Entity linking (EL) is the task to disambiguate entity names from text by linking them to a specific entity in a knowledge base (KB). While more and more features are exploited to help the task, document context, as the most important hint, is currently handled using traditional Bag-of-Words methods. These traditional methods have limitations in handling ambiguity and logical relationship of vocabulary terms. / We propose a new context model for EL that makes use of ontology structure to handle ambiguity and logical relationship of document context. We also decompose the semantic meaning of entities and use a different ontology for each semantic dimension. This model benefits from: 1) sensitivity to ambiguous terms; 2) sub-concept relationship captured by the ontology structure and; 3) precise representation of semantic information by the multiple dimension design. The new model yields significantly better performance over existing methods from the experiments. / 實體鏈接是為文本中的本體名稱鏈接一個在知識庫內的實體，以消除歧義。雖然有越來越多的特徵項被提出及採用，當中最重要的上下文資訊卻一直透過經典的詞語包方法來運用。這頖經典方法在處理歧義和詞彙的邏輯關係方面存在一定限制。 / 為此我們提出一個新的用於實體鏈接的上下文模型。它運用本體結構來處理上下文中的歧義和邏輯關係，更把實體的語意分拆，並使用不同的本體來分別處理不同語意方向。此模型的優勢來自以下幾點︰一、對帶歧義詞項的感受性；二、通過本體結構妥善利用子概念關係；三、被分拆在不同語意方向讓語意可得到更精確的處理。實驗證實，使用此模型取代現有方法能顯著地提升實體鏈接表現。 / Chung, On Yip. / Thesis M.Phil. Chinese University of Hong Kong 2014. / Includes bibliographical references (leaves 77-79). / Abstracts also in Chinese. / Title from PDF title page (viewed on 18, October, 2016). / Detailed summary in vernacular field only. / Detailed summary in vernacular field only.