# An Approach to Inference-Driven Dialogue Management within a Social Chatbot

**Authors:** Sarah E. Finch, James D. Finch, Daniil Huryn, William Hutsell, Xiaoyuan Huang, Han He, Jinho D. Choi  
**arXiv:** [2111.00570](https://arxiv.org/abs/2111.00570)  
**Date:** 2021-10-31  
**Summary page:** https://www.summarizepaper.com/en/paper/2111.00570v1/

## Key points

- Dialogue management approach based on logical inference
- Conversation modeled as collaborative inference process
- User utterances translated into symbolic predicate representation
- Synthesis of new predicates using graph matching techniques
- Selection and translation of synthesized predicates into English responses
- Enables understanding of latent semantics and flexible initiatives in conversations
- Collaboration with users to synthesize new knowledge
- Authors: Sarah E. Finch, James D. Finch, Daniil Huryn, William Hutsell, Xiaoyuan Huang, Han He, Jinho D. Choi
- Published in the 4th Proceedings of Alexa Prize (Alexa Prize 2020)
- Falls under categories: computer science (cs.CL) and artificial intelligence (cs.AI)
- Innovative approach for more intelligent and contextually relevant chatbot responses

## Summary

The paper titled "An Approach to Inference-Driven Dialogue Management within a Social Chatbot" presents a novel dialogue management approach for chatbots based on logical inference. Unlike traditional approaches that treat conversation as a sequence of response generation tasks, this approach models conversation as a collaborative inference process where speakers share information to generate new knowledge in real-time. The chatbot pipeline proposed in the paper consists of three stages. In the first stage, user utterances are translated into a symbolic predicate representation. This structured representation is then used in the second stage along with a larger knowledge base to synthesize new predicates using efficient graph matching techniques. Finally, in the third and final stage, the chatbot selects a small subset of these synthesized predicates and translates them into coherent English responses. This unique approach enables the chatbot to understand the latent semantics of user inputs and take flexible initiatives during conversations. By modeling conversation as an inference process, the chatbot can collaboratively synthesize new knowledge with users, resulting in responses that are both novel and coherent with the dialogue context. The authors of this paper include Sarah E. Finch, James D. Finch, Daniil Huryn, William Hutsell, Xiaoyuan Huang, Han He, and Jinho D. Choi. The paper was published in the 4th Proceedings of Alexa Prize (Alexa Prize 2020) and falls under the categories of computer science (cs.CL) and artificial intelligence (cs.AI). In summary, this paper introduces an innovative approach to dialogue management for chatbots by leveraging logical inference which allows for real-time synthesis of new knowledge during conversations leading to more intelligent and contextually relevant responses from the chatbot.