Natural Language Processing Algorithms for Academic Content Generation

Paredes-Vargas A.; Salinas-Llorca P.; Santisteban J. · Proceedings of the 2024 IEEE 4th International Conference on Advanced Learning Technologies for Education and Research, ICALTER 2024 (2024) · DOI: 10.1109/icalter65499.2024.10819214

Resumen

This study has conducted several activities to analyze and compare the performance of five Natural Language Processing (NLP) models: GPT-3, T5, ERNIE, BERT, and XLNet for generating academic content in universities located in Lima, Peru. Some NLP models, including GPT and BERT, were assessed based on accuracy, quality, speed, and customization capabilities. The research examined the use of various algorithms for academic tasks like essay and summary creation. The findings show that all models perform adequately; however, some excel in generating coherent content tailored to different academic styles, thereby optimizing the time of both students and teachers. In conclusion, the ERNIE model emerges as one of the best, as it enables the generation of diverse and high-quality content efficiently.

Instituciones: Universidad Peruana de Ciencias Aplicadas