Exploiting Structural Similarity for Effective Web Information Extraction