A Practical Dialogue-Act-Driven Conversation Model for Multi-Turn Response Selection

Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing, pages 1980–1989, Hong Kong, China, November 3–7, 2019. c⃝2019 Association for Computational Linguistics 1980 A Practical Dialogue-Act-Driven Conversation Model for Multi-Turn Response Selection Harshit Kumar IBM Research, Delhi, India harshitk@in.ibm.com Arvind Agarwal IBM Research, Delhi, India arvagarw@in.ibm.com Sachindra Joshi IBM Research, Delhi, India jsachind@in.ibm.com Abstract Dialogue Acts play an important role in con- versation modeling. Research has shown the utility of dialogue acts for the response se- lection task, however, the underlying assump- tion is that the dialogue acts are readily avail- able, which is impractical, as dialogue acts are rarely available for new conversations. This paper proposes an end-to-end multi-task model for conversation modeling, which is op- timized for two tasks, dialogue act prediction and response selection, with the latter being the task of interest. It proposes a novel way of combining the predicted dialogue acts of con- text and response with the context (previous utterances) and response (follow-up utterance) in a crossway fashion, such that, it achieves at par performance for the response selection task compared to the model that uses actual dialogue acts. Through experiments on two well known datasets, we demonstrate that the multi-task model not only improves the accu- racy of the dialogue act prediction task but also improves the MRR for the response selection task. Also, the cross-stitching of dialogue acts of context and response with the context and response is better than using either one of them individually. 1 Introduction Response selection remains at the core of con- versation modeling, with the objective of select- ing an appropriate response utterance from a set of candidate utterances, for a given conversation history consisting of previous utterances (context). Decades of research in this task includes tradi- tional methods such as (Kitano, 1991; Ritter et al., 2011) and recent deep learning based methods (Ji et al., 2014; Chaudhuri et al., 2018; Xu et al., 2018; Chen et al., 2017; Song et al., 2018; Lowe et al., 2015; Zhao et al., 2018; Wen et al., 2016). Underlying these methods, a fundamental need is to capture the semantics of the context and use it for selecting the appropriate response. While the context provides essential clues as to what could be a follow-up response, research (Kumar et al., 2018) has further shown that any additional infor- mation available in the form of dialogue acts can also be helpful for response selection. Such in- formation when used along with the context im- proves the performance of the response selection task. However, the above method assumes that di- alogue acts are available at the time of response selection, which is rarely the case —as dialogue acts are usually not available for new conversa- tions in a live setting— thus making them imprac- tical for practitioners. In this paper, we propose a novel model that bridges this gap between theory and practice. In other words, our proposed model leverages the …

Content truncated to 500 words