A Few-shot Approach to Task-oriented Dialogue Enhanced with Chitchat
Armand Stricker, Patrick Paroubek · 2024
Large language models (LLMs) tuned for chat have recently been adopted for few-shot end-toend task-oriented dialogue (TOD), with some success.To further assess this method, we conduct experiments on two, more complex, taskoriented benchmarks that integrate elements of chitchat into the conversation.We enhance a few-shot baseline by adding zero-shot chitchat detection and implementing function calling for dialogue state tracking (DST).We focus on this initial step in the TOD pipeline as errors due to added chitchat at this stage have a higher chance of impacting overall performance.We find that this prompting method shows increased resilience to mixed-mode inputs and our enhanced pipeline allows for natural inter-mode conversations, as assessed through human evaluation.Our findings also suggest that the performance gap between fewshot prompting and supervised task-specific models is narrowing.