Web Topic Representation Based on Multi-layer Semantic Model

Web topic is the theme indicated by Web pages. Some methods have been proposed to represent Web topic based on text mining. However, these methods canpsilat denote multimedia contents on Web page, such as images, audios, videos and so on. Consequently, text-based methods can´t represent Web topic exactly because most Web pages consist of many multimedia contents. This paper proposes a new approach, named multi-layer semantic model, to represent Web topic. Using this model, the semantics of varied contents contained by Web page can be involved. The model is composed of several semantic layers, including text layer, image layer, audio layer, video layer and other extensible layers. Web resources are located on different layers according to their types. Their relationships within one layer and between layers are represented by inner-layer links and cross-layer links respectively. This method can also bring benefits for the similarity computing between Web topics.