Source

British Museum — Department of Egypt and Sudan

Egypt and Sudan · Predynastic through Islamic Egypt

1,644

Objects

Inscriptions

613

Images

3,316

Cross-references

Browse all 1,644 objects → Visit source →

Sample objects

Source details

Type: museum_collection
Access: wikidata_sparql_fallback
Stack: proprietary-cms-cf-gated
License: Mixed (BM site terms; Wikidata gateway data is CC0)
Modality: english_prose
ID system: BM registration number (EA prefix common) + Wikidata Q-ID
ID tier: tier-2
Inscription richness: 0/10
Cleanliness: 3/10

Form breakdown

jewelry: 1,235
vessel: 177
other: 141
tool: 56
statue: 25
architecture: 6
ostracon: 3
document: 1

Notes

## Discovered schema (British Museum via Wikidata fallback, surveyed 2026-05-23) ### HEADLINE FINDING: Direct BM access paths are dead The brief asked us to compare (a) collection.britishmuseum.org collection API and (b) ResearchSpace SPARQL. Both confirmed unreachable as of 2026-05-23: - **collection.britishmuseum.org** TIMES OUT at TCP on every path including /sparql. Related hosts (`ld.britishmuseum.org`, `data.britishmuseum.org`, `iiif.britishmuseum.org`, `sparql.britishmuseum.org`) all NXDOMAIN. The legacy linked-data infrastructure is decommissioned. - **researchspace.org/sparql** returns HTTP 404 on every common path. The site is alive but no longer hosts a BM SPARQL endpoint. - **www.britishmuseum.org** is fully behind Cloudflare Turnstile — both /collection/object/{ID} and /api/_search return the "Just a moment..." JS challenge for every UA we tried (custom, Mozilla, python-requests). Full probe outcomes stored as `_probe:*` samples on this source. ### Fallback: Wikidata SPARQL — the only working path Survey via Wikidata's SPARQL endpoint (https://query.wikidata.org/sparql), unioning six filters (EA-prefix inventory, location-of-discovery in Egypt, ditto Sudan, has TM text ID P8532, has TLA object ID P12185, has /collection/ permalink P973). ### Coverage reality (post-filter) - 57 unique Wikidata items returned by the union. - **13/57 are actually Egyptian** (judged by EA inv, TM/TLA cross-ref, Egyptian/Sudanese find spot, or Egypt/Pharaoh/Nubia in label). - The remaining 44 are noise — Wikidata's BM coverage isn't department-tagged, so the /collection/ permalink filter overshoots into Japanese woodcut prints, European engravings, Babylonian tablets, etc. - BM's actual Egyptian collection is ~100,000 objects. **Effective Wikidata coverage of BM Egyptian: ~0.02%.** Wildly insufficient for serious downstream use; usable only as a high-value cross-ref seed set. ### Likely-Egyptian sample (the comparable subset to MET, n=13) By auto-categorisation from P31 (instance of) + P186 (material): - statue n=5 - papyrus n=3 - other n=3 - architecture n=1 - stela n=1 ### Schema observed (Wikidata entity JSON; one record per Q-ID) Identity: id (Q-ID), labels.en, descriptions.en BM-side identifiers: P217 inventory number (multi-valued, e.g., 'EA24', 'EA10470,3', '1911,0901.1'). P973 described-at-URL (often the BM permalink at britishmuseum.org/collection/object/Y_{INV}). Type: P31 instance of (multi-valued — Rosetta Stone has FOUR: 'stele', 'bilingual inscription', 'Überrest', 'Epigraphic monument'). Reconfirms Flag 1. Material: P186 (granodiorite, limestone, papyrus, faience, ...) Provenance: P189 location of discovery, P575 time of discovery, P495 country of origin Date: P571 inception Dimensions: P2048 height, P2049 width, P2610 thickness (mm, structured) Images: P18 (Commons filename) — dereference via Special:FilePath/{filename} Cross-refs (when present): - **P8532 Trismegistos text ID** — direct link to TM - **P12185 Thesaurus Linguae Aegyptiae object ID** — direct link to TLA - P282 writing system (hieroglyphs Q132659, demotic Q188519, ...) - P407 language of work or name Other: P195 collection (always Q6373 for our filter), P276 location ### Four-flag comparison vs MET (source #3) MET (#3) BM-via-Wikidata (#4) total surveyed objects 77 57 (13 likely Egyptian) with inscription text 0/77 0/57 with TM cross-ref 0/77 **2/57** with TLA cross-ref 0/77 **1/57** with writing-system tag n/a 2/57 with image 77/77 48/57 collection coverage on path 100% (open API) <0.1% (Wikidata fallback only) licensing model per-object PD uniformly free via Commons; BM-side rights statements unknowable from gateway ### Flag 1 — Object vs inscription modeling (CONFIRMED, same as MET) Wikidata models inscriptions only by metadata (P282 writing system + P407 language) and never by content. 2/57 BM items have P282 set. No transcribed text anywhere in Wikidata's BM coverage. The decision to model `object` and `inscription` as separate tables is reinforced across both museum sources. ### Flag 2 — Transcription presence (CONFIRMED ABSENT, same as MET) Zero transcribed text. Museum-curated data (whether direct or via Wikidata) cannot be the inscription corpus. Same conclusion as MET. ### Flag 3 — Cross-ref status (NOTABLY BETTER WHERE WIKIDATA COVERS IT) 2/57 samples have a TM text ID via P8532; 1/57 have a TLA object ID via P12185. **For Wikidata-covered BM items, the museum→TM gap closes.** For the other ~99.98% of BM Egyptian items not on Wikidata, the gap is identical to MET (no cross-refs at all). Concrete examples in this sample: - Rosetta Stone EA24: TM 8809 + TLA FF4FQ4TTYBAE3L7VW22DCNJ3XI - Papyrus of Ani EA10470,3: TM 134357 (no TLA) These two pairs should seed the fuzzy-match pipeline as ground-truth (object_id, TM-Text) tuples when we build it. ### Flag 4 — Image licensing (DIFFERENT MODEL) Wikidata images come via Wikimedia Commons. Uniformly free reuse, but the underlying rights statement varies per file. 48/57 samples carry at least one image. License resolution requires a second hop to Commons per file. Practically: the BM-side rights (stricter than MET) are irrelevant for Wikidata-mediated images since those are already filter-passed for free reuse on Commons. ### Find spot vs current location P189 (location of discovery) vs P276 (location). Many items have P276 = Q6373 (the museum itself). The well-curated ones also have a finer P276 pointing to a gallery or storage location. ### Operational gotchas - Wikidata SPARQL requires a UA with product name + contact per https://meta.wikimedia.org/wiki/User-Agent_policy. The `python-requests/...` UA used elsewhere is REJECTED. Use `egyptian-inscriptions-survey/0.1 (mailto:contact)`. - Wikidata's wbgetentities API caps at 50 IDs per call — bulk label lookups must chunk. - Complex UNIONs over property paths (`?item wdt:P189/wdt:P17 wd:Q79`) HTTP 504 (Gateway Timeout) on Wikidata's public endpoint. Split into separate small queries and union in code. - The /collection/ permalink filter overshoots — BM items on Wikidata aren't department-tagged in a way SPARQL can filter on. Post-hoc filter on inv prefix. ### Open questions - BM may publish bulk dumps via data.gov.uk, AHRC repositories, or Zenodo. Worth a one-off search before declaring this gap permanent. - Brooklyn Museum (#5) is next. Their open API has been alive in the past and they publish Egyptian object data without Cloudflare gating. Confirm whether the museum-API pattern works for Brooklyn before treating MET as the high-water mark. - Coordinate with Wikidata Cultural Heritage WikiProject — the 2-item BM curation suggests someone has been actively adding TM cross-refs and we should meet them.

British Museum — Department of Egypt and Sudan

Sample objects

amphora

amphora

amphora

model

altar

ostracon (?)

ostracon (?)

amphora

amphora

amphora

amphora

ostracon (?)