Scholarlink

Reader discretion is recommended as our study tackles topics such as child sexual abuse, molestation, etc. Online Social Media (OSM) is increasingly being used by individuals with different mental health problems for finding support groups. As such extensive research has been carried out for understanding the mental health of individuals by observing their activities on OSM. However, previous studies haven’t put much focus on studying mental health in victims of Childhood Sexual Abuse (CSA). CSA is a menace to society and has long-lasting effects on the mental health of the survivors. Proper care and attention towards CSA survivors facing mental health problems can drastically improve their mental health. Our work fills this gap by studying Reddit posts related to CSA. To this end, we collect and create a dataset of around 8192 CSA-related posts. For further understanding the characteristics of CSA-related posts, we performed a comparative analysis with 9198 non-CSA mental health-related posts using various natural language processing (NLP) techniques such as word-shift, word cloud, topic analysis, and emotion analysis. We found that observable differences exist between them in terms of topics being discussed and emotions. Additionally, we propose a modeling framework, Mental Feature-interaction Transformer (MentalFiT) for identifying mental health problems in posts associated with CSA. It follows the notion that mental health problems and emotions are related to each other. The modeling approach involves the use of a transformer encoder to facilitate interaction among features at the intra-level, while the feature interaction block is employed to facilitate interaction among features at the inter-level. Thorough and extensive experimentation conveys the efficacy of the proposed framework on our novel curated dataset. The proposed method demonstrates impressive performance compared to models that do not incorporate emotional features. Our study opens up a new perspective toward understanding mental health problems in CSA victims and will serve as a frontier for upcoming work in this direction. Datasets and codes will be made available here ( https://github.com/orchidchetiaphukan/CSA_ASONAM2024 ).

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Whispers of Trauma: Leveraging Social Media for Assessing Mental Health in Victims of Childhood Sexual Abuse

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Orchid Chetia Phukan,
  • Rajesh Sharma,
  • Arun Balaji Buduru

摘要

Reader discretion is recommended as our study tackles topics such as child sexual abuse, molestation, etc. Online Social Media (OSM) is increasingly being used by individuals with different mental health problems for finding support groups. As such extensive research has been carried out for understanding the mental health of individuals by observing their activities on OSM. However, previous studies haven’t put much focus on studying mental health in victims of Childhood Sexual Abuse (CSA). CSA is a menace to society and has long-lasting effects on the mental health of the survivors. Proper care and attention towards CSA survivors facing mental health problems can drastically improve their mental health. Our work fills this gap by studying Reddit posts related to CSA. To this end, we collect and create a dataset of around 8192 CSA-related posts. For further understanding the characteristics of CSA-related posts, we performed a comparative analysis with 9198 non-CSA mental health-related posts using various natural language processing (NLP) techniques such as word-shift, word cloud, topic analysis, and emotion analysis. We found that observable differences exist between them in terms of topics being discussed and emotions. Additionally, we propose a modeling framework, Mental Feature-interaction Transformer (MentalFiT) for identifying mental health problems in posts associated with CSA. It follows the notion that mental health problems and emotions are related to each other. The modeling approach involves the use of a transformer encoder to facilitate interaction among features at the intra-level, while the feature interaction block is employed to facilitate interaction among features at the inter-level. Thorough and extensive experimentation conveys the efficacy of the proposed framework on our novel curated dataset. The proposed method demonstrates impressive performance compared to models that do not incorporate emotional features. Our study opens up a new perspective toward understanding mental health problems in CSA victims and will serve as a frontier for upcoming work in this direction. Datasets and codes will be made available here ( https://github.com/orchidchetiaphukan/CSA_ASONAM2024 ).