# Islam West Africa Collection

- **Years:** Since 2023
- **Methods:** Omeka S, PHP, Python, TypeScript, Svelte 5, Typesense, Semantic Search, ECharts, D3.js, MapLibre GL, OCR, OpenRefine, Metadata Standards, Linked Data, Wikidata, WikiCite, SPARQL, IIIF, Docker, Linux, Nginx, Rclone, spaCy, pandas, Hugging Face, Embeddings, Topic Modelling, Web Scraping
- **Web page:** https://www.frederickmadore.com/digital-humanities/iwac

*An open-access archive of nearly 21,000 newspapers, Islamic publications, recordings and documents on Islam and Muslims in Benin, Burkina Faso, Côte d'Ivoire, Niger, Nigeria and Togo.*

How have Muslim communities in West Africa written about themselves, and how has the press written about them? The *Islam West Africa Collection* (IWAC) gathers the sources that answer that question: newspapers, Islamic magazines and pamphlets, archival documents, audiovisual recordings and photographs from Benin, Burkina Faso, Côte d'Ivoire, Niger, Nigeria and Togo, from the 1960s to today.

Much of this print culture is fragile and hard to consult, even in the countries that produced it. IWAC puts it online without paywalls or registration. It now holds nearly 21,000 items from 86 newspapers and 24 archives, in ten languages, most of them in French. More than 32 million words are searchable in full, an index of over 5,000 people, places, organisations, events and topics ties the materials together, and a bibliography of 867 references points to the wider scholarship.

I direct and develop the collection, which is hosted by the [Leibniz-Zentrum Moderner Orient (ZMO)](https://www.zmo.de/en) in Berlin. It grew out of the award-winning [Islam Burkina Faso Collection](https://www.frederickmadore.com/digital-humanities/islam-burkina-faso-collection.md) I launched in 2021.

## Finding your way in

The search is built to forgive. It ignores accents, so "Cote d'Ivoire" finds the same documents as "Côte d'Ivoire", and it matches meaning as well as words: a search for "laïcité au Burkina" surfaces relevant articles even when those exact words never appear. Results filter by country, newspaper, person, place, subject and year, can be shown on a map, and keep their own web address, so any search can be bookmarked, shared or cited.

## Seeing patterns

Interactive dashboards turn the archive into something you can read at a distance:

- **Overviews** of how the collection has grown, which countries and languages it covers, and which newspapers were published when
- **Topics and words over time**: thirty themes traced across the decades, the vocabulary that sets each newspaper apart, and a calendar that follows the Islamic months
- **Press attitudes**: how several AI models read the tone of coverage of Islam, compared side by side in a [dedicated study](https://www.frederickmadore.com/digital-humanities/iwac-sentiment-analysis.md)
- **Networks and maps** of the people, organisations and places written about together
- **A page of charts for every record**: when and where a person or place appears, what it connects to, the closest articles, and the academic works that discuss the same subject

## Ask with AI

Researchers can also explore IWAC by putting questions to an AI assistant they already use, in any language. The [IWAC MCP server](https://www.frederickmadore.com/digital-humanities/iwac-mcp-server.md) gives the assistant read-only access to the collection, and every answer links back to the records it draws on, so you can check the sources yourself. The collection's [Ask with AI page](https://islam.zmo.de/s/westafrica/page/ai-access) explains how to connect.

## How it's built

IWAC runs on [Omeka S](https://omeka.org/s/), an open-source platform for digital collections, self-hosted at ZMO. I wrote the custom parts, all released as open source:

- **[IWAC-theme](https://github.com/fmadore/IWAC-theme)**: the bilingual French and English interface, with light and dark modes
- **[IwacSearch](https://github.com/fmadore/IwacSearch)**: the search, powered by Typesense
- **[IwacVisualizations](https://github.com/fmadore/IwacVisualizations)**: the charts, maps and networks
- **[IWAC-SEO](https://github.com/fmadore/IWAC-SEO)**: the metadata that helps search engines find each record and lets Zotero or Google Scholar cite it properly
- **[iwac-vocabulary](https://github.com/fmadore/iwac-vocabulary)**: a small vocabulary that records which AI model produced each transcription, summary or reading of tone

Much of the transcription, indexing and summarising runs through [AI pipelines](https://www.frederickmadore.com/digital-humanities/iwac-ai-pipelines.md), and their results are written back into the archive alongside the name of the model that produced them.

## The collection as a dataset

Omeka S is built for browsing a collection, not for computing over one. A Python pipeline therefore republishes IWAC on [Hugging Face](https://huggingface.co/datasets/fmadore/islam-west-africa-collection) as seven citable tables, each loadable in a single line of code. They also carry what the archive itself does not: embeddings that capture meaning, topics, word counts, Islamic-calendar dates and the AI models' readings of tone. The dashboards read these same columns, so a figure on the site and a figure computed by an outside researcher rest on the same numbers.

Rights vary item by item, so the public dataset withholds the full text wherever the source is restricted. That leaves the text of a little under half the newspaper articles and most Islamic publications, while the computed columns are released for every public item. What is open tends to follow publisher and period, so counts from the dataset describe what can be shared rather than everything the collection holds.

## The bibliography on Wikidata

Since 2023, in the spirit of open science and the [WikiCite](https://meta.wikimedia.org/wiki/WikiCite) initiative, I have been publishing the collection's bibliography on [Wikidata](https://www.wikidata.org/), the free knowledge base behind Wikipedia. Of its 867 references, 846 now have a Wikidata item, and each IWAC record links to it.

There the references join a shared web of linked data, in which their authors, journals, publishers and subjects connect to everything else in the knowledge base. The bibliography can be found and reused beyond the archive, and asked questions a printed list cannot answer: who writes about Islam in Niger, with whom, and in which journals. A [ready-made query](https://qlever.dev/wikidata/?query=%23%20Works%20on%20Islam%20in%20the%20six%20countries%20of%20the%20Islam%20West%20Africa%20Collection%0APREFIX%20wd%3A%20%3Chttp%3A%2F%2Fwww.wikidata.org%2Fentity%2F%3E%0APREFIX%20wdt%3A%20%3Chttp%3A%2F%2Fwww.wikidata.org%2Fprop%2Fdirect%2F%3E%0APREFIX%20rdfs%3A%20%3Chttp%3A%2F%2Fwww.w3.org%2F2000%2F01%2Frdf-schema%23%3E%0ASELECT%20%3FtopicLabel%20%3Fwork%20%3Ftitle%20%3Fyear%20WHERE%20%7B%0A%20%20VALUES%20%3Ftopic%20%7B%20wd%3AQ3155361%20wd%3AQ3081878%20wd%3AQ4119820%20wd%3AQ3087639%20wd%3AQ2637700%20wd%3AQ522809%20%7D%0A%20%20%3Fwork%20wdt%3AP921%20%3Ftopic%20.%0A%20%20%3Ftopic%20rdfs%3Alabel%20%3FtopicLabel%20FILTER%28LANG%28%3FtopicLabel%29%20%3D%20%22en%22%29%0A%20%20OPTIONAL%20%7B%20%3Fwork%20wdt%3AP1476%20%3Ftitle%20%7D%0A%20%20OPTIONAL%20%7B%20%3Fwork%20wdt%3AP577%20%3Fdate%20BIND%28YEAR%28%3Fdate%29%20AS%20%3Fyear%29%20%7D%0A%7D%0AORDER%20BY%20%3FtopicLabel%20DESC%28%3Fyear%29%0A&exec=true) lists the works Wikidata files under Islam in each of the collection's six countries, and can be edited to ask others.

## Addresses

- [islam.zmo.de/s/westafrica](https://islam.zmo.de/s/westafrica)
- [islam.zmo.de/s/afrique_ouest](https://islam.zmo.de/s/afrique_ouest)
- [IwacSearch](https://github.com/fmadore/IwacSearch)
- [IwacVisualizations](https://github.com/fmadore/IwacVisualizations)
- [IWAC-SEO](https://github.com/fmadore/IWAC-SEO)
- [IWAC-theme](https://github.com/fmadore/IWAC-theme)
- [iwac-vocabulary](https://github.com/fmadore/iwac-vocabulary)
- [IWAC-Hugging-Face](https://github.com/fmadore/IWAC-Hugging-Face)
- [huggingface.co/datasets/fmadore/islam-west-africa-collection](https://huggingface.co/datasets/fmadore/islam-west-africa-collection)
- [Wikidata query](https://qlever.dev/wikidata/?query=%23%20Works%20on%20Islam%20in%20the%20six%20countries%20of%20the%20Islam%20West%20Africa%20Collection%0APREFIX%20wd%3A%20%3Chttp%3A%2F%2Fwww.wikidata.org%2Fentity%2F%3E%0APREFIX%20wdt%3A%20%3Chttp%3A%2F%2Fwww.wikidata.org%2Fprop%2Fdirect%2F%3E%0APREFIX%20rdfs%3A%20%3Chttp%3A%2F%2Fwww.w3.org%2F2000%2F01%2Frdf-schema%23%3E%0ASELECT%20%3FtopicLabel%20%3Fwork%20%3Ftitle%20%3Fyear%20WHERE%20%7B%0A%20%20VALUES%20%3Ftopic%20%7B%20wd%3AQ3155361%20wd%3AQ3081878%20wd%3AQ4119820%20wd%3AQ3087639%20wd%3AQ2637700%20wd%3AQ522809%20%7D%0A%20%20%3Fwork%20wdt%3AP921%20%3Ftopic%20.%0A%20%20%3Ftopic%20rdfs%3Alabel%20%3FtopicLabel%20FILTER%28LANG%28%3FtopicLabel%29%20%3D%20%22en%22%29%0A%20%20OPTIONAL%20%7B%20%3Fwork%20wdt%3AP1476%20%3Ftitle%20%7D%0A%20%20OPTIONAL%20%7B%20%3Fwork%20wdt%3AP577%20%3Fdate%20BIND%28YEAR%28%3Fdate%29%20AS%20%3Fyear%29%20%7D%0A%7D%0AORDER%20BY%20%3FtopicLabel%20DESC%28%3Fyear%29%0A&exec=true)

## Reviews

- [Anaïs Wion, "Islam West Africa Collection: Dataset, Distant Reading, and Uses of AI for Discourse Analysis", *The Digital Orientalist* 23 September 2025.](https://digitalorientalist.com/2025/09/23/islam-west-africa-collection-dataset-distant-reading-and-uses-of-ai-for-discourse-analysis/)

  > On the one hand, it is a database providing access to material that is difficult to access and highly specific, which can be reused by others and addressed with other research questions. But it is also an exemplary project that can inspire researchers to use the open access tools available to create their own platforms for analyzing and sharing their data. The care taken to comply with open access best practices is truly admirable.
