近年來，人們已習慣以網路進行資訊與知識之獲取，卻容易造成資訊過量等問題產生，因此延伸出關鍵字搜尋、文件分類等相關研究與技術並協助知識搜尋者能快速且有效取得資訊。雖有資訊搜尋窗口以縮小搜尋範圍，然而面對特定領域網站時，若無相關領域背景之知識搜尋者仍需不斷嘗試並取得回饋，關鍵字搜尋與文件分類乃缺乏有效地協助。因此，本研究乃以「勞工安全衛生知識網」為基礎，針對特定領域知識文件發展一套「結合語意解析與文件摘要技術之知識文件檢索」模式，以協助知識搜尋者有效地篩選資料與知識。 本研究所建立「結合語意解析與文件摘要技術之知識文件檢索」模式乃包含「知識文件表達項目解析模組」、「知識文件問答解析模組」與「知識文件結構化摘要推論模組」。首先，「知識文件表達項目解析模組」乃針對「勞工安全衛生知識網」之人因工程工作場所研究報告與技術叢書進行文件表達項目解析，以擷取此類型知識文件之表達方式、呈現內容以及相關性語彙，進而建構知識語彙庫並取得文件之觀念性語句，以作為知識文件問答解析之關鍵字語意比對與結構化摘要推論之依據。其次，「問答解析模組」乃進行文件關鍵字詞與搜尋字串語意相似性分析，透過語意相似性進而取得確實文件，最後，「結構化摘要模組」乃計算並擷取具有文件之代表性語句，並以制式摘要方式呈現於搜尋者，以避免個人閱讀偏好影響文件之篩選，進而加強勞工安全衛生知識網搜尋技術以提高知識網知識分享之成效。本研究除發展模式與方法論外，並依此方法論建構一套「結合語意解析與文件摘要技術之知識文件檢索」系統以進行案例驗證，從而確認方法論與技術之可行性。 整體而言，藉由知識語彙庫建立並透過觀念性語句擷取，並整合關鍵字索引與結構化摘要，以提供更方便、有效地方式解除知識搜尋者於專業領域中搜尋瓶頸，並能藉以上述整合技術更容易取得專業知識。 It is a common practice to acquire information and knowledge from the Internet; thus, keyword searching, document classification and other technologies have been developed to facilitate searching. Although the search engine sites can narrow down the scope of search, knowledge demanders without background knowledge in the specific fields need to continuously search and receive feedbacks. Hence, this paper develops a Knowledge Document Retrieval model using semantic analysis and document summarization technologies for domain knowledge documents. First, this paper analyzes the ergonomic technology reports from the website of “Institute of Occupational Safety and Health” to capture the expressions and related vocabulary of domain knowledge documents to develop the knowledge vocabulary database. Second, through the Question and Answer Analysis (QAA) module, the correlations between proper names and query strings can be obtained. Third, based on Conceptual Vocabulary Determination (CVD) module, the most conceptual or representative sentences of domain documents can be derived and serve as candidate sentences for structured summarization. Finally, the Document Structured Summarization (DSS) module is used to calculate and retrieve representative sentences of the documents and integrate them into summary for knowledge demanders. It is expected that knowledge demanders can directly read the desired parts according to problems to ensure they can find document they want within a short time. In order to demonstrate applicability of the proposed methodology, a web-based knowledge document retrieval system is also established based on the proposed model. Furthermore, the knowledge documents (i.e., ergonomic technology reports) from the website of “Institute of Occupational Safety and Health” are applied as examples to evaluate the proposed model. The verification results show that the developed system is a high-performance knowledge document retrieval system. As a whole, this research provides an approach for knowledge demanders to efficiently and accurately acquire the domain knowledge documents.
