Date: 2021-11-12 / 3:00 ~ 4:00 PM
Emotion plays an important role in building more human-like social chatbots. A lot of recent research has focused on building an empathetic dialogue system. However, many deep learning models can be less interpretable, and it is sometimes hard to identify causes as well as emotional changes of both speakers and listeners in conversations. The goal of this project is to extract emotional reasoning structures that are useful and interpretable for enhancing empathetic capabilities of social chatbots. Specifically, this project intends to use Reddit data to achieve three goals: (1) extract emotion collocations, (2) extract emotion-cause pairs using information from posts (as situations) and comments (as responses), and (3) generate a framework that contains situation-response pairs across Reddit posts with corresponding emotions.