Reimagining the Archive in the Post-Truth Era · Rhodes University, Makhanda · 1 July 2026

Reconfiguring the Archive

Relationality and discovery in the Africa Multiple Interactive Research Atlas (AMIRA)

Frédérick Madore Data Curator Africa Multiple Cluster of Excellence

University of Bayreuth · German Excellence Strategy · since 2019

The Africa Multiple Cluster of Excellence

A large African studies cluster, reconfiguring the field both conceptually and structurally to confront power imbalances in how knowledge is produced and transmitted.

Three working concepts
  • Multiplicity: relationships before fixed entities
  • Relationality: worlds made, unmade, silenced
  • Reflexivity: our own positions, and the power behind them
In figures
300+Researchers, cluster-wide
5Research Centres, five countries
2026–32Second phase, under way
Known within the cluster as "Reconfiguring African Studies" — reworking the field's concepts and its structures at once.

About us · Africa Multiple Research Centres (AMRCs)

A network of five centres

AMRCs
  • Université Joseph Ki-Zerbo (Burkina Faso)
  • University of Bayreuth (Germany)
  • Moi University (Kenya)
  • University of Lagos (Nigeria)
  • Rhodes University (South Africa)

Shown in green on the map. africamultiple…/centres

Digital Research Environment (DRE) — the cluster's digital infrastructure unit

We build the cluster's data infrastructure

Four areas of work — one goal: research data that is findable, reusable and presentable.

Data management & curation

Metadata and ontologies, Wikidata reconciliation, African-language taxonomies, DOIs and archiving.

Data science & methods

Text mining, NLP, network analysis, web scraping.

AI-assisted document processing

OCR, transcription, entity extraction, tagging — on local, open-source, GDPR-compliant models.

Dissemination & presentation

Dashboards, maps, network graphs, Omeka S databases, websites.

Grounded in FAIR & CARE and multiple knowledge systems. Full range: africamultiple…/digital-solutions

DRE · how we work

Two principles guide how we work

Description is a joint effort

Curation and description are shared with our AMRC partners. Each centre describes the data it knows best; the DRE runs the shared infrastructure that connects them.

Storage stays distributed

Data stays in its local repository; Bayreuth holds the metadata layer that points to it. Research data becomes findable without being relocated.

Not an extractive model A considered response to debates on data colonialism and extractivism: the field's material is not drawn from Africa onto a server in Germany.

The first phase

We began with WissKI

A semantic research-data system built on knowledge graphs — entities and their relations, tied to shared ontologies.

WissKI — Wissenschaftliche Kommunikationsinfrastruktur, "scientific communication infrastructure".

WissKI@UBT — searching and exploring the cluster's research data through a web interface.
Searching and exploring research data across the centres.
A WissKI knowledge graph — entities connected to each other and to shared ontologies through typed relations.
Records as a knowledge graph: entities linked by typed relations.

The new platform · replacing WissKI@UBT

Enter AMIRA

The Africa Multiple Interactive Research Atlas — the Cluster's new home for its research data, publications and outputs.

Browse

Items, people, projects, publications and media — with type-ahead autocomplete and faceted filters.

Visualise

Maps, network graphs and charts across the whole collection — not one record at a time.

Cross-linked

Every person, project, subject and place is its own page you can open.

Now live Built on Omeka S

QR code linking to the AMIRA website Scan to explore AMIRAdata.africamultiple.uni-bayreuth.de

Africa Multiple Interactive Research Atlas

data.africamultiple.uni-bayreuth.de Open AMIRA ↗

International Library of African Music (ILAM)

A collection from Rhodes, in AMIRA

data.africamultiple.uni-bayreuth.de/s/amira/page/international-library-of-african-music-ilam Open the collection ↗

An external collection from ILAM at Rhodes — its catalogue is described in AMIRA

Facets, filters, autocomplete — built on Typesense

data.africamultiple.uni-bayreuth.de/s/amira/dre-search Open search ↗

Here an entity is the sum of its relationships, rather than a standalone record.

Relationality, one of the cluster's three concepts, treats phenomena as formed through their relationships — and attends to the worlds those relationships leave unmade or silenced.

Relationality, made visible

Connections, not records

data.africamultiple.uni-bayreuth.de/s/amira/item/293 Open record ↗

Open a record — here, Prof. Ute Fendler — and AMIRA lists every resource linked to it, each labelled with the nature of the connection.

71 linked resources on this record

Projects, publications, co-authors, places and subjects — each its own page, and each link naming the relationship.

Custom Omeka S module · built in-house

Spatial exploration

One of the DRE visualisation modules I build — every georeferenced place in the collection, on one map you can pan, zoom and open.

data.africamultiple.uni-bayreuth.de/s/amira/page/spatial-exploration Open the map ↗

Custom Omeka S module · knowledge-graph links

Entities and their links

data.africamultiple.uni-bayreuth.de/s/amira/page/networks Open the network ↗

The whole collection as one network — entities such as people, projects, subjects and places, drawn together by the links recorded between them in the knowledge graph.

The catch, already visible Every link here was described by someone. The network is only ever as good as that description — and subject description is the thinnest of it.

An atlas of relations is only as strong as its weakest metadata.

The weakest part is almost always subject description — the part that researchers supply the least of all.

When terms do exist

Whose subject headings?

Where subject terms exist, they tend to come from Library of Congress Subject Headings.

  • Our pipeline matches subjects against the LCSH API
  • LCSH was built elsewhere, for other collections
  • African concepts and languages fit it poorly
  • Borrowed terms reshape what can be found
The vocabulary problem Two people pick the same term for the same thing under a fifth of the time (Furnas et al., 1987). One imposed vocabulary makes that worse, not better.

Risam (2018): standard schemas can carry epistemic violence — categories that erase as they describe.

Still exploratory

Building the entities the vocabulary lacks

Wikidata, not only Library of Congress

Open, multilingual, editable by communities rather than one national library.

  • Reconcile people, places, institutions to Wikidata
  • Create the African entities it still lacks
AI to help, not to decide

Open-source local LLMs draft candidates at scale; people review and approve.

  • Draft labels, definitions, links for review
  • Curators and centres keep the final say
Exploratory by design To move authority over description closer to the people the data is about.

Thank you

Who has the authority to decide how knowledge is described, and therefore how it is found?

In an era of contested truth, this is the challenge that reimagining the archive must face.

Explore AMIRA: data.africamultiple.uni-bayreuth.de/s/amira