Extraction matrix and PRISMA record for a 96-study systematic review of cognitive computing in construction contract and claims management

Data appendix for the review article Cognitive Computing in Construction Contract and Claims Management: A Systematic Review from Rule-Based Parsing to Neurosymbolic GraphRAG. The workbook contains: (1) the 96 included studies with an 18-field extraction matrix (venue, SJR 2024 quartile, thematic pillar, technical wave, architecture, dataset scale, metrics, limitations, DOI); (2) the PRISMA 2020 selection arithmetic (2,720 identified, 650 duplicates removed, 2,070 screened, 1,615 excluded at title and abstract, 455 full texts assessed, 359 excluded across four eligibility gates); (3) the full-text exclusion counts by gate (142 / 115 / 73 / 29); (4) the cross-pillar benchmark table underlying Table 8; (5) the quality-appraisal sheet with item-score template; and (6) an audit notes sheet recording every correction made to the metadata, quartile labels and quality scores. Quartiles follow the SJR 2024 release, best-quartile rule. Contract form is named in 20 of the 96 studies; 24 study no contract corpus; 52 report no governing form. Per-study quality scores are left blank pending the original item-level scores; the aggregate in Table 7 of the article is the authoritative record.