A Conceptual Model for Retrieval of Chinese Frequently Asked Questions in Healthcare

摘要

Frequently asked questions (FAQs) in healthcare provide general readers with both reliable and readable healthcare information. In this paper, we present a conceptual retrieval technique that serves as a supplement to enhance existing FAQ retrievers to find Chinese healthcare FAQs for each input query. By analyzing the structures and goals of Chinese healthcare FAQs, we identify three types of essential concepts in healthcare FAQs: event, condition, and aspect, as a Chinese healthcare FAQ often cares about some aspects (e.g., cause) of some events (e.g., cardiovascular disease) under some condition (e.g., patients of the periodontal disease). The proposed conceptual retrieval technique is thus named ECA (Event, Condition, and Aspect). Given healthcare FAQs annotated by the three types of concepts, ECA can measure the conceptual similarities between an input query and the FAQs. Empirical evaluation on real-world Chinese healthcare FAQs shows that the conceptual similarity information provided by ECA is helpful for an FAQ retriever to have significantly better performance in identifying relevant FAQs for input queries.