# Emily: Developing An Emotion-affective Open-Domain Chatbot with Knowledge Graph-based Persona

**Authors:** Weixuan Wang, Xiaoling Cai, Chong Hsuan Huang, Haoran Wang, Haonan Lu, Ximing Liu, Wei Peng  
**arXiv:** [2109.08875](https://arxiv.org/abs/2109.08875)  
**Date:** 2021-09-18  
**Summary page:** https://www.summarizepaper.com/en/paper/2109.08875v1/

## Key points

- Authors present their work on developing Emily, an emotion-affective open-domain chatbot
- Goal of Emily is to perceive a user's negative emotion state and provide support by positively converting their emotions
- Proposed two-step approach: finetuning pretrained dialogue model and leveraging knowledge graphs for handling personal information queries
- Emily understands and responds appropriately to users' emotional states
- Differentiates between general open-domain dialogue utterances and personal information queries
- Effectiveness of Emily evaluated against several state-of-the-art chatbots, showing positive effects in influencing emotions and addressing personality inconsistencies
- Innovative approach for developing an emotion-affective open domain chatbot named Emily

## Summary

In the paper titled "Emily: Developing An Emotion-affective Open-Domain Chatbot with Knowledge Graph-based Persona," authors Weixuan Wang, Xiaoling Cai, Chong Hsuan Huang, Haoran Wang, Haonan Lu, Ximing Liu and Wei Peng present their work on developing Emily - an emotion-affective open-domain chatbot. The goal of Emily is to perceive a user's negative emotion state and provide support by positively converting their emotions. To achieve this, the authors propose a two-step approach. First, they finetune a pretrained dialogue model using data that captures dialogue contexts and desirable emotion state transitions across turns. This allows Emily to understand and respond appropriately to users' emotional states. Second, Emily is able to differentiate between general open-domain dialogue utterances and questions related to personal information. To handle personal information queries, the chatbot leverages a question-answering approach based on knowledge graphs. This ensures that Emily maintains consistency in its responses while handling sensitive user data. The effectiveness of Emily is evaluated by comparing it against several state-of-the-art open-domain chatbots. The results demonstrate the positive effects of the proposed approaches in influencing emotions and addressing personality inconsistencies. Overall, this paper presents an innovative approach for developing an emotion-affective open domain chatbot named Emily. By incorporating techniques such as finetuning dialogue models and leveraging knowledge graphs for handling personal information queries, Emily demonstrates improved performance in understanding and responding to users' emotions while maintaining consistency in its interactions.