# Towards Knowledge-Based Recommender Dialog System

**Authors:** Qibin Chen, Junyang Lin, Yichang Zhang, Ming Ding, Yukuo Cen, Hongxia Yang, Jie Tang  
**arXiv:** [1908.05391](https://arxiv.org/abs/1908.05391)  
**Date:** 2019-08-15  
**Summary page:** https://www.summarizepaper.com/en/arxiv-id/1908.05391v1/

## Key points

- Authors propose a framework called KBRD (Knowledge-Based Recommender Dialog System)
- Framework integrates recommender system with dialog generation system
- Dialog system improves recommendation system by incorporating knowledge-grounded information about users' preferences
- Recommender system enhances dialog generation system by providing recommendation-aware vocabulary bias
- Experimental results show significant advantages over baseline models in both dialog generation and recommendation evaluation
- Integrated systems bring mutual benefits by bridging the gap between them
- KBRD framework improves recommendation accuracy and dialog relevance
- Opens up possibilities for developing advanced knowledge-based recommender systems

## Summary

In this paper titled "Towards Knowledge-Based Recommender Dialog System," authors Qibin Chen, Junyang Lin, Yichang Zhang, Ming Ding, Yukuo Cen, Hongxia Yang, and Jie Tang propose a novel framework called KBRD (Knowledge-Based Recommender Dialog System). The framework integrates a recommender system with a dialog generation system to enhance their respective performances. The authors highlight that the dialog system improves the recommendation system by incorporating knowledge-grounded information about users' preferences. This integration allows for more personalized and accurate recommendations. On the other hand, the recommender system enhances the dialog generation system by providing recommendation-aware vocabulary bias. This ensures that the generated dialogs are relevant and aligned with users' preferences. To evaluate the effectiveness of their proposed model, experimental results are presented which demonstrate significant advantages over baseline models in both dialog generation and recommendation evaluation. Furthermore, a series of analyses reveal that the two integrated systems bring mutual benefits to each other by effectively bridging the gap between them. Overall, this paper introduces an innovative approach to recommender systems and dialog generation by integrating them into a unified framework called KBRD. The experimental results validate its effectiveness in improving both recommendation accuracy and dialog relevance while opening up new possibilities for developing advanced knowledge-based recommender systems with enhanced user experiences.