Measuring Semantic Relations of Web Sites by Clustering of Local Context

Our contribution in this paper is an approach to measure semantical relations within a web site. We start with a web page description by key words. The implementation of structural and content information reduces the variety of key words. Thereby, the document-key-word-matrix is smoothend and similarities between web pages are emphasized. This increases the possibility of cluster key words and identif topics successfully. To do so, we implement a probabilistic clustering algorithm. To assess semantic relations, we introduce a number of measures and interpret them.