Syndicom: Improving Conversational Commonsense With Error-Injection and Natural Language Feedback

Christopher Richardson, Larry Heck


In Sessions:

Sigdial Oral Session 3: Dialogue modeling and evaluation: (Thursday, 10:30 CEST, Sun I , Watch on Zoom , Chat on Discord )

Abstract: Commonsense reasoning is a critical aspect of human communication. Despite recent advances in conversational AI driven by large language models, commonsense reasoning remains a challenging task. In this work, we introduce &#92;textsc{Syndicom} - a method for improving commonsense in dialogue response generation. &#92;textsc{Syndicom} consists of two components. The first component is a dataset composed of commonsense dialogues created from a knowledge graph and synthesized into natural language. This dataset includes both valid and invalid responses to dialogue contexts, along with natural language feedback (NLF) for the invalid responses. The second contribution is a two-step procedure: training a model to predict natural language feedback (NLF) for invalid responses, and then training a response generation model conditioned on the predicted NLF, the invalid response, and the dialogue. <p>&#92;textsc{Syndicom} is scalable and does not require reinforcement learning. Empirical results on three tasks are evaluated using a broad range of metrics. &#92;textsc{Syndicom} achieves a relative improvement of 53&#92;% over ChatGPT on ROUGE-1, and human evaluators prefer &#92;textsc{Syndicom} over ChatGPT 57&#92;% of the time. We will publicly release the code and the full dataset.