Bilingual Intention-Driven Evidence Corpus for Prediabetes

This record provides the public v1.0 release of a bilingual intention-driven evidence corpus for prediabetes. The current release contains 3,242 structurally curated evidence units derived from 48 source documents, together with source-level metadata, release documentation, and field definitions. Each evidence unit is organized into standardized fields including document mapping, page anchors, bilingual claims (Chinese and English), evidence type, transferability, target population, and audience. The corpus is designed to support knowledge base construction, bilingual evidence retrieval, question answering, and just-in-time adaptive intervention (JITAI) / ecological momentary assessment (EMA) scenarios for health behavior support in populations at high risk of diabetes. Files included in this record:- public corpus files in JSONL, CSV, and XLSX formats;- source-document mapping files;- release notes and field definitions. This public release excludes reviewer-identifying information, internal quality-control traces, and non-public review metadata. It does not redistribute source full texts. Users should cite this record when using the public corpus in research, system development, or evaluation.