基於向量空間模型之智慧型文件搜尋系統開發-以台灣醫療糾紛判決書為例

本研究使用1995年至2020年台灣司法院法學資料檢索系統之判決書文件，開發台灣醫療糾紛判決書文件搜尋系統，因近年來消費者意識抬頭，在台灣健保制度下醫療糾紛發生的可能性與風險日益增加，系統將有助於醫界進行醫療糾紛案例快速檢索，期望能成為醫療糾紛案例討論與實證研究的重要基礎建設。研究運用文件中的特徵詞表示成空間向量，根據空間向量相似度來找出關聯性較高的文件，將醫療糾紛判決書分為9大類作為分類效能比較，並比較預訓練之空間向量模型、案例文件庫訓練之空間向量模型量與TF-IDF詞權重調整效能，研究結果證明案例文件庫訓練之空間向量模型合併TF-IDF詞權重調整取得較高的分類精確度。

This research uses the judgment documents of the legal information retrieval system of the Taiwan Judicial Court from 1995 to 2020 to develop a search system for the judgment documents of Taiwan's medical disputes. In recent years, consumer awareness has risen, and the possibility and risks of medical conflicts under the Taiwan health insurance system are increasing. The system will help the medical community quickly retrieve medical dispute cases and become an essential infrastructure for case discussion and empirical research. The research uses the feature words in the documents to be expressed as space vectors, finds the documents with higher relevance according to the similarity of the space vectors, divides the medical dispute judgments into nine categories for classification efficiency comparison, and compares the pre-trained space vector models and cases The space vector model amount of the file database training and the TF-IDF word weight adjustment performance. The research results prove that the space vector model of the case file database training combined with the TF-IDF word weight adjustment achieves higher classification accuracy.