example of response selection. complete a prediction, the IM approach has to per- form N independent matches, which means N gra- dient computations (where N is the number of re- sponse options). Besides, the dialogue context is repeatedly encoded N times, which further con- tributes to the inefficiency. The other drawback is that options in these models are independent and agnostic of each other. In reality, humans …

Content truncated to 500 words