DepressionFeature: Underlying ontology for user-specific depression analysis

Sumit Dalal, Sarika Jain, Mayank Dave

Abstract: Smart healthcare systems utilize machine learning (ML) and deep learning (DL) techniques to analyze patient data encompassing social media data, sensor data, and clinical records for depression detection. However, existing methodologies predominantly focus on syntactic cues, often overlooking semantic nuances. Despite numerous efforts to address this, prevailing ontologies lack comprehensive coverage of social media-oriented concepts and terms related to depression. To bridge the gap, this study extends the DepressionFeature ontology to incorporate user-specific data, offering an exhaustive range of textual social media-centric concepts. A semi-automatic approach is leveraged to scrutinize concepts and terms from textual social media data and pertinent literature. The resulting ontology is implemented in Ontology Web Language (OWL) using Protégé. Evaluation proceeds in three phases: initially, validation via criteria-based and metric-based methodologies; subsequently, validation through execution of knowledge graph (KG)-based SPARQL queries and RDF validator; finally, comparison with existing ontologies. Beyond depression detection, DepressionFeature facilitates sentiment analysis, user profiling, and other social data classification tasks. Future endeavors include its integration for interpreting ML and DL predictions.