Show, Don't Tell: Demonstrations Outperform Descriptions for Schema-Guided Task-Oriented Dialogue
This work addresses the problem of reducing overhead for service developers in task-oriented dialogue systems, though it appears incremental as it builds on existing schema-guided approaches.
The paper tackles the challenge of building universal dialogue systems that generalize across domains by proposing to use labeled example dialogues instead of natural language descriptions to convey schema semantics, achieving state-of-the-art performance on zero-shot generalization benchmarks like the Schema-Guided Dialogue dataset and MultiWOZ leave-one-out benchmark.
Building universal dialogue systems that operate across multiple domains/APIs and generalize to new ones with minimal overhead is a critical challenge. Recent works have leveraged natural language descriptions of schema elements to enable such systems; however, descriptions only indirectly convey schema semantics. In this work, we propose Show, Don't Tell, which prompts seq2seq models with a labeled example dialogue to show the semantics of schema elements rather than tell the model through descriptions. While requiring similar effort from service developers as generating descriptions, we show that using short examples as schema representations with large language models results in state-of-the-art performance on two popular dialogue state tracking benchmarks designed to measure zero-shot generalization - the Schema-Guided Dialogue dataset and the MultiWOZ leave-one-out benchmark.