Supporting Asynchronous Discussion Text Analysis with an Automatic Coding Approach

Content analysis is an important method in the research of computer supported collaborative learning. However, the general manual coding method for deeper content analysis is time consuming, which greatly limits the amount and scale of the analyzed content. On the other hand, the traditional text categorization technology has been used to support the interaction texts auto-coding, but the effect is not satisfied. Therefore, this paper proposes a self-learning method based on symbol, syntax and context to automatically code the interaction text. First, according to the symbol rule base, the interaction text is automatically pre-coded with several candidate codes, then the probably best code is selected from the candidates based on the syntax rule base, and finally the code type result is revised by considering the context rule base. To verify the proposed method, we took 702 interaction records as the corpus and conducted an experiment. The experiments results show that our method is feasible and effective to support text auto-coding. Furthermore, compared with the traditional categorization method, the presented method achieved a better coding result with the precision rate 76.5% and recall rate 85.8%.