Abstract: With technological advancements, intelligent question answering has become crucial across various domains, with answer precision being a critical metric for evaluating these systems. Enhancing precision necessitates enhancing the accuracy of question similarity matching and the ranking of recall results. Building on current text vectorization techniques, experimental research has demonstrated that employing elastic hybrid matching and threshold segmentation processing of recall results can boost the precision of domain-specific question answering. Analysis of the experimental data from private domain Q&A on home appliances, we distill high-precision strategies for domain-specific question answering, offering theoretical insights to guide future intelligent question-answering approaches.

Key words: Intelligent question answering, Precision, Ranking, Elastic hybrid matching, Recall, Vectorization, Home appliances