Cross-Course and Multi-course Sentiment Classification of Student Posts

Foteini Dolianiti, Dimitrios Iakovakis, Sofia B. Dias, Sofia Hadjileontiadou, José A. Diniz, Georgia Natsiou, Melpomeni Tsitouridou, Leontios Hadjileontiadis

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review


Affective Computing is one of the most active research topics in education. Increased interest in emotion recognition through text channels makes sentiment analysis (i.e., the Natural Language Processing task of determining the valence in texts) a state-of-the-practice tool. Considering the domain-dependent nature of sentiment analysis as well as the heterogeneity of the educational domain, development of robust sentiment classifiers requires an in-depth understanding of the effect of the teaching-learning context on model performance. This work investigates machine learning-based sentiment classification on datasets comprised of student posts in forums, pertaining to two different academic courses. Different dataset configurations were tested, aiming to compare performance: i) between single-course and multi-course classifiers, ii) between in-course and cross-course classification. A sentiment classifier was built for each course, exhibiting a fair performance. However, classification performance dramatically decreased, when the two models were transferred between courses. Additionally, classifiers trained on a mixture of courses underperformed single-course classifiers. Findings suggested that sentiment analysis is a course-dependent task and, as a rule of thumb, less but course-specific information results in more effective models than more but non-specialized information.

Original languageBritish English
Title of host publicationBrain Function Assessment in Learning - 2nd International Conference, BFAL 2020, Proceedings
EditorsClaude Frasson, Panagiotis Bamidis, Panagiotis Vlamos
PublisherSpringer Science and Business Media Deutschland GmbH
Number of pages11
ISBN (Print)9783030607340
StatePublished - 2020
Event2nd International Conference on Brain Function Assessment in Learning, BFAL 2020 - Heraklion, Greece
Duration: 9 Oct 202011 Oct 2020

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume12462 LNAI
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349


Conference2nd International Conference on Brain Function Assessment in Learning, BFAL 2020


  • Education
  • Natural language processing
  • Sentiment analysis


Dive into the research topics of 'Cross-Course and Multi-course Sentiment Classification of Student Posts'. Together they form a unique fingerprint.

Cite this