Automation of discussion board evaluation through keyword extraction techniques: a comparative study
Gomathi Thiyagarajan, S. R. Mahadeva Prasanna, B. V. Uma · IOP Conference Series Materials Science and Engineering · 2021
Abstract With the current trend in lieu of the ongoing pandemic, online learning and technology-based pedagogy receive worldwide attention. There are several pedagogical practices one can use to evaluate students mindful and the method will vary based on the learning objective. In this paper, our goal is to dive deeper into identifying the best keyword extraction technique which can be closely mapped with human evaluation. We report the results of three popular keyword extraction techniques specifically RAKE, TF-IDF, and Semantic Fingerprinting on the dataset generated from the discussion board post of a Learning Management System. The results illustrate that the TF-IDF algorithm shows the highest correlation with a human evaluation with 0.76 correlations, RAKE with 0.36, and Semantic Fingerprinting with 0.58. It was also identified Semantic Fingerprinting had a lowest mean square error of 18.26.