Large language models (LLMs) are increasingly prevalent in conversational systems due to their advanced understanding and generative capabilities in general contexts. However, their effectiveness in task-oriented dialogues (TOD), which requires not only response generation but also effective dialogue state tracking (DST) within specific tasks and domains, remains less satisfying. In this work, we propose a novel approach FnCTOD for solving DST with LLMs through function calling. This method improves zero-shot DST, allowing adaptation to diverse domains without extensive data collection or model tuning. Our experimental results demonstrate that our approach achieves exceptional performance with both modestly sized open-source and also proprietary LLMs: with in-context prompting it enables various 7B or 13B parameter models to surpass the previous state-of-the-art (SOTA) achieved by ChatGPT, and improves ChatGPT’s performance beating the SOTA by 5.6% average joint goal accuracy (JGA). Individual model results for GPT-3.5 and GPT-4 are boosted by 4.8% and 14%, respectively. We also show that by fine-tuning on a small collection of diverse task-oriented dialogues, we can equip modestly sized models, specifically a 13B parameter LLaMA2-Chat model, with function-calling capabilities and DST performance comparable to ChatGPT while maintaining their chat capabilities. We have made the code publicly available at https://github.com/facebookresearch/FnCTOD.

Frame semantic parsing is an important component of task-oriented dialogue systems. Current models rely on a significant amount training data to successfully identify the intent and slots in the user’s input utterance. This creates a significant barrier for adding new domains to virtual assistant capabilities, as creation of this data requires highly specialized NLP expertise. In this work we propose OpenFSP, a framework that allows for easy creation of new domains from a handful of simple labels that can be generated without specific NLP knowledge. Our approach relies on creating a small, but expressive, set of domain agnostic slot types that enables easy annotation of new domains. Given such annotation, a matching algorithm relying on sentence encoders predicts the intent and slots for domains defined by end-users. Experiments on the TopV2 dataset shows that our model trained on these simple labels have strong performance against supervised baselines.