Compares the relationships people asserted in the original Linked Jazz project
(base.semlab.io Wikibase: 52nd Street crowdsourcing + supervised name tagging)
with the LLM/audit relationships in ~/git/ch-jazz-mashup/linked_jazz.sqlite,
on the 46 transcripts both corpora share (Hamilton College, Smithsonian, Rutgers).
REPORT.md— findings.site/index.html— explorer (open directly in a browser; data embedded insite/data.js). Published at https://semanticlab.github.io/linked-jazz-2026-qa/ via GitHub Pages (.github/workflows/pages.ymldeployssite/on every push tomain).
# 1. transcript list + property labels (SPARQL; small queries only, bulk ones 524 behind Cloudflare)
curl -sG -A semlab-data-process "https://query.semlab.io/proxy/wdqs/bigdata/namespace/wdq/sparql" -H "Accept: text/csv" --data-urlencode 'query=...' > semlab_transcripts.csv
# 2. SQLite interviewees of the three archives (sqlite_interviewees_rhs.txt), then:
uv run python harvest_semlab.py # wbgetentities API, cached in raw/entities/ -> data/semlab_harvest.json
uv run python harvest_reverse.py # statements on object items (found none)
uv run python extract_sqlite.py # -> data/sqlite_side.json
uv run python compare.py # -> data/comparison.json
uv run python build_site.py # -> site/data.js
uv run python write_report.py # -> REPORT.mdAll requests to base.semlab.io need User-Agent: semlab-data-process (Cloudflare rule).
- transcript item:
P1instance ofQ1960(interview transcript),P13interviewee,P14interviewer,P15source (Q2031 Hamilton College, Q2033 Smithsonian, Q2034 Rutgers, Q2032 U. Michigan, Q21579 Weeksville),P16legacy MD5,P7slug,P98date. - block item:
P1=Q2013,P24parent transcript,P17local id,P19text,P20text URL (semlab.s3.amazonaws.com/texts/<md5>/<n>.txt),P23speaker,P21associated entities. - relationship statements live on the interviewee's person item (
P29knows of …P39influenced by), qualified byP25generation method (Q18800supervised NER,Q18801crowdsourced 52nd St) andP40user consensus, referenced byP26block items.