Reflective Writing Analysis Approach Based on Semantic Concepts: An Evaluation of WordNet Affect Efficiency

Automatic analysis of reflective writing involves identifying indicator strings and using string matching or rule matching processes, which flag sections of a text containing reflective material. The problem with the string-based approach is its inability to deal with knowledge inference from the text, such as the content, context, relevance, clarity, and interconnection, which can be identified by semantic analysis. The semantic analysis depends mainly on mapping the text into stored knowledge sources, such as WordNet, and analyzing the associations in the underlying knowledge source. In this paper, a semantic-based approach for reflective writing analysis is proposed, in which the input text, which is being analyzed, is mapped into semantic concepts. Moreover, a machine learning (ML) approach for reflective writing identification and analysis has been implemented to overcome the limitations of rule execution and keyword matching. The proposed approach addresses the efficiency of using several effective concepts, correlated with effective words that are identified in WordNet-Affect. The input text is classified into reflective or non-reflective categories, after which the input text is classified into various reflective classes, based on the type of the document. Moreover, the concepts in WordNet-Affect are evaluated and analyzed to demonstrate their effects on classification and labeling tasks.