TripleNet: Triple Attention Network for Multi-Turn Response Selection in Retrieval-Based Chatbots

Proceedings of the 23rd Conference on Computational Natural Language Learning, pages 737–746 Hong Kong, China, November 3-4, 2019. c⃝2019 Association for Computational Linguistics 737 TripleNet: Triple Attention Network for Multi-Turn Response Selection in Retrieval-based Chatbots Wentao Ma†, Yiming Cui‡†, Nan Shao†, Su He†, Wei-Nan Zhang‡, Ting Liu‡, Shijin Wang†§, Guoping Hu† †State Key Laboratory of Cognitive Intelligence, iFLYTEK Research, China ‡Research Center for Social Computing and Information Retrieval (SCIR), Harbin Institute of Technology, Harbin, China §iFLYTEK AI Research (Hebei), Langfang, China †§{wtma,ymcui,nanshao,suhe,sjwang3,gphu}@iflytek.com ‡{ymcui,wnzhang,tliu}@ir.hit.edu.cn Abstract We consider the importance of different ut- terances in the context for selecting the response usually depends on the current query.1 In this paper, we propose the model TripleNet to fully model the task with the triple ⟨context, query, response⟩ instead of ⟨context, response⟩ in previous works. The heart of TripeNet is a novel attention mecha- nism named triple attention to model the rela- tionships within the triple at four levels. The new mechanism updates the representation for each element based on the attention with the other two concurrently and symmetrically. We match the triple ⟨C, Q, R⟩ centered on the re- sponse from char to context level for predic- tion. Experimental results on two large-scale multi-turn response selection datasets show that the proposed model can significantly out- perform the state-of-the-art methods. 2 1 Introduction To establish a human-machine dialogue system is one of the most challenging tasks in Artifi- cial Intelligence (AI). Existing works on build- ing dialogue systems are mainly divided into two categories: retrieval-based method (Yan et al., 2016; Zhou et al., 2016), and generation-based method (Vinyals and Le, 2015). The retrieval- based method retrieves multiple candidate re- sponses from the massive repository and selects the best one as the system’s response, while the generation-based method uses the encoder- decoder framework to generate the response, which is similar to machine translation. 1In this paper, we define the last message which is waiting for a response as the ‘query,’ the conversation history includ- ing the query as ‘context,’ and each message in the context as an ‘utterance.’ 2 TripleNet source code is available at https:// github.com/wtma/TripleNet. A: i downloaded angry ip scanner and now it doesn’t work and i can’t uninstall it B: you installed it via package or via some ::::: binary :::::: installer A: i installed from ubuntu soft center B: hm i do n’t know what package it is but it should let you remove it the same way A: ah makes sense then ... hm ::: was :it:a:::: deb :: file True Response: i think :it:::: was :::::: another ::::: format mayge sth starting with r False Response: thanks i appreciate it try sudo apt-get install libxine-extracodecs Figure 1: A real example in the Ubuntu Corpus. The upper part is the conversation between speaker A and B. The speaker A want to uninstall the ip scanner and the current query is about the format of the package, so the true response is about the format, but the existing conversation model …

Content truncated to 500 words