Towards a System that Predicts the Category of Educational and Vocational Guidance Questions, Utilizing Bidirectional Encoder Representations of Transformers (BERT)

In today's complex job landscape, educational and vocational guidance has emerged as a critical factor in determining successful integration. Families are increasingly acknowledging its value, actively participating in shaping their child's educational path. Responding to this need, we have created a system leveraging the BERT technique to categorize queries pertaining to educational and vocational orientation, drawing upon the principles of Holland's test. Text classification, particularly when it pertains to question classification, is a crucial component in the realm of natural language processing and represents a foundational aspect of artificial intelligence. Given the vast amount of textual data available today, a robust word processing system is essential. Transformers models like Bidirectional Encoder Representations of Transformers (BERT) have gained immense popularity in NLP due to their remarkable performance in various tasks. This article demonstrates the implementation of a multi-class classification using BERT, specifically focusing on questions related to educational and vocational guidance following Holland's RIASEC typology. Our model effectively categorizes each input question into one of four classes: The components of Activity, Occupations, Abilities, and Personality make up our dataset. The findings suggest that our methodology demonstrates competitive efficacy.