基於注意力的問答檢索與醫學實體識別模型之互動式健康照護機器人

Abstract

在醫療機構中，回答病患與陪病者關於疾病的問題被視為一個很重要的任務。而在現今醫療人力短缺與護病比增加的情況下，醫護人員可以為每位病患回答問題的時間也越來越少。然而，有研究顯示，正確的健康教育信息能積極改善患者的知識、態度和行為，因此，透過問答的方式將正確的健康照護資訊傳遞至關重要。本論文使用醫療院所提供的問答集，設計了一個兼顧效率與準確性的健康照護問答系統。 大多數現有方法在檢索階段時常會著重在詞彙匹配，未能關注醫學領域中的關鍵實體。在本論文中，我們開發了一個使用多個基於注意力機制模型來回答健康照護相關疑問的人機互動問答系統。基於注意力機制的Transformer模型將使用者的問題分別進行語義編碼和抽取醫學實體。系統會結合這兩個特徵到我們設計的融合模組中，與健康照護問答及進行比對，最後即時地提供使用者最準確的回覆。透過與使用者互動的歷史紀錄中提取的醫學實體資訊，本系統還會推薦相關的健康照護知識給使用者，以增強使用者與機器人系統之間的互動性，這有別於以往只針對使用者問題回覆的系統。

In healthcare facilities, answering the questions from the patients and their companions about the health problems is regarded as an essential task. With the current shortage of medical personnel resources and an increase in the nurse-to-patient ratio, staff in the medical field have consequently devoted less time to answering questions for each patient. However, studies have shown that correct healthcare information can positively improve patients' knowledge, attitudes, and behaviors. Therefore, delivering correct healthcare knowledge through a question answering system is crucial. This thesis focuses on designing an efficient and accurate healthcare question answering system, utilizing the special question-and-answer knowledge set provided by healthcare facilities as well as sources from the general web. Most existing works heavily rely on query’s lexical matches at the retrieval stage but fail to focus on the critical entities in the medical field from the query. In this thesis, we develop a healthcare question answering system that uses attention-based models to answer healthcare-related questions. Attention-based transformer models are utilized to efficiently encode semantic meanings and extract the medical entities inside the user query individually. These two features are integrated through our designed fusion module to match against the pre-collected healthcare knowledge set, so that our system will finally give the most accurate response to the user in real-time. By incorporating the extracted medical entities from the historical records of users’ entities of questions, the system will also recommend the relevant healthcare knowledge to augment the interaction between users and the question-answering robot system, which is different from the previous systems using traditional approaches that only give users replies to the specific questions.