Twentieth International Conference on Technology, Knowledge, and Society

Abstract

This paper presents the automatic evaluation of academic documents using one of the techniques of natural language processing, the word embedding technique. The aim is to analyze three previously selected algorithms (FastText, Word2Vec and GloVe) by applying evaluation metrics associated with the word embedding technique. According to the results obtained, the two most efficient algorithms will be selected; to subsequently test them in a working model for academic documents; and that can be considered in the development of academic evaluations. That is, distractors will be automatically generated for multiple-choice questions in academic papers by entering a question and the correct answer.