Abstract
Students' real-time feedback has numerous advantages in education, however, analysing feedback while teaching is both stressful and time consuming. To address this problem, we propose to analyse feedback automatically using sentiment analysis. Sentiment analysis is domain dependent and although it has been applied to the educational domain before, it has not been previously used for real-time feedback. To find the best model for automatic analysis we look at four aspects: preprocessing, features, machine learning techniques and the use of the neutral class. We found that the highest result for the four aspects is Support Vector Machines (SVM) with the highest level of preprocessing, unigrams and no neutral class, which gave a 95 percent accuracy.
| Original language | English |
|---|---|
| Title of host publication | 2014 IEEE 26th international conference on tools with artificial intelligence |
| Subtitle of host publication | ICTAI 2014 10-12 November 2014, Limassol, Cyprus |
| Place of Publication | Piscataway |
| Publisher | Institute of Electrical and Electronics Engineers Inc. |
| Pages | 419-423 |
| ISBN (Print) | 9781479965724 |
| DOIs | |
| Publication status | Published - 2014 |
Keywords
- Accuracy
- Analytical models
- Education
- Niobium
- Real-time systems
- Sentiment analysis
- Support vector machines
- Educational Data Mining
- Feature Selection
- Real-time Feedback
- Sentiment Analysis