檔案內容主題自動分類及其成效評估之研究

Research on the Automatic Subject Classification of Archives Content and its Effectiveness Evaluation

檔案分類係檔案管理工作之基礎，為瞭解自動分類應用於檔案實務之可行性，本研究採用兩種數位工具對《總裁批簽》檔案進行主題概念擷取與分群，並輔以專家學者調查評估歸類結果。研究目的在於：一、運用不同自動分類工具進行檔案內容概念分析及其主題自動分群實務之探討；二、歸納人工與數位工具分類的差異，促進檔案分類作業。研究發現：一、採用共現字歸類法的CATAR文獻內容探勘工具，其所類聚產生之分類表較符合《總裁批簽》檔案主題分類需求；二、專有名詞在人工或數位工具分類過程中，皆為判斷主題的重要依據；三、檔案主題分類應兼顧單件層級之主題詞彙擷取，較能滿足使用者需求。後續研究建議針對不同檔案內容屬性，可尋求更貼近需求之自動分類工具，並鼓勵強化檔案內容描述與全文數位化，以開展檔案內容探勘更多元化的運用。

Subject classification is the basis of archives management. In order to understand the feasibility of applying automatic classification to archives, this study used two digital tools to extract and group the subjects and concepts of the“President’s Approval”archives, and supplemented with assessment by content experts. The aims of this study include: first, using different automatic classification tools to analyze the concept of archives content and discuss the practice of automatic clustering of subjects; second, summarizing the differences between manual and digital tools to facilitate archival subject classification. The results indicate that: first, the CATAR clustered result via the co-words method is more consistent with the requirements of the“President’s Approval”for archives subject classification; second, proper nouns are important indicators in both manual and digital tool classification; finally, in addition to the overall assessment of the entire archives, the subject classification should also take into account the extraction of subject terms at the item to meet the needs of users. According to the results, follow-up research suggestions can be aimed at different archives attributes to find a better automatic classification tool closer to practical needs, and to encourage the description of archives content and the development of full-text digitization in archives to carry out more diversified applications of archive content exploration.