Student stress is a problem that hinders the teaching-learning processes, and that has increased considerably since the beginning of the Covid-19 pandemic. This article introduces a framework for the development of an emotion-based decision support tool for learning processes. As a case study, we consider undergraduate students starting their academic year virtually in the context of a pandemic. Through the application of the PANAS questionnaire and NLP techniques on free-text responses, students' emotions are automatically classified as positive and negative, as well as a level of basic emotions of the Plutchik model. The results allow to identify the most frequent sentiments in students. Also, they show concordances between both measurement instruments and a high capacity for the classification of emotions.