Incorporating Interlocutor-Aware Context into Response Generation on Multi-Party Chatbots

Proceedings of the 23rd Conference on Computational Natural Language Learning, pages 718–727 Hong Kong, China, November 3-4, 2019. c⃝2019 Association for Computational Linguistics 718 Incorporating Interlocutor-Aware Context into Response Generation on Multi-Party Chatbots Cao Liu1,2, Kang Liu1,2, Shizhu He1,2, Zaiqing Nie3, Jun Zhao1,2 1 National Laboratory of Pattern Recognition, Institute of Automation, Chinese Academy of Sciences, Beijing, 100190, China 2 University of Chinese Academy of Sciences, Beijing, 100049, China 3 Alibaba AI Labs, Beijing, 100029, China {cao.liu, kliu, shizhu.he, jzhao}@nlpr.ia.ac.cn zaiqing.nzq@alibaba-inc.com Abstract Conventional chatbots focus on two-party re- sponse generation, which simplifies the re- al dialogue scene. In this paper, we strive toward a novel task of Response Generation on Multi-Party Chatbot (RGMPC), where the generated responses heavily rely on the inter- locutors’ roles (e.g., speaker and addressee) and their utterances. Unfortunately, complex interactions among the interlocutors’ roles make it challenging to precisely capture con- versational contexts and interlocutors’ infor- mation. Facing this challenge, we present a response generation model which incorpo- rates Interlocutor-aware Contexts into Recur- rent Encoder-Decoder frameworks (ICRED) for RGMPC. Specifically, we employ interac- tive representations to capture dialogue con- texts for different interlocutors. Moreover, we leverage an addressee memory to enhance contextual interlocutor information for the tar- get addressee. Finally, we construct a corpus for RGMPC based on an existing open-access dataset. Automatic and manual evaluations demonstrate that the ICRED remarkably out- performs strong baselines. 1 Introduction Human computer conversation has been an impor- tant and challenging task in NLP and AI since the Turing Test was proposed in 1950 (Turing, 1950). Recently, with the rapid growth of social conver- sation data available on the Internet, data-driven chatbots are able to learn to generate responses di- rectly and have attracted much more attention than before (Li et al., 2016a; Tian et al., 2017). Researches in this area mostly focus on the dia- log with two interlocutors (Ma´ıra Gatti de Bayser et al., 2017). However, the real-life interaction in- volves a substantial part of Multi-Party Chatbot- s (MPC, such as internet forum and chat group), which is a form of conversation with multiple in- t Speaker Addressee Utterance 1 Alan ( ) Bert ( ) the main ubuntu 2 Carl ( ) - i n-1 Carl ( ) Jack ( ) i n Carl ( ) n+1 Jack ( ) Carl ( ) mate is a recent reincarnation i (Generated) Response Responding Speaker Target Addressee Context Figure 1: An example of Multi-Party Chatbots (MPC). At each turn, a speaker said one utterance to an ad- dressee. There are many interlocutors (e.g., Alan, Bert and so on) in a conversation, where ai represents inter- locutor’s ID. terlocutors (Ouchi and Tsuboi, 2016). For ex- ample, there are more than three interlocutors (a1, a2, a3...am) involved in the conversation in Figure 1, and their roles (e.g., speaker and ad- dressee) may change across different dialog turns. As shown in Figure 1, at each turn, the core is- sue of MPC is to capture who (speaker) talks to whom (addressee) about …

Content truncated to 500 words