Bangla Mental Health Corpora: English-Aligned and Mixed-Provenance Datasets with Analysis Code

This release contains two Bangla mental-health text resources and their analysis code. The English-Aligned Bangla Mental Health Corpus contains 65,188 aligned English–Bangla records across 17 content categories, with translation-quality features. The Bangla Mixed-Provenance Mental Health Corpus contains 14,372 Bangla texts across 13 categories: 1,372 publicly visible comments and 13,000 generated texts. The package includes the v1 and v2 synthetic components, record-level provenance, generator and analysis scripts, results, a datasheet, and a SHA-256 manifest. Labels describe text content for research and are not clinical diagnoses. Generated records describe no real person and must not be presented as testimony. Public comments may remain searchable; do not use these data to identify, profile, screen, or diagnose individuals. File-level rights and limitations are documented in the included README and datasheet.