Abstract:
Online collaborative learning is a student-centered learning that students can express their opinions via an online discussion board. The discussion can reflect the collaboration behavior among them. However, such discussions do not have concrete patterns which make difficult of interpretations. The classification of discussion sentences on the board is an essential mechanism used to describe students interaction patterns, in order to consider the progress and/or the problems of collaboration. Text mining technic was used to classify the discussion sentences into 5 types, according to the sentence function, including declarative sentence, compound sentence, negative sentence, interrogative sentence, and imperative sentence. In order to overcome the problems of uncertain patterns of discussion sentences, feature extraction was used to extract the features of Thai sentences before classification step.
The performance of classification models were compared in 2 dimensions. The first dimension was performance of feature extraction, which can extract the sentences into 2 main features, i.e. Term Frequency (TF) of part-of-speech (POS), and Term Binary (TB) of key phrase with Term Frequency (TF) of part-of-speech (POS). The second dimension was performance of classification, which 4 algorithms Support Vector Machine (SVM), Decision Tree, Naïve Bayes, and k-Nearest Neighbor were used. The results of comparisons showed that, in the 1st dimension, the performance of TB with TF of POSwas higher than the performance of only TB feature extraction. For the 2nd dimension, the performance of SVM was highest, then Decision Tree, Naïve Bayes, and k-Nearest Neighbor respectively.
The research experiment was performed on the discussion sentences of collaborative learning group. The result showed that analysis of Thai sentences in collaborative learning can identify the interaction behavior in group using the frequency of each type of sentences. Such identification can be divided into 2 levels, including the sentence frequency in each group and the sentence frequency that each student used for communication in group. The research results can be used to recommend for collaborative learning activities in the future.