An Information Retrieval Approach to Short Text Conversation
Zongcheng Ji, Zhengdong Lu, Hang Li
TL;DR
This paper treats short text conversation (STC) as an information retrieval problem, leveraging a massive Weibo corpus to retrieve suitable responses rather than generating them. It proposes a three-stage retrieval framework with a suite of matching models, including a translation-based language model, a deep matching network, and a topic-word model, all learned within a learning-to-rank setup (RankingSVM). The authors release a large STC dataset and show that combining multiple features significantly improves retrieval performance (best MAP ≈ 0.654, P@1 ≈ 0.637), while also detailing limitations such as entity association and logic consistency. The work demonstrates that IR approaches can yield human-like responsiveness at scale and provides concrete directions for enhancing STC with semantics, discourse, and entity-aware reasoning.
Abstract
Human computer conversation is regarded as one of the most difficult problems in artificial intelligence. In this paper, we address one of its key sub-problems, referred to as short text conversation, in which given a message from human, the computer returns a reasonable response to the message. We leverage the vast amount of short conversation data available on social media to study the issue. We propose formalizing short text conversation as a search problem at the first step, and employing state-of-the-art information retrieval (IR) techniques to carry out the task. We investigate the significance as well as the limitation of the IR approach. Our experiments demonstrate that the retrieval-based model can make the system behave rather "intelligently", when combined with a huge repository of conversation data from social media.