Simulated Annealing for Emotional Dialogue Systems
This work addresses the challenge of ensuring emotional accuracy in dialogue systems, which is important for applications like empathetic personal companions, but it is incremental as it builds on existing search-based methods.
The paper tackled the problem of generating emotionally accurate dialogue responses by proposing a simulated annealing-based search method that iteratively edits general responses to improve emotional correctness while maintaining coherence. The result was a 12% improvement in emotion accuracy over the previous state-of-the-art on the NLPCC2017 dataset, with no degradation in generation quality as measured by BLEU.
Explicitly modeling emotions in dialogue generation has important applications, such as building empathetic personal companions. In this study, we consider the task of expressing a specific emotion for dialogue generation. Previous approaches take the emotion as an input signal, which may be ignored during inference. We instead propose a search-based emotional dialogue system by simulated annealing (SA). Specifically, we first define a scoring function that combines contextual coherence and emotional correctness. Then, SA iteratively edits a general response and searches for a sentence with a higher score, enforcing the presence of the desired emotion. We evaluate our system on the NLPCC2017 dataset. Our proposed method shows 12% improvements in emotion accuracy compared with the previous state-of-the-art method, without hurting the generation quality (measured by BLEU).