Platform for automating the calculation of subjects similarity in academic transfers using Natural Language Processing (NLP)

• Grant number: 23/06156-4
• Support Opportunities: Research Grants - Innovative Research in Small Business - PIPE
• Duration: February 01, 2024 - October 31, 2024
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computer Systems
• Principal Investigator: Bruno Berchielli
• Grantee: Bruno Berchielli
• Host Company: Blox - Sistema Gamificado de Educação Por Competência Ltda
• CNAE: Desenvolvimento e licenciamento de programas de computador customizáveis
• City: São Paulo
• Associated scholarship(s): 23/16843-9 - Data collection - Input data; Data validation/cleaning; Data validation/cleaning; Development; Validation and Conclusion., BP.TT; 23/16648-1 - Platform for automating the calculation of subjects similarity in academic transfers using Natural Language Processing (NLP), BP.TT

Abstract

The objective of the project "Platform for automating the calculation of subjects similarity in academic transfers using Natural Language Processing (NLP)" is the automation of the analysis processes of transferring students between Educational institutions, a procedure currently carried out by academic professionals in a manual way. There are a growing number of students who seek educational institutions with the intention of validating studies that have already been carried out in other institutions. This process, however, is still carried out manually by educational institutions, which makes it expensive and slow. In a survey carried out by Blox, the average time for analysis by coordinators is from one to three hours for new processes - those in which there is no historical record of analysis carried out previously for the same institution and course of origin. A large higher education institution performs more than 100,000 analysis of this type annually. The objective of this project by Blox - Sistema Gamificado de Educação de Competencia Ltda., candidate for PIPE I, is the validation of the sROBERTa language model (scale-aware Bidirectional Encoder Representations from Transformers) of Natural Language Processing (NLP) in the analysis of equivalence between subjects. To achieve this goal, it will be necessary to have high accuracy in the calculation of semantic similarity and the consequent similarity between the results obtained automatically and those performed manually by specialists in the educational field. This project is a continuation of the work already developed at Blox, whose platform has already achieved a 3.6-fold reduction in the time taken to complete courses when compared to that spent by specialists. (AU)