Artificial intelligence approach for detection and classification of depression among refugees in selected diasporic novels

Abstract

The primary objective of this research is to develop and implement an artificial intelligence (AI) approach for the detection and classification of mental breakdowns in literary texts. The study employs text analytics techniques, utilizing natural language processing (NLP) to extract and analyze data from six novels written by Afghan and Pakistani diasporic writers. The aim is to identify and classify the topics and sentiments related to depression in the selected narratives. To achieve these objectives, four algorithms for topic modelling are utilized, namely Latent Dirichlet Allocation (LDA), Latent Semantic Index (LSI), Hierarchical Dirichlet Process (HDP), and Non-negative Matrix Factorization (NMF). Additionally, a rule-based technique is applied for sentiment analysis using two Python libraries, VADER and TextBlob. For the classification of depression, four machine learning models are employed: Decision Tree (DT), Naïve Bayes (NB), K-Nearest Neighbor (KNN), and Support Vector Machine (SVM). The results indicate that HDP has the highest score in topic modelling with a score of 0.79. Furthermore, Vader provides more insightful sentiment analysis results. With a classification model accuracy of 68%, Naïve Bayes outperforms the other machine learning models. The findings suggest that the proposed model can efficiently predict all classes of depression, particularly when the dataset is balanced.