This dissertation investigated the application of text mining process from techniques of computing intelligence and machine learning in the problem of word sense ambiguity. The work in the methods of decision support area aimed to develop techniques capable of doing a word meaning disambiguation automatically and also to construct a prototype based on the application of such techniques. Special attention was given to the process of ambiguity detection and, for this reason, a differentiated approach was used. Unlikely the most common type of disambiguation, in which the machine is trained to do it in determined terms, the present work aimed to address the ambiguity problem without the need of knowing the meaning of the term used, and thus, to make the system more robust and generic. In order to achieve that, specific heurists were developed based on computing intelligence techniques. The semantic criteria used to identify the ambiguous terms were extracted from grouping techniques employed in lexis built after some term normalization process.