Analysing the potential of seq-to-seq models for incremental interpretation in task-oriented dialogue
This work addresses incremental interpretation in dialogue systems, but it is incremental as it builds on prior research with limited new findings.
The study investigated how encoder-decoder models handle disfluencies in task-oriented dialogues, finding that disfluencies have minimal impact on task success but adding them to training data improves model representations.
We investigate how encoder-decoder models trained on a synthetic dataset of task-oriented dialogues process disfluencies, such as hesitations and self-corrections. We find that, contrary to earlier results, disfluencies have very little impact on the task success of seq-to-seq models with attention. Using visualisation and diagnostic classifiers, we analyse the representations that are incrementally built by the model, and discover that models develop little to no awareness of the structure of disfluencies. However, adding disfluencies to the data appears to help the model create clearer representations overall, as evidenced by the attention patterns the different models exhibit.