1 of 10

EAP1721 · ENDANGERED ARCHIVES PROGRAMME

From Epigraphic Data to Archival Regime

Armenian Inscriptions in India and the EAP1721 Digital Corpus

Hamest Tamrazyan

École polytechnique fédérale de Lausanne (EPFL)

Gayane Hovhannisyan

Yerevan State University (YSU)

Gravestone of Anna Cachkhatoon, Armenian Church of the Holy Nazareth, Kolkata

2 of 10

RESEARCH PROBLEM

How should inheriting institutions read a digital corpus?

Research question

How is colonial-era diaspora memory reorganised when it enters contemporary metadata infrastructures?

Thesis

EAP1721 is both a salvage record of endangered minority heritage and an archival regime shaped by standards, vocabularies, depositories and metadata fields.

“Colonial archive” is used analytically, not accusatorily: it names the representational effects of standards and custodial structures, not the intent of the project.

128

file-level records with images and bilingual descriptions

4

cities: Kolkata, Madras, Hyderabad, Mumbai

1641–1917

date range of the 118 dated inscriptions

3

digital depositories beyond the British Library EAP

2 / 10

3 of 10

THEORY → MODEL

Empirical archive versus archival regime

Empirical archive

What researchers consult

  • Inscriptions and photographs
  • Transcriptions
  • Descriptive metadata records

Archival regime

Through what the corpus circulates

  • Standards: ISAD(G), LC subject headings
  • Controlled country and language fields
  • Depository networks, access and IP rules

“What remains tacit in a paper archive becomes declarative in a digital one.” The cost: ideological choices can appear as technical defaults. The gain: once encoded, they can be measured and revised.

Grounding

Stoler 2002

Archives as technologies of governance and classification

Mbembe 2002

Archives regulate what can be preserved and spoken

Derrida 1996

Consignation: gathering, ordering, authorising

Bowker & Star 2000

Classification as socio-technical infrastructure

Hamilton et al. 2002

Decolonising ≠ relocating documents

3 / 10

4 of 10

LITERATURE → CORPUS

Prior documentation and the place of EAP1721

1897 / 1937

M. J. Seth

History of the Armenians in India; Armenians in India. Funerary texts read as biographical witnesses of merchant families.

Parish work

Community records

Selective transcriptions by parishes and the Armenian Apostolic Church; no schema-validated cross-city database.

2011

S. Aslanian

New Julfa as a transregional trade network: Isfahan, the Mediterranean, the Indian Ocean.

2024

R. Kortoshyan

Most extensive recent printed corpus; indispensable for philological and historical work.

2026

EAP1721

EAP-standardised metadata corpus for preservation, search, comparison and infrastructure integration.

Methodological asymmetry. Merchant correspondence in European archives is catalogued and digitised; the inscriptions remain physically dispersed across Indian churches and cemeteries. EAP1721 complements printed epigraphy rather than replacing it.

4 / 10

5 of 10

EVIDENCE BASE

The pilot corpus: 128 inscriptions, four custodial sites

Records by city

Kolkata

1697–1917

75

Armenian Church of the Holy Nazareth

Madras

1706–1830

20

Armenian Church of St Mary

Hyderabad

1641–1869

18

Armenian Cemetery

Mumbai

1798–1900

15

Armenian Cemetery, Antop Hill

Two clusters: the late-Safavid New Julfa world (1670s–1790s) and the larger nineteenth-century colonial moment (peak 1840s–1860s). Ten undated records excluded; decades with no records shown as zero.

5 / 10

6 of 10

MODEL → METHOD

Method: a directed multi-layer network from the metadata

334

nodes

1,957

typed edges

9

node types

18

edge types

1

Workbook fields

Titles, Related Towns, Country of Origin, Locations of Original and Digital Copies, Gregorian date

2

Deterministic parsing

Regular expressions on English titles: “Gravestone of X, daughter of Y, wife of Z”

3

Typed edges

Multiple relations in one title yield multiple edges; weights record co-occurrence

4

Diagnostic graph

Bridges, singletons and dominant categories become properties of the schema itself

Only 2 subject and 2 language nodes: the thinness of the vocabulary is visible in the graph’s own shape.

6 / 10

7 of 10

IMPLEMENTATION

One record, one subgraph

Holy Nazareth, Kolkata

“Gravestone of Anna Cachkhatoon, wife of Owen Moses, daughter of Mariam Moses and granddaughter of Avetis Apgar”

commemorates

Inscription

Anna Cachkhatoon

Owen Moses

wife_of

Mariam Moses

daughter_of

Avetis Apgar

granddaughter_of

Kolkata

located_in_city

1900s

dated_to

Also: written_in → Armenian, English · held_at → custodial site · has_subject

7 / 10

8 of 10

EVIDENCE

Five diagnostics: how categories shape visibility

Geography

Isfahan and Julfa dominate related-town nodes, yet the country field links all 128 records to “Armenia”.

Anachronism for 17th-c. New Julfa merchants

Kinship

26 person-to-person edges, 5 relation types. Men via descent; women via fathers, husbands, widowhood.

Epigraphic convention silently inherited

Custodianship

7 institutional nodes, 3 digital depositories. Descendant communities appear only in acknowledgements.

Robust but socially narrow

Vocabulary

Two subject terms (“Burials”, “Diaspora”) and two languages for a multi-genre, Persianate corpus.

Genres and naming practices unstructured

Time

One Gregorian date field; production regimes (Safavid, Mughal, Deccan, EIC, Raj) are not expressible.

Chronology without history

8 / 10

9 of 10

IMPLICATIONS FOR LIBRARY PRACTICE

From diagnosis to practice: four extension layers

Current schema

Proposed extension layer

Analytical gain

Modern country field

Temporally indexed historical geography: Safavid Iran, Mughal Empire, Persianate Deccan, EIC India, British India

Provenance without nation-state anachronism

Kinship in titles / free text

Structured kinship fields, extended to Armenian titles and transcriptions

Family networks, gendered patterns, prosopography

Institutional custodians only

Community-partner register with documented governance roles

Consent and local knowledge recorded as data

Two LC subject terms

Project vocabulary over LC headings: epitaphs, memorial and donor plaques, bilingual, damaged or illegible

Genre-level retrieval with interoperability intact

No redesign of the EAP schema. A declared, backward-compatible extension layer: modest technical effort, substantial conceptual gain — the archive describes the conditions under which inscriptions become data.

9 / 10

10 of 10

CONCLUSION

Two durations

Lithic duration

Stone survives across Safavid, Mughal, Company and Crown regimes.

Infrastructural duration

Metadata can outlive the stone, but inherits the political shape of its own time.

Schemas, vocabularies and custodial networks are not packaging around the archive; they are part of the archive itself. The network makes the regime measurable — and therefore revisable.

Next phase, 2026–2029

  • Historical-geography thesaurus
  • Structured kinship fields
  • Community-partner register
  • Linked Open Data federation with merchant letters, New Julfa parish registers, diocesan and British Indian records
  • Communities as co-curators of memory

EAP1721 project: eap.bl.uk/project/EAP1721 · Network: hamesttamrazyan.github.io/eap1721-network