Publication

Abstract : The rise of suicidal thoughts posted on social media platforms means that future strategies should be devised to create automated detectors that will be able to detect at-risk individuals and deliver the necessary interventions in time. Although there is a lot of research in English language content, there is acute deficiency in processing code-mixed text, specifically Hindi-English which finds wide usage in South Asian online communities. This void is filled in this paper by presenting a new code-mixed Hindi-English dataset of suicidal ideation detection based on 6,533 samples of data gathered on Reddit and supplemented with known methods of code-mixing generation. Our experiments are done with two transformerbased models, such as English-BERT, and IndicBERT. Our findings indicate that English-BERT has slightly better overall statistics, with an F1-score as 0.87 and an accuracy of 88.2, whereas IndicBERT has a higher recall (0.89) when used in code-mixed contexts, so it is better suited to the applications demanding a high recall rate. The contribution of the present work is not only a valuable language resource but also empirical information that can be used to create culturally and linguistically sensitive mental health monitoring systems. © 2025 IEEE.