The digital mediation resources used in Distance Education can hinder the teacher's perception about the student's state of mind. However, the textual expression in natural language is widely encouraged in most Distance Education courses, through the use of Virtual Learning Environments and other digital tools. This fact has motivated research efforts in order to identify feelings expressed by students in textual messages. A significant part of the known approaches in this area apply textual analysis without a deep linguistic representation, which can lead to some weakness in the results obtained. This paper reports an approach using theories of Computational Lexical Semantics for the representation of the lexicon of emotion. The methodology was developed through studies regarding corpus analysis, lexical unit description, and the implementation of a computational system to identify the feelings expressed in the textual messages in natural language, using the lexicon of emotion. This system was used in evaluation experiments that indicate improvements when comparing the adopted approach with other similar approaches.