Reproduce “CARE: A Clue-guided Assistant for CSRs to Read User Manuals” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Weihong Du, Jia Liu, Zujie Wen, Dingnan Jin, Hongru Liang, Wenqiang Lei
A clue-guided assistant named CARE was developed to help Customer Service Representatives read user manuals by providing question-relevant responses and explicit 'clue chains' as explanations. This system reduced the time CSRs spent answering questions by 35.7% and improved response accuracy and reliability in an online deployment for training new CSRs.
Key insights
- Jointly reasoning over both procedural and factual knowledge, represented in a unified heterogeneous graph, is crucial for accurate and comprehensive understanding of user manuals.
- Providing explicit 'clue chains' as explanations enhances the utility of AI assistants by enabling human users (CSRs) to verify predictions, build trust, and maintain service quality.
- Self-supervised learning offers an effective approach to generate explainable Q&A data for complex, specialized documents like user manuals, mitigating the need for extensive manual annotation.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.