KeyGraph and WordNet hypernyms for topic detection

The Vast number of publicly available unstructured information on web and their rapid growth pose a great challenge in understanding, managing and structuring the information. Topic modeling algorithms have been developed with the purpose of analyzing these unstructured data and obtain abstract topics and clusters from these data collections. KeyGraph is a word co-occurrence based algorithm for topic modeling. We provide an extension for KeyGraph algorithm by incorporating WordNet hypernyms for Keywords in the data collection. Our results show that incorporating hypernyms for KeyGraph algorithm would result improved topic and document clusters.