Among the various challenges regarding distance education is the necessity of reducing the student dropout rate. In this sense, the present research aimed to contribute to the design of a lexical database focused on emotions and opinions that can be incorporated into a predictive evasion software. For the database design, we used the Scup tool to collect 150 tweets containing distance education students’ opinions and analyzed them in the light of Martin and White’s Appraisal Framework, along with five resources related to the sentiment Analysis field, which were taken from Liu’s work. In addition, we used the Aulete dictionary to describe the lexical units found in our corpus to better fit them into the analysis categories. Results showed 220 opinion tokens, which were identified and labeled according to their polarity. Moreover, these tokens were included in the domains attitude (judgment and appreciation) and graduation (sharp and strong) from the linguistic framework used. The results also indicated the necessity of another resource to help identify the use of figurative language, slangs, and extralinguistic elements, such as GIFS and emojis.