Sentiment analysis on educational datasets: a comparative evaluation of commercial tools

FOTEINI S. DOLIANITI, DIMITRIOS IAKOVAKIS, SOFIA B. DIAS, SOFIA J. HADJILEONTIADOU, JOSE A. DINIZ, GEORGIA NATSIOU, MELPOMENI TSITOURIDOU, PANAGIOTIS D. BAMIDIS, LEONTIOS J. HADJILEONTIADIS

Abstract

Sentiment analysis systems have been gaining increasing popularity for extracting information regarding students' affective state. Developing such systems from scratch is a challenging task, thus, many studies employ commercial, general-purpose tools that are not domain-specific. The aim of the present work is to comparatively evaluate the performance of five well-known commercial/academic sentiment analysis tools on two educational datasets and contrast it with the performance of educational domain-specific tools, at document and sentence level. Findings suggest that: a) different tools work better for specific datasets and analysis levels, b) depending on the dataset, a general-purpose tool might be a viable solution, and c) any method, domain-specific or general-purpose one, should be evaluated before employed.

Keywords

Sentiment analysis, opinion mining, education, commercial systems, benchmark

Full Text:

PDF

DOI: https://doi.org/10.26220/une.2987

View Counter: Abstract | 2 | times, and PDF | 1 | times



Educational Journal of the University of Patras UNESCO Chair | ISSN: 2241-9152 | Department of Educational Sciences and Early Childhood Education University of Patras

Pasithee | Library & Information Center | University of Patras