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

Existing open-domain human-computer conversation systems are typically: they either synthesize or retrieve a reply provided a human-issued. It is generally presumed that humans should take the role to lead conversation and introduce new content when a stalemate occurs, and that computer only needs to "respond. " In this paper, we propose, a conversation system that can proactively introduce new when appropriate. We design a pipeline to determine when, what, and how introduce new content during human-computer conversation. We further propose novel reranking algorithm Bi-PageRank-HITS to enable rich interaction between context and candidate replies. Experiments show that both the-introducing approach and the reranking algorithm are effective. Our StalemateBreaker model outperforms a state-of-the-practice conversation by +14. 4% p@1 when a stalemate occurs.