Ranking Responses Oriented to Conversational Relevance in Chat-bots

Proceedings of COLING 2016, the 26th International Conference on Computational Linguistics: Technical Papers, pages 652–662, Osaka, Japan, December 11-17 2016. Ranking Responses Oriented to Conversational Relevance in Chat-bots Bowen Wu1, Baoxun Wang1, Hui Xue2 1Microsoft AI and Research Group, Beijing, China 2Microsoft Research, Beijing, China {bowenwu, baoxwang, xuehui}@microsoft.com Abstract For automatic chatting systems, it is indeed a great challenge to reply the given query considering the conversation history, rather than based on the query only. This paper proposes a deep neural network to address the context-aware response ranking problem by end-to-end learning, so as to help to select conversationally relevant candidate. By combining the multi-column convolutional layer and the recurrent layer, our model is able to model the semantics of the utterance sequence by grasping the semantic clue within the conversation, on the basis of the effective representation for each sentence. Especially, the network utilizes attention pooling to further emphasis the importance of essential words in conversations, thus the representations of contexts tend to be more meaningful and the performance of candidate ranking is notably improved. Meanwhile, due to the adoption of attention pooling, it is possible to visualize the semantic clues. The experimental results on the large amount of conversation data from social media have shown that our approach is promising for quantifying the conversational relevance of responses, and indicated its good potential for building practical IR based chat-bots. 1 Introduction There exist two query intentions in Intelligent Agents: the task completion oriented intention and the open-domain chat intention. As the applications of dialog systems, task completion oriented agents are designed to accomplish users’ requirements in a few rounds of conversations. This kind of intentions reflect users’ basic needs on the agents, thus studies on dialog systems have a longer history and achieved great process (Weizenbaum, 1966; Ferguson et al., 1996; Shawar and Atwell, 2007; Williams, 2010). The open-domain chat intention, by contrast, represents users’ communicating needs. Apparently, automatic chatting systems with good using experience will significantly attract people’s interests, even make people form the habits of communicating with agents, hence it is possible to be a new platform for any task-oriented services to plug in (see Duer1). One challenge directly brought by open-domain Chat-bots is, indeed, user queries can be related to any topic in any possible forms, that is, it’s NOT wise to transform chatting queries into slot-value sequences to further trace users’ intentions within the conversation, as task-oriented dialog systems do. The even more essential challenge chat-bots have to face is to guarantee the semantic and logic con- tinuity of conversations, that is, a response from bots should be relevant with both the adjacent query and the corresponding short conversation history. Actually, such “context-aware” chatting ability is the critical feature of a human-like chat-bot, thus much attention has been paid on this task. The basic re- quirement for chat-bots is to semantically understand conversations like humans, which is abstracted as the conversation modeling problem. This paper discusses the approaches to addressing the context-aware chatting problem, by …

Content truncated to 500 words