Creation and annotation of a code-mixed Kannada english dataset with accurate labels for detecting depression and major depressive disorder categories

Abstract The research addresses mental health challenges within multilingual communities, with focus on individuals who frequently share their emotion, feelings and depressive thoughts by blending Kannada and English languages. The study includes creating a dataset that will be annotated with a label to detect depression and major depressive disorder categories. A code-mixed dataset was collected from various sources to construct the dataset. After data collection, it is processed using multilingual tools, and clinical experts provide guidance in annotating the dataset to accurately detect the depressive disorders. By examining key linguistic features related to depression—including emotional tone, grammatical structures, and code-switching patterns—the study offers a valuable insight into how depression is expressed across languages. The labelled dataset is crucial for developing deep learning models capable of accurately identifying and quantifying depression, particularly in multilingual environments. Additionally, the study emphasizes the role of language in mental health discourse and its potential impact on clinical practices and interventions for linguistically diverse populations. The practical implications of the research include using the dataset for digital mental health monitoring, supporting clinical decision-making, and creating culturally mental health tools for underrepresented multilingual communities.