Content based visual mining of document collections using ontologies

Document collections are important data sets in many applications. It has been shown that content based visual mappings of documents can be done effectively through projection and point placement strategies. An important step in this process is the creation of a vector space model, in which terms selected from the text and weighted are used as attributes for the vector space. That step in many cases impairs the quality of the projection due to the existence, in the data set, of many terms that are frequent but do not represent important concepts in the user's particular context. This paper proposes and evaluates the use of ontologies for content based visual analysis of textual data sets as a means to improve the displays for the analysis of the collection. The results show that when the ontology effectively represents the data domain it increases quality of maps.