Intellimote: a hybrid classifier for classifying learners' emotion in a distributed e-learning environment

Abstract

A huge collection of textual, graphical, audio, and video contents are readily available on the Internet to be used for the purpose of learning. Sentimental feedbacks of learners posted at the end of many of these contents may be considered as genuine reactions of the learners who have gone through the contents. Such learners' sentiments are important inputs for judging the acceptability of a learning material. Analyzing such feedbacks using sentiment analysis techniques can identify the best reusable learning contents that may be used for developing new courseware. This can significantly reduce the time and effort of authoring, which is otherwise a difficult, time-consuming, and costly affair. This methodology can also be used for continuous assessment on the learning materials released for use. This paper presents a machine learning-based approach for emotion analysis in e-learning materials. It describes the design and experimental use of a sentiment analysis classifier that uses classifier combination rules to combine polarity scoring and a support vector machine (SVM). The present approach also gives an opportunity for users to train a lexicon on a very specialized set of data, pertaining to the domain of usage. This helps either to enhance polarity scores of certain words that appear more frequently or to add words that are completely missing from the lexicon, and contributes a great deal in determining the polarity scores.