以症狀語意資訊建構中醫標準化症狀詞庫之研究

中醫在「辨證論治」的過程中，症狀的收集是辨證成功的關鍵。在中醫自動診斷專家系統建構中，標準化的症狀收集更是重要的一環，也就是要有一致化的症狀描述語詞，然後再編整成中醫症狀本體，才能有效地進行中醫自動診斷專家系統的辨證，否則無統一的症狀描述則不利於辨證系統的運作。本研究依此原因建立了一套中醫症狀詞庫系統，用以完成中醫症狀標準化的問題。材料與方法：運用「症狀字詞比對」及「關鍵字詞比對」的模式進行自動化的症狀標準化。運用「中醫症狀知識本體」記錄標準化症狀之間的語意關係以利未標準化的中醫症狀標準化。研究結果：第一、已經標準化過的症狀，則會存於系統資料庫中，系統可透過症狀字詞比對正確地查詢到其所對應的標準症狀。第二是若未經標準化過的症狀，則不在中醫症狀詞庫中，系統則會自動化地依字詞語意將該症狀分割成一系列已知的關鍵字詞，再透過關鍵字詞比對的方式，選取出語意最接近的標準化症狀。第三是運用語意資訊，可以明顯地提昇系統查詢標準症狀的正確率由35%提昇至69%。討論：經過症狀字詞比對及關鍵字詞比對法的建立，可有效提昇自動標準化症狀的成功率。在標準化的過程中，症狀的屬性是很重要的一項，能在建入中醫症狀本體後，建立起症狀的語意關係。它們有包含的關係。「症狀類別」包含了許多不同的「症狀屬性」，有些相同，有些相異，而「症狀屬性」又包含了許多不同的「症狀屬性值」。只要利用少許的關鍵字詞，就能透過它們所形成的語義關係，找出所有可能的標準化症狀。只要這個症狀的知識本體越完整，系統所能找到的標準症狀將會越精確。結論：我們找出一個研究方法以快速地建立標準化的中醫症狀，不需要再為了蒐集整理資料而耗費大量時間。

關鍵字

中醫症狀詞庫；症狀標準化；語意資訊

並列摘要

In Tradition Chinese Medicine (TCM), the collection of symptoms is the key to the success of syndrome differentiation. In constructing the TCM automatic diagnosis expert system, standardized symptom collection is essential. That is to have a consistent description of the symptoms and then compiled into the TCM symptom ontology. In this way, the TCM automatic diagnosis expert system can be used for useful differentiation. For this reason, a set of TCM symptom thesaurus was established to complete the problem of TCM symptom standardization. Materials and methods: automated symptom standardization was performed using the model of ＂symptomatic word comparison＂ and ＂keyword comparison.＂ Using the ontology of TCM symptom knowledge to record the semantic relationship between standardized symptoms to facilitate the standardization of unstandardized TCM symptoms. Results: first, the standardized symptoms will be stored in the system database, and the system can correctly query the similar standard symptoms through the comparison of symptom words. Second, if the unstandardized symptoms are not in the TCM symptom database, the system will automatically divide the symptoms into a series of known keywords according to the meaning of the word, and then select the most similar standardized symptoms through keyword comparison. Third, the use of semantic information can significantly improve the accuracy of system query standard symptoms from 35% to 69%. Discussion: the success rate of automatic standardization of symptoms can be effectively improved through the establishment of symptom word comparison and keyword comparison method. In the process of standardization, the attribute of symptoms is fundamental, and the semantic relationship of symptoms can be established after symptoms are built into the TCM symptom ontology. The ＂symptom category＂ contains many different ＂symptom attributes,＂ some identical, some different, and ＂symptom attributes＂ contains many different ＂symptom attributes values.＂ With a few keywords, all the possible standardized symptoms can be identified through the semantic relationships they form. The more complete the knowledge of the symptoms, the more accurate the standardized symptoms can be found.