Improving Information-Carrying Data Capacity in Text Mining

摘要

In this article the relation between the selection of textual data representation and text mining quality has been shown. Due to this, the information-carrying capacity of data has been formalized. Then the procedure of comparing information-carrying data capacity with different structures has been described. Moreover, the method of preparing the y -gram representation of a text involving machine learning methods and ontology created by the domain expert, has been presented. This method integrates expert knowledge and automatic methods to develop the traditional text-mining technology, which cannot understand text semantics. Representation built in this way can improve the quality of text mining, what was shown in the test research.