Text Analytics Model to Identify the Connection Between Theme and Sentiment in Literary Works
Nurul Najiha Jafery, Pantea Keikhosrokiani, Moussa Pourya Asl
Abstract
The rapid advancements in data science techniques and approaches have influenced disciplines, such as literary studies, that are particularly engaged in qualitative text analysis. This chapter aims to apply natural language preprocessing (NLP) to identify the connection between theme and sentiment in a corpus of six life writings by or about Iraqi people. To do so, the study uses Latent Dirichlet Allocation (LDA) from topic modeling and the two models of Gensim and Mallet. It also implements TextBlob dictionary to calculate the polarity and subjectivity scores to measure the sentiment for detected themes. Nine topics are extracted from both models. The extracted themes point to the prevalence of traumatic events that the authors have personally endured. Gensim works better than Mallet as it has high coherence score and relevant terms. In sentiment analysis, most of the themes appeared as positive. The application of LDA using Gensim also revealed that the selected life writings are shaped and influenced by the authors' personal feelings. It is hoped that the analytical models can encourage future studies to improve existing qualitative methods in literary studies.