Analysis on reflective writing using natural language processing and sentiment analysis

Abstract

Natural Language Processing (NLP) opens up the possibility for a machine to help us human process the vast data out there. There are multiple branches of NLP, but this thesis focuses on sentiment analysis, more specifically for reflective writing analysis. However, to obtain accurate results, NLP model needs to be tailored for the specific application. In this thesis, an NLP model is developed to process survey results obtained from psychology course by building a word database and categorise the level of reflection using fuzzy logic system. The main processes involved in this work are mainly performing intensive literature review on reflective writing models and to design fuzzy logic rules which can categorise the various levels of reflection demonstrated by the students. The model is then used to analyse a total of 47 reflective journals collected from the survey responses of the students from the School of Education in UTM. These surveys need to be pre-processed using some tools such as Natural Language Tool Kit (NLTK) to be fed to the developed model. There is the need to build a word database with words specific to the psychology field to improve the accuracy. From the analysis of the students' survey, most of the students exhibit level 1-2 reflection.