Computación y Sistemas
versión impresa ISSN 1405-5546
Knowledge is the most valuable treasure of humankind. Most of this knowledge exists in natural language format, for instance, in books, journals, reports, etc. The real possession of all this knowledge depends on our capabilities to perform different tasks with texts, such as: searching for interesting texts, comparing different documents, and summarizing them. Text mining, an emerging research area that can be roughly characterized as knowledge discovery in large text collections, is focused on automatically analyzing a set of texts. Mainly, it is concerned with the discovery of interesting patterns such as clusters, associations, and deviations from large text collections. Current methods of text mining tend to use simplistic and shallow representations of texts, e.g., keyword sets or keyword frequency vectors. On one hand, such representations are easy to obtain from texts and easy to analyze, but on the other hand, however, they restrict the knowledge discovery results to the topic level. To obtain more useful and meaningful results, richer text representations are necessary. On the basis of this assumption, we propose a new method for doing text mining at detail level. This method uses conceptual graphs for representing text content and relies on performing some tasks on these graphs, allowing the discovery of more descriptive patterns.
Palabras llave : Text Mining; Conceptual Graphs; Conceptual Clustering; Knowledge Discovery.