Response Generation in Longitudinal Dialogues: Which Knowledge Representation Helps?
This addresses the challenge of building dialogue systems for longitudinal conversations, which involve sparse, personal interactions over long periods, but the approach is incremental as it adapts existing models to a specific domain.
The paper tackled response generation in longitudinal dialogues by fine-tuning pre-trained language models (GePpeTto and iT5) and experimenting with different knowledge representations, including graph-based ones, to improve grounded responses, with evaluation through automatic metrics and human assessments of contextualization, appropriateness, and engagement.
Longitudinal Dialogues (LD) are the most challenging type of conversation for human-machine dialogue systems. LDs include the recollections of events, personal thoughts, and emotions specific to each individual in a sparse sequence of dialogue sessions. Dialogue systems designed for LDs should uniquely interact with the users over multiple sessions and long periods of time (e.g. weeks), and engage them in personal dialogues to elaborate on their feelings, thoughts, and real-life events. In this paper, we study the task of response generation in LDs. We evaluate whether general-purpose Pre-trained Language Models (PLM) are appropriate for this purpose. We fine-tune two PLMs, GePpeTto (GPT-2) and iT5, using a dataset of LDs. We experiment with different representations of the personal knowledge extracted from LDs for grounded response generation, including the graph representation of the mentioned events and participants. We evaluate the performance of the models via automatic metrics and the contribution of the knowledge via the Integrated Gradients technique. We categorize the natural language generation errors via human evaluations of contextualization, appropriateness and engagement of the user.