개념적 거리와 밀도를 기반으로 한 의미적 웹 문서 검색시스템

Abstract

Nowadays, there is so much quantity of lecture data acquired through the development and spread of internet techniques that we can easily get lecture data. We can obtain lots of data formed in various patterns ranging from simple text to multimedia data such as voice, movie and etc. But the data of custom templates can't be obtained easily using query in the web. So describing and organizing this vast amount of content is essential for realizing the full potential of the web as an information resource. We should consider classification within larger context of subject-based metadata. Automatic classification is needed for at least two important reasons. The first is the sheer scale of resources available on the web and their ever-changing nature. And the second reason is that classification itself is a subjective activity. Different classification schemes are needed for different applications. Specific domain classification schemes, which can be quickly applied to large amounts of content using automatic methods, hold great promise for generating effective metadata. If web content shifts from text-based pattern to multimedia oriented one, metadata will become even more important. Semantic interpretation of multimedia is incomplete without some mechanism in understanding semantic content which is not directly visible. For this reason, human assisted content-annotation using natural language is one of the most common methods, particularly in multimedia retrieval applications, and it provides the means for exploiting syntactic, semantic as well as lexical information. A simple form of human-assisted semantic annotation is an attachment of textual description (such as keyword, or a simple sentence) to multimedia data. Problems in text (such as keywords) based image retrieval require exact remembrance of annotating words for retrieval and efficient management of semantic or concept clusters. More advanced technique like thesaurus-based term rewriting can be used to relieve the problem of traditional keyword-based information retrieval. However, keyword-based retrieval is still in the level of syntactic pattern matching. In other words, dissimilarity computation among terms is usually done by string matching not concept matching. In this paper, we propose the way of classifying document by the subject using similarity measure or conceptual density based on subject classification of U-WIN from dispersed web data. Specially, we focus on similarity measure and conceptual density for automatic subject classification.