Using Multi-Encoder Fusion Strategies to Improve Personalized Response Selection
This work improves personalized dialogue systems by enhancing response selection accuracy, though it is incremental as it builds on existing BERT-based models with new fusion strategies.
The paper tackled the problem of personalized response selection by addressing the correlation between persona and empathy and improving faithfulness to conversation context, achieving a new state-of-the-art with improvements of over 2.3% on original personas and 1.9% on revised personas in hits@1 accuracy on the Persona-Chat dataset.
Personalized response selection systems are generally grounded on persona. However, there exists a co-relation between persona and empathy, which is not explored well in these systems. Also, faithfulness to the conversation context plunges when a contradictory or an off-topic response is selected. This paper attempts to address these issues by proposing a suite of fusion strategies that capture the interaction between persona, emotion, and entailment information of the utterances. Ablation studies on the Persona-Chat dataset show that incorporating emotion and entailment improves the accuracy of response selection. We combine our fusion strategies and concept-flow encoding to train a BERT-based model which outperforms the previous methods by margins larger than 2.3 % on original personas and 1.9 % on revised personas in terms of hits@1 (top-1 accuracy), achieving a new state-of-the-art performance on the Persona-Chat dataset.