Analyzing Topics Similarity in Undergraduate Syllabus using Embedding Techniques

In the academic environment, the syllabus document set up the work plan of learning-teaching process and the way of being developed. A fundamental part is the subjects that must be elaborated carefully care in order to correlate properly the sequence between related courses and not generate a complete overlapping among them. In order to detect the level of similarity between courses topics, this research implements an evaluation using embedding techniques such as Word2Vec, FastText and BERT. For this, courses syllabus belonging to the Professional School of Systems Engineering of the Universidad Católica de Santa Mariá (UCSM) were collected. According to the metrics values, the better approach was BERT because that includes attention mechanisms that put more value on the words around a word. With this proposal, decisions can be made at the managerial level. If the content similarity is greater than the established threshold, then they must be analyzed and readjusted to meet the quality requirements in teaching and the student develops the competencies of the graduate profile.