Recently, the revolutionary transformations in social and political landscapes as well as the remarkable developments in artificial intelligence reinforced the importance of geography and spatial analyses in literary and cultural studies. This chapter proposes an analytical framework of topic modelling and sentiment analysis for exploring the connection between theme, place, and sentiment in 36 autobiographical narratives by or about women from the Middle East. In the proposed framework, a latent Dirichlet allocation and latent semantic analysis algorithm from topic modelling, TextBlob library for sentiment analysis are employed to detect the place names that come together and to point out the associated themes and emotions throughout the data source. The model gives a scoring of each topical clusters and reveals that the diasporic authors are more likely to write about their hometown than their current host land. The authors hope that the merging of topic modelling and sentiment analysis would be beneficial to literary critics in the analysis of long texts.
TopIntroduction
The present century’s fast sociopolitical upheavals have moved the issues of space and spatiality to the forefront of social, political, cultural, and literary studies (Tally Jr, 2017). In recent years, the explorations of auto-/biographical narratives by Middle Eastern writers have been marked with a growing focus on matters related to local and global migration, geographical displacement, and public opinion and sentiment (Asl, 2020). The contents of the diasporic women’s life accounts are set in locations that evoke distinct emotions and reactions: misogynist dystopian spaces of the homeland (Anishchenkova, 2014); place as sites of discipline and punish (Asl, 2019; Pourya Asl, 2022); emancipatory utopia of the (Western) host land (Sassoon, 2016); heterotopic reality of the cyberspace; utopian longing for the homeland (Golley, 2007); transgressive sites of the third spaces (Asl, 2018), etc. The field of life writing by/about Middle Eastern women has remained significantly under-researched: First, the available volume of critical reviews does not complement the large quantity of fictional works written by Middle Eastern women over the past few decades. Secondly, existing reviews are often rejected for their biased and inauthentic views of the women’s stories as presenting solely a subjective narrative of their country of birthplace. This particular problem emanates from the conventional data collection and analysis techniques that fail to provide accurate analysis, especially in relation to the analysis of spatial representations and emotion expressions in connection to dominant themes. Hence the significance of further investigation into the field with a computerized method, for analyzing literary works, remains a challenge more objectively with regard to the notions of theme, place and emotion. The intention of this study is to develop and adopt a computerized analytical method to explore contemporary auto-/biographical narratives by/about Middle Eastern women and examine the interconnection between the three aspects of content, space, and emotion. The goal is to identify the frequent spatial topics in the selected works and describe the common concerns as prevalent among those.
This study uses data science technique Natural Language Processing (NLP) to propose an analytical model of Topic Modelling and Sentiment Analysis for identifying theme and sentiment related to different spaces in a corpus of 36 auto-/biographical narratives. In the model, Latent Dirichlet Allocation (LDA), Latent Semantic Analysis (LSA) and Non-Negative Matrix Factorization (NMF) are us for Topic Modelling and results are compared in terms of accuracy. In order to detect the place names that come together and explain the existing connection between themes and sentiments as pointed out from the dataset, the TextBlob package for Sentiment analysis and Python are utilized as systematic techniques for analysis. Natural Language Processing is useful in parsing human language into structures that can be linked to other values.
This chapter starts with an introduction and a statement of the goal of study followed by literature review. Literature review focuses on text mining and opinion mining, topic modelling and sentiment analysis. After review of related literature, materials and methodology are stated which include the whole process of data science life cycle and techniques used to achieve the main goal of this study. Next, results and discussion are added and finally, paper concludes with remarks on implications of the study and future direction.