| Summary: | 碩士 === 國立嘉義大學 === 資訊管理學系碩士班 === 97 === The role of content provider has shifted to the users on the website of Web 2.0. Such change has enriched and varied the information of content website, and the great quantity of information also reinforces the users’ needs for more effective search engine. Recently, it has become more and more popular for the web surfers to share videos or photos on the content-sharing website. Though traditional search engines focused on the analysis of textual content and have attained to a satisfactory effectiveness, they cannot effectively support the search of non-textual content, i.e., videos. Most of content-sharing websites have provided the tagging mechanism to enable users to classify and annotate their sharing contents. Basically, the user-provided tags can reflect the user’s perspective on the contents, and thus, different users may provide various tags for the same videos. As a result, using keyword match approach to find out the relevant videos makes the search engine suffer the problem of word mismatch and word ambiguity. Word mismatch refers to the phenomenon in which a concept is described by different terms, and word ambiguity means that a term may refer to different concepts. In video search, word mismatch will deteriorate the search precision, and on the other hand, word ambiguity will deteriorate the search utility (i.e., the ranking of search results). In this study, we propose a social-tag-based approach to identify the context words of user’s query. Our approach then uses the context words to expand user’s query and re-ranks the expansion search results. Our evaluation results suggest the performance of the proposed social-tag-based approach is comparable to or even more effective than the existent video search engine, e.g., YouTube.com under some evaluation scenario.
|
|---|
