Constrained Decoding for Neural NLG from Compositional Representations in Task-Oriented Dialogue
This work addresses the challenge of controlling and improving semantic correctness in neural NLG for task-oriented dialogue systems, though it appears incremental by building on existing Seq2Seq methods.
The paper tackles the problem of generating natural language responses from structured semantic representations in task-oriented dialogue by proposing tree-structured representations and a constrained decoding approach for Seq2Seq models, demonstrating promising results on a new weather dataset and the E2E dataset.
Generating fluent natural language responses from structured semantic representations is a critical step in task-oriented conversational systems. Avenues like the E2E NLG Challenge have encouraged the development of neural approaches, particularly sequence-to-sequence (Seq2Seq) models for this problem. The semantic representations used, however, are often underspecified, which places a higher burden on the generation model for sentence planning, and also limits the extent to which generated responses can be controlled in a live system. In this paper, we (1) propose using tree-structured semantic representations, like those used in traditional rule-based NLG systems, for better discourse-level structuring and sentence-level planning; (2) introduce a challenging dataset using this representation for the weather domain; (3) introduce a constrained decoding approach for Seq2Seq models that leverages this representation to improve semantic correctness; and (4) demonstrate promising results on our dataset and the E2E dataset.