Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

PaRT: Enhancing Proactive Social Chatbots with Personalized Real-Time Retrieval

Zihan Niu, Zheyong Xie, Shaosheng Cao, Chonggang Lu, Zheyu Ye, Tong Xu, Zuozhu Liu, Yan Gao, Jia Chen, Zhe Xu, Yi Wu, Yao Hu

TL;DR

PaRT tackles the engagement limitations of passive social chatbots by introducing a proactive framework that combines personalized real-time retrieval with generation. It integrates user profiling, an intent-guided query refiner, and retrieval-augmented generation to produce topic-driven, knowledge-grounded responses using real-time data from RedNote, with top-$k$ retrieved passages guiding the reply. Offline evaluations demonstrate clear gains in retrieval precision and generation quality, while online deployment reports a robust 21.77% increase in average dialogue duration, confirming improved user engagement in production. Overall, PaRT proves the viability of real-world, proactive chatbots that sustain long-form conversations through personalized retrieval and context-aware prompting.

Abstract

Social chatbots have become essential intelligent companions in daily scenarios ranging from emotional support to personal interaction. However, conventional chatbots with passive response mechanisms usually rely on users to initiate or sustain dialogues by bringing up new topics, resulting in diminished engagement and shortened dialogue duration. In this paper, we present PaRT, a novel framework enabling context-aware proactive dialogues for social chatbots through personalized real-time retrieval and generation. Specifically, PaRT first integrates user profiles and dialogue context into a large language model (LLM), which is initially prompted to refine user queries and recognize their underlying intents for the upcoming conversation. Guided by refined intents, the LLM generates personalized dialogue topics, which then serve as targeted queries to retrieve relevant passages from RedNote. Finally, we prompt LLMs with summarized passages to generate knowledge-grounded and engagement-optimized responses. Our approach has been running stably in a real-world production environment for more than 30 days, achieving a 21.77\% improvement in the average duration of dialogues.

PaRT: Enhancing Proactive Social Chatbots with Personalized Real-Time Retrieval

TL;DR

PaRT tackles the engagement limitations of passive social chatbots by introducing a proactive framework that combines personalized real-time retrieval with generation. It integrates user profiling, an intent-guided query refiner, and retrieval-augmented generation to produce topic-driven, knowledge-grounded responses using real-time data from RedNote, with top- retrieved passages guiding the reply. Offline evaluations demonstrate clear gains in retrieval precision and generation quality, while online deployment reports a robust 21.77% increase in average dialogue duration, confirming improved user engagement in production. Overall, PaRT proves the viability of real-world, proactive chatbots that sustain long-form conversations through personalized retrieval and context-aware prompting.

Abstract

Social chatbots have become essential intelligent companions in daily scenarios ranging from emotional support to personal interaction. However, conventional chatbots with passive response mechanisms usually rely on users to initiate or sustain dialogues by bringing up new topics, resulting in diminished engagement and shortened dialogue duration. In this paper, we present PaRT, a novel framework enabling context-aware proactive dialogues for social chatbots through personalized real-time retrieval and generation. Specifically, PaRT first integrates user profiles and dialogue context into a large language model (LLM), which is initially prompted to refine user queries and recognize their underlying intents for the upcoming conversation. Guided by refined intents, the LLM generates personalized dialogue topics, which then serve as targeted queries to retrieve relevant passages from RedNote. Finally, we prompt LLMs with summarized passages to generate knowledge-grounded and engagement-optimized responses. Our approach has been running stably in a real-world production environment for more than 30 days, achieving a 21.77\% improvement in the average duration of dialogues.
Paper Structure (14 sections, 2 equations, 2 figures, 4 tables)

This paper contains 14 sections, 2 equations, 2 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Details of the Proposed PaRT
  3. User Profiling
  4. Intent-guided Query Refiner
  5. Retrieval-Augmented Generation
  6. Experiments
  7. Implementation Details
  8. Evaluation Metrics
  9. Offline Experimental Results
  10. Retrieval Performance
  11. Generation Performance
  12. Impact of Retrieval Quantity on PaRT Performance
  13. Online Experimental Result
  14. Conclusion

Figures (2)

  • Figure 1: The interface of PaRT. From left to right, we sequentially present the greeting interface, the dialogue page, and the user profile interface. The greeting scenario is initiated by chatbot at the beginning, while the dialogue scenario aims to proactively guide ongoing conversation.
  • Figure 2: The figure provides an overview of PaRT. It illustrates the different dialogue experiences based on traditional chatbot (left) and our PaRT method (right).