NLP Text Analysis on Influential Factors of Persistence for Female Computer Science Undergraduates: A Comparative Study

Abstract

Despite a steady increase in the number of Computer Science (CS)/Software Engineering (SE) bachelor’s degrees conferred by postsecondary institutions in the past decade, a critical gap exists globally between enrollment and degree completion rates among female students. The purpose of this research was to comparatively investigate, describe and analyze influencing factors of persistence correlated to the experiences and perceptions of female undergraduates in pursuit of their CS/SE degree. The study was conducted in two geographically, culturally, and economically distinct public universities in the U.S. & Ethiopia. Can persistence factors in pursuit of a computing degree in Ethiopia be predicted by factors reported to be effective in the U.S. or vice-versa?Automatic thematic analysis of the response data using sentiment analysis and Latent Dirichlet Allocation (LDA) topic modeling from the domain of Natural Language Processing (NLP) was used to provide a degree of objectivity to the interpretation of interview transcripts and write-in portions of online surveys. Quantitative NLP text analysis of qualitative data has scant literature precedence. Independent samples t-tests were applied to the computed mean of average sentiment scores for each auto-transcribed interview response, evaluating the hypotheses that state no gender difference in the perceptions of cognitive gains, the institutional difference in sentiment scores of course review, and mean difference in sentiment scores between the first two introductory CS courses. Moreover, this study identified and enumerated twelve contributing factors of success for graduating female seniors in CS/SE using LDA topic modeling through the algorithmic computation of auto-transcribed interview responses.