Supervised Term Weight Training for Improving Question-Knowledge Matching in Chatbots

Shuangyong Song, Chao Wang, Haiqing Chen, Huan Chen

Abstract: We design several supervised methods of training term weights for improving results of question-knowledge matching task on a chatbot system-AliMe. Considering demand of high QPS (Query Per Second), we take a traditional regression model as the matching model rather than deep models.