StalemateBreaker: A Proactive Content-Introducing Approach to Automatic\n Human-Computer Conversation

Xiang Li, Lili Mou, Rui Yan, Ming Zhang

arXiv (Cornell University) · 2016 · 24 citations · 31 references

DOIFull text

Open access

Concepts

Abstract

Existing open-domain human-computer conversation systems are typically\npassive: they either synthesize or retrieve a reply provided a human-issued\nutterance. It is generally presumed that humans should take the role to lead\nthe conversation and introduce new content when a stalemate occurs, and that\nthe computer only needs to "respond." In this paper, we propose\nStalemateBreaker, a conversation system that can proactively introduce new\ncontent when appropriate. We design a pipeline to determine when, what, and how\nto introduce new content during human-computer conversation. We further propose\na novel reranking algorithm Bi-PageRank-HITS to enable rich interaction between\nconversation context and candidate replies. Experiments show that both the\ncontent-introducing approach and the reranking algorithm are effective. Our\nfull StalemateBreaker model outperforms a state-of-the-practice conversation\nsystem by +14.4% p@1 when a stalemate occurs.\n

References

31