Summarizing Web sites automatically

Abstract. This research is directed towards automating the Web Site summarization task. To achieve this objective, an approach, which ap-plies machine learning and natural language processing techniques, is employed. The automatically generated summaries are compared to man-ually constructed summaries from DMOZ Open Directory Project. The comparison is performed via a formal evaluation process involving human subjects. Statistical evaluation of the results demonstrates that the au-tomatically generated summaries are as informative as human authored DMOZ summaries and significantly more informative than home page browsing or time limited site browsing. 1