Modelling Context Emotions using Multi-task Learning for Emotion Controlled Dialog Generation

Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics, pages 2919–2931 April 19 - 23, 2021. ©2021 Association for Computational Linguistics 2919 Modelling Context Emotions using Multi-task Learning for Emotion Controlled Dialog Generation Deeksha Varshney, Asif Ekbal, Pushpak Bhattacharyya Department of Computer Science and Engineering Indian Institute of Technology Patna Patna, India {1821cs13, asif, pb}@iitp.ac.in Abstract A recent topic of research in natural language generation has been the development of auto- matic response generation modules that can automatically respond to a user’s utterance in an empathetic manner. Previous research has tackled this task using neural generative meth- ods by augmenting emotion classes with the in- put sequences. However, the outputs by these models may be inconsistent. We employ multi- task learning to predict the emotion label and to generate a viable response for a given ut- terance using a common encoder with multi- ple decoders. Our proposed encoder-decoder model consists of a self-attention based en- coder and a decoder with dot product attention mechanism to generate response with a speci- fied emotion. We use the focal loss to handle imbalanced data distribution, and utilize the consistency loss to allow coherent decoding by the decoders. Human evaluation reveals that our model produces more emotionally perti- nent responses. In addition, our model outper- forms multiple strong baselines on automatic evaluation measures such as F1 and BLEU scores, thus resulting in more fluent and ade- quate responses. 1 Introduction One of the key skills for dialogue agents in a dialog system is to acknowledge the feelings of the user and respond accordingly. It is quite instinctive for humans to identify and understand other people’s emotions but is quite hard for Artificial Intelligence (AI) systems due to the lack of representative pub- licly available data sets for training and evaluating an intelligent and robust dialog management sys- tem. Table 1 shows an example of emotion labelled conversation from the dataset. The example shows how two different emotionally inclined responses can lead a conversation in two different directions. An engaging conversation usually involves empa- Agent 1 Do you like wearing hats? It has so many functions. Curious Agent 2 I don’t like them on myself but I know a lot of people that can pull them off. Neutral Agent 1 Yes me as well. In the military hats denote a nationality, branch of service, rank or regiment. Curious Agent 2 Yes. I love hats! I have a wide variety of hats and wear them for different reasons. Happy Agent 1 Yes. . . Even I like it too !! Specially I am on vacation, roaming around I do carry 2–3 hats. And I wear it according to my dressing style. Happy Table 1: A snippet of two different emotionally in- clined conversations with a common query. thetic responses by conversing partners which can have varied emotion labels. It is important to capture user’s affective infor- mation by any dialog agent to build an intelligent and socially engaging open-domain chatbot. …

Content truncated to 500 words