黃帝內經貫珠詞典庫之研發與應用

DEVELOPMENT AND APPLICATION OF HUANG-DI-NEI-JING CONCORDANCE THESAURUS

中醫經典黃帝內經之原文資料，經常被歷代醫學家引用到醫藥論著中，以闡述學術思想或醫學理論。本研究選取中醫經典資料庫NW2001，其內容包括：黃帝內經、金元四大家著作、景岳全書及張氏醫通等中醫經典約四百萬字。所有NW2001 典籍都交由一個文字剖析器，以BruteForce Algorithm 切割成N-連字串（N-Gram）詞典庫。結果發現4-Gram 是最佳化的庫詞典在NW2001的精確率（Precision）及回收率（Recall）分別為0.96 及0.86；在中醫典籍網際網路貫珠集應用上效率很高，資料庫經過索引，網頁顯示的空等時間（Downtime）能控制在8 秒內。由於內經詞句的專一性（Specificity）相當高，所以被萃取出來的內經4-Gram 詞典知識庫，便可經由反轉檔（Inverted File）索引到NW2001 資料庫有關中醫學術理論之出處。Huang-Di-Nei-Jing, one of the Classical Traditional Chinese Medicinal Literature (CTCML), was frequently referenced by other CTML authors to write their medical principles or ideas. In this study, a CTCML database NW2001 including: Huang-Di-Nei-Jing , Jin-Yuan-Si-Da-Jia, Jing-Yue-Quan-Shu, Cheng-Si-Yi-Tong totally around four million Chinese characters was used. A word parser, developed by a computer program, can separate words into N-Grams thesaurus, which was driven by the Brute Force Algorithm. Finally, we found that 4-Grams thesaurus indexed on the NW2001 database had Precision and Recall of 0.96 and 0.86 respectively. This optimized 4-Gram thesaurus was used to develop the CTCML world-wide-web. The database was indexed and well formed that the downtime has been controlled under 8 seconds to show each homepage on the WWW. The 4-Grams data set, which extracted from Huang-Di- Nei-Jing , formed a knowledge-based database. By using inverted file, the 4-Gram thesaurus of Huang-Di- Nei-Jing could be mapped into the NW2001 database. The mapped information had a high specificity relating to medical principles of CTCML.