The consistency of a response to a given post at semantic-level and emotional-level is essential for a dialogue system to deliver human-like interactions. However, this challenge is not well addressed in the literature, since most of the approaches n
eglect the emotional information conveyed by a post while generating responses. This article addresses this problem by proposing a unifed end-to-end neural architecture, which is capable of simultaneously encoding the semantics and the emotions in a post for generating more intelligent responses with appropriately expressed emotions. Extensive experiments on real-world data demonstrate that the proposed method outperforms the state-of-the-art methods in terms of both content coherence and emotion appropriateness.
The consistency of a response to a given post at semantic-level and emotional-level is essential for a dialogue system to deliver human-like interactions. However, this challenge is not well addressed in the literature, since most of the approaches n
eglect the emotional information conveyed by a post while generating responses. This article addresses this problem by proposing a unifed end-to-end neural architecture, which is capable of simultaneously encoding the semantics and the emotions in a post and leverage target information for generating more intelligent responses with appropriately expressed emotions. Extensive experiments on real-world data demonstrate that the proposed method outperforms the state-of-the-art methods in terms of both content coherence and emotion appropriateness.
Scalability for handling unknown slot values is a important problem in dialogue state tracking (DST). As far as we know, previous scalable DST approaches generally rely on either the candidate generation from slot tagging output or the span extractio
n in dialogue context. However, the candidate generation based DST often suffers from error propagation due to its pipelined two-stage process; meanwhile span extraction based DST has the risk of generating invalid spans in the lack of semantic constraints between start and end position pointers. To tackle the above drawbacks, in this paper, we propose a novel scalable dialogue state tracking method based on slot tagging navigation, which implements an end-to-end single-step pointer to locate and extract slot value quickly and accurately by the joint learning of slot tagging and slot value position prediction in the dialogue context, especially for unknown slot values. Extensive experiments over several benchmark datasets show that the proposed model performs better than state-of-the-art baselines greatly.
