About the SciBorg UK Editorial Board, Research Methods, and Corpus Sources
Kevin Nelson
Scientific Oversight and Natural Language Processing Governance
The SciBorg UK project applies natural language processing to extract discourse structure, named entities, and chemical reaction data from peer-reviewed scientific literature. Research standards are overseen by computational linguists and domain specialists who ensure rigorous methodological fidelity across all extracted corpora and open-access resources.
Corpus Coverage and Data Provenance
Corpus Layer
Source Provenance
Validation Standard
Chemical Entity Corpus
Full-text articles across 15 chemistry journals
Dual expert manual annotation with Cohen kappa above 0.85
Discourse Structure Set
Abstracts and introductions from 500 peer-reviewed papers
Cross-validated rhetorical zone classification (AZ/CZM)
Reaction Extraction Benchmarks
Patent chemistry archives and open-access datasets
Formula parsing evaluated against Gold Standard IUPAC records
Four-Stage Verification Protocol for Annotated Data
Source Selection: Filtering literature based on explicit chemical synthesis protocols and verified peer-review status.
Linguistic Tokenization: Parsing text using domain-adapted chemical lexicons and dependency parsers.
Inter-Annotator Agreement: Conducting parallel reviews with a minimum 85% agreement threshold before corpus integration.
Provenance Documentation: Assigning permanent identifiers and publishing metadata records for reproducible scientific access.
Evaluation Criteria Rubric
Criterion
Verification Standard
Primary Evidence
Cross-checked against archival & institutional records
Terminology
Aligned with domain-specific technical nomenclature