標題: 一個以本體論為基礎的自學影片推薦機制
An Ontology-based Recommendation Mechanism for Self-study Videos
作者: 楊媛如
羅濟群
資訊管理研究所
關鍵字: 自學影片;混合式推薦方法;本體論;相關回饋;Self-study Video;Hybrid Recommendation Mechanism;Ontology;Relevance Feedback
公開日期: 2011
摘要: 知識經濟的時代要如何創造有價值的知識,藉由知識分享與管理進而提供有效的學習,已成為一個重要的議題。影片具有比文字更容易使人理解的特性,因此愈來愈多人採用影片做為自學的媒介。然而,目前大部分的自學影片是預先製作的,故內容無法包含自學者本身的使用經驗。隨著網際網路普及化,自學者利用網路搜尋學習資源的需求將會持續增長。然而,目前的推薦系統缺乏一個可以提供推薦自學影片的推薦機制。故本研究提出一個以本體論為基礎的自學影片推薦機制,此機制是結合以項目為基礎的過濾、內容式過濾及協同過濾的混合式方法,透過本體論分析使用者正在瀏覽的自學影片,並結合使用者過去對於自學影片的相關回饋,將其他關聯使用者興趣的自學影片推薦給使用者。本研究針對提出的混合式推薦機制實作一自學影片推薦網站,並與其他相關推薦機制進行準確度的比較。實驗結果證明本研究提出的機制較其他推薦機制能將準確度提升57%,能更準確的推薦適合的自學影片給使用者。
How to create valuable knowledge and provide users for self-study through knowledge sharing and knowledge management in the knowledge-driven economy times has become an important issue. As the video is easier to understand for people than the text, more and more people use video as their self-study medium. But the present self-study videos are mainly pre-production, they can’t consist of learners’ self-study experience. With Internet getting more popular, the need to obtain information through search engine when learning something will increase than before. However, there are only a few video recommendation mechanism adapted to the self-study environment. In this paper, we propose an ontology-based self-study video recommendation mechanism (OSVRM) which integrates adopted ontology item-based filtering, content-based filtering, and collaborative filtering. According to the ontology of the current self-study video and the user feedback of the past self-study videos, OSVRM takes charge of analyzing and recommending self-study videos to users that correspond with their interests. We also implement a web-based system to accomplish the OSVRM and compare to other recommendation mechanism. Our experiment results show that the OSVRM can enhance 57% accuracy. Therefore, the OSVRM manifests its own better recommendation than conventional recommendation mechanism.
URI: http://140.113.39.130/cdrfb3/record/nctu/#GT079934519
http://hdl.handle.net/11536/50143
Appears in Collections:Thesis