Towards a system for predicting the category of educational and vocational guidance questions using bidirectional encoder representations of transformers (BERT)

Abstract

© 2020, World Academy of Research in Science and Engineering. All rights reserved. Educational and vocational guidance is a particularly important issue today, as it strongly determines the chances of successful professional integration into the increasingly difficult labor market. Families have understood this well since they have been interested, in the educational orientation of their child. In this sense, we have set up a system for classifying questions of educational and professional orientation in based on Holland's test using the BERT method. Text classification, particularly the classification of questions is a basic task in natural language processing which is a very profound concept in the field of artificial intelligence. As the most abundant data in the world today is in the form of texts, having a powerful word processing system is essential and is more than just a necessity. Recently, Transformers models such as Bidirectional encoder representations of transformers or BERT are a very popular NLP models known for producing remarkable results compared to other methods in a wide variety of NLP tasks. In this article, we demonstrate how to implement a multi-class classification using BERT. In particular, we explain the classification of questions concerning the field of educational and vocational guidance following the RIASEC typology of Holland. Our model allows us to obtain the category of each input question. In our case, we define four classes (Activity, Occupations, Abilities, and Personality) for the set of questions, which constitute our data set. The results of this approach demonstrate that our model achieves competitive performance.