TRACE: A New Framework for Trustworthy Public Service Chatbots
A recent study presents TRACE (Trustworthy Retrieval-Augmented Conversational Engine), a framework aimed at enhancing constraint-aware suggestions in public service chatbots. This system tackles the issue of unreliable and chaotic public service directories by converting user inquiries into structured and semantic constraints for retrieval processes, utilizing a dual data representation model. The effectiveness of TRACE is assessed through a carefully selected statewide pantry directory and a synthetic query benchmark, showcasing its ability to improve the reliability of conversational systems in public services. The research can be found on arXiv with the identifier 2608.10176.
Key facts
- TRACE is a retrieval-based, constraint-aware framework for public service chatbots.
- It parses user queries into structural and semantic constraints for downstream retrieval.
- It uses a dual data representation schema.
- The system is designed to handle noisy and heterogeneous service directories.
- Evaluation was conducted using a curated statewide pantry directory and a synthetic query benchmark.
- The paper is titled 'TRACE: Trustworthy Retrieval-Augmented Conversational Engine'.
- The paper is available on arXiv with identifier 2608.10176.
- The research addresses the issue of unreliable recommendations from general-purpose LLM-based chatbots.
Entities
Institutions
- arXiv