Research on the Key Issues in Web Appearance Mining
Mar 14, 2014Author：
Current Web vision and mining communities focus on the content analysis and understanding for Web texts, images and videos.There is scant research on the analysis of the appearances of Web pages. Nevertheless, users'access for Web pages is a process of human-computer interaction. Besides content, there are several other factors such as visual attributes (e.g., aesthetics, complexty, readability, accessibility) of Web appearances that significantly affect users'access. The analysis and mining the visual attributes of Web pages is sumarized as a new research topic, namely, Web appearance mining. This project aims to investigate the key issues in the Web appearance mining. (1)Feature extraction for Web appearance. How to apply state-of-the-art image feature extraction methods,specifically, the local descriptor, into the feature extraction for Web appearances. (2)The analysis of multiple attributes of Web appearance. How to quantitively describe the correlations among visual attributes. (3)The evaluation/measurement model construction. How to train effective models in order to classify, score,or rank Web pages according to their appearances. (4)Applications. How to combine Web appearance mining with computer graphic techniques such as color transfer in order to automatically edit and enhance the appearances of Web pages. Web appearance mining is a cross-discipline research topic. This project will utilizes thoeries and methods from multiple research areas (including Web mining, machine learning, human-computer interaction, vision and graphcis) to promote the research on Web appearance mining.