Target-Guided Dialogue Response Generation Using Commonsense and Data Augmentation

Anonymous

Abstract: Targeted-guided response generation enables dialogue systems to smoothly guide a conversation from a dialogue context toward a target sentence. Such control is useful for designing dialogue systems that direct a conversation toward specific goals, such as providing counselling and creating non-obtrusive recommendations. In this paper, we introduce a new technique for target-guided response generation, which first finds a bridging path of commonsense knowledge concepts between the source and target, and then uses the identified bridging path to generate transition responses. Additionally, we propose techniques to re-purpose existing dialogue datasets for target-guided generation. Finally, we demonstrate the shortcomings of existing automated metrics for this task, and propose a novel evaluation metric that we show is more effective for target-guided response evaluation. Our experiments show that our proposed evaluation metric is reliable and our techniques outperform baselines on the generation task. Our work generally enables dialog system designers to exercise more control over the conversations that their systems produce.