A Measure of Similarity between Graph Vertices: Applications to Synonym Extraction and Web Searching