Representing document semantics by means of graphs

Text files (documents) are the principal means for representing information and knowledge in several domains. Commonly, users store documents increasely in hard disks or online storage units. The users follow a personal organization based on one topic, but documents can contain several topics. Current search engines for documents are based on the file name or content, a search word or part of such word must exactly match. This paper proposes a method for representing document semantics taking into account the topics that documents contain. For this, a meta-graph is generated considering synonyms, hyponyms, and hypernyms from the relevant terms contained in the documents. The proposal have been compared against Google Desktop with interesting results.