<?xml version='1.0' encoding='UTF-8'?><metadata xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:dcterms="http://purl.org/dc/terms/" xmlns="http://dublincore.org/documents/dcmi-terms/"><dcterms:title>EvidenceNet Dataset (for HCC and CRC diseases)</dcterms:title><dcterms:identifier>https://doi.org/10.7910/DVN/649TSE</dcterms:identifier><dcterms:creator>Zong, Chang</dcterms:creator><dcterms:publisher>Harvard Dataverse</dcterms:publisher><dcterms:issued>2026-03-31</dcterms:issued><dcterms:modified>2026-03-31T20:30:19Z</dcterms:modified><dcterms:description>This is the public derived data release for EvidenceNet, a framework for constructing disease-specific, evidence-centric knowledge graphs from full-text biomedical literature published during 2010-2025 (around 500 articles for each disease). This package contains two released resources: EvidenceNet-HCC and EvidenceNet-CRC. For each disease, it provides record-level JSON files (evidence_nodes.json) and graph-level JSON files (evidence_graph.json) that preserve structured evidence records, normalized biomedical entities, typed evidence-evidence relations, provenance metadata, and evidence quality scores. The package also includes selected derived evaluation outputs for component validation, question answering, future link prediction, and target prioritization. Raw source PDFs, large third-party knowledge resources, and local cache files are not included.</dcterms:description><dcterms:subject>Engineering</dcterms:subject><dcterms:subject>Medicine, Health and Life Sciences</dcterms:subject><dcterms:subject>Evidence Knowledge Graph</dcterms:subject><dcterms:subject>Information Extraction from Full-Text Literature</dcterms:subject><dcterms:subject>Biomedical Reasoning</dcterms:subject><dcterms:date>2026-03-31</dcterms:date><dcterms:contributor>Zong, Chang</dcterms:contributor><dcterms:dateSubmitted>2026-03-30</dcterms:dateSubmitted><dcterms:license>CC0 1.0</dcterms:license></metadata>