End-to-end sequence generation is a popular technique for developing open domain dialogue systems, though they suffer from the safe response problem. Researchers have attempted to tackle this problem by incorporating generative models with the returns of retrieval systems. Recently, a skeleton-thenresponse framework has been shown promising results for this task. Nevertheless, how to precisely extract a skeleton and how to effectively train a retrieval-guided response generator are still

more »

... ging. This paper presents a novel framework in which the skeleton extraction is made by an interpretable matching model and the following skeleton-guided response generation is accomplished by a separately trained generator. Extensive experiments demonstrate the effectiveness of our model designs. 1 Previous studies (Weston et al., 2018; Wu et al., 2019; Cai et al., 2019) alleviated the problems by putting hard constraints on the data, which, however, greatly reduces the amount of usable data (e.g., Cai et al. (2019) required the Jaccard distance between the retrieved response and the target response should be in the range [0.3, 0.7]).