Reprexbase | Daniel Antal

Help Us Build a Truly Inclusive European Music Observatory

Thu, 19 Jun 2025 18:45:00 +0200

Across Europe, music libraries are under pressure: greater expectations for digital services, growing metadata burdens, and increasingly fragmented infrastructure. At the same time, vital parts of our musical heritage—especially regional or minority repertoires—remain hidden from search, discovery, and policy.

Please meet us at 👉 IAML2025 in Salzburg on 7 or 8th July. Our presentation takes place in the session of Music Libraries of Tomorrow: Reaching out to Wider Audiences at the Mozarteum University E.001 HS Thomas Bernhard room on 7 July 2025, 16:00–17:30. The day after you can meet us in the Gallery for the poster session.

We initiated the Open Music Europe project, because we believe that in the music ecosystem, data centralisation always fails, and a new kind of cooperation is needed—one that respects local control while enabling international reuse.

Our Slovak pilot, the SKCMDb, connects libraries, music centres, rights organisations, and platforms through a shared metadata backbone based on open ontologies. Built as a national data sharing space, it enables coordinated cataloguing and discovery across public and private systems—from streaming services and printed scores to CD loans and digital archives.

Please visit our poster and talk with our team members, Daniel Antal, Anna Márta Mester (librarian-data steward) and Anna Zilkova (chairperson of IAML Slovakia) on 8 July 2025 10:30–11:00 in the Gallery. You can download our poster in PDF here.

But we also know that cultural and music policy is not only national. It is often regional, local, or community-based. That’s why we follow the principle of subsidiarity: letting decisions and innovation happen at the lowest competent level, close to the collections and communities themselves.

Our Finno-Ugric Data Sharing Space, including the LīvMDb (Livonian Music Database), shows how even the smallest communities—without formal cultural infrastructure—can take part in high-quality metadata production and digital discovery. We provide the tools and models to empower local custodians, in their language, on their terms, and without the need for large institutional support.

Please check out the demo version of the Finno-Ugric Dataspace or read the long-form project description.

Now we invite IAML members—national libraries, regional centres, municipal collections, and independent music librarians—to join us in building a federated, decentralised European Music Observatory. One that reflects Europe’s diversity. One that reduces data curation costs and improves visibility. One that connects music libraries with the open data and open science infrastructures already transforming other sectors.

Our platform is open-source, built on FAIR principles and the European Interoperability Framework. We use tools like Wikibase, Blazegraph, Sampo-UI, and R—packaged to work for libraries with limited technical capacity.

Sneak peak: http://135.181.91.51:3007/en/

If you care about interoperability, cultural equity, and the future of library relevance in the streaming era—this is your moment to get involved.

Not present at IAML2025?
👉 Presentation
👉 Poster
👉 Please contact us directly.

Let’s ensure music libraries remain vital entry points to Europe’s rich and evolving cultural soundscape.

Metadata Groundhog Day: What a Moribound Language Can Teach Spotify and Shopify

Thu, 19 Jun 2025 18:45:00 +0200

And if you want to fix these errors, you may find that you are back to the Data Sisyphus.

When you build systems in the cloud, or in your local architecture, at one point you will realise that naming things — places, people, products — or updating their whereabouts is probably the most time-consuming, most expensive, and most error-prone workflow.

In this blogpost, we want to talk about what seems like the easiest part of a location: the name of the city, town, or village.

Mazirbe Is Missing Again

We recently built a multilingual gazetteer — essentially a reconciled database of place names — for a tiny stretch of the Livonian coast in Latvia. At first glance, this might seem like a project rooted deeply in the digital humanities.

But here’s the twist: the very same problems we tackled here are the ones plaguing the music industry, global e-commerce platforms, and enterprise software stacks.

This is Gross-Irben 👉 Lielirbe / Īra / Irben / Suur-Irben 👉 Familiar with RDF: see in TTL; Klein-Irben will be near Gross-Irben, and Irē is almost Īra!

Mazirbe is a small, big place. It definitely exists, and it is the cultural center of a small nation: the Livonians. Yet, when you are looking for clothing, music, or photographs that should come from Mazirbe in a relevant database, you often find nothing. Not even the place.

But Mazirbe exists!

Depending on the record, it might appear as:

Mazirbe (Latvian) • Irē (Livonian) • Мазирбе (Russian) • Klein-Irben (German) • Suur-Irben (Finnish-German hybrid) • Мазирбе (Russian) • Mazirbė (Lithuanian)

Meyer‘s Zeitungsatlas 050 – Russland- Gouvernement Sankt Petersburg, Esthland, Liefland, Kurland

This kind of variation isn’t just a cultural footnote — it breaks databases, mismatches search results, and silently corrupts analytics.

If you’re in music metadata, this is your “JAY Z” vs. “Jay-Z” vs. “Shawn Carter” problem.

If you’re in e-commerce, it’s “Red Crewneck XXL” vs. “Crewneck, crimson, 2XL”.

Same data structure. Same unresolved chaos.

A Gazetteer That Works Like Real Life

We created a semantic, multilingual, multiscript gazetteer for the Livonian coast. Each place entry includes:

All known name variants across time, languages, and scripts
Structured links to global authority services (Wikidata, VIAF, GeoNames)
Canonical IDs, multilingual labels, and machine-readable formats (RDF, TTL, etc.)
Context about administrative boundaries, historical changes, and source provenance

Try us:

👉 Mazirbe / Irē / Klein-Irben / Мазирбе / Mazirbė 👉 Familiar with RDF: see in TTL

We published it using Wikibase — the same technology that powers Wikidata. It’s not just a spreadsheet; it’s a small, dynamic knowledge graph.

And we also put it into BlazeGraph, so you can find all these villages — and also the music, the clothing, or photographs that come from them.

So What?

Here’s why this matters outside the northern shores of Kurzeme, or beyond the borders of Latvia:

In global supply chains, location names and vendor names drift constantly. While country boundaries are relatively stable, subnational boundary changes — counties, parishes, provinces, municipal borders — happen thousands of times per year, even within Europe.
In streaming metadata, artists get duplicated, misspelled, or transliterated inconsistently. It’s not unusual to find dozens of same-named artists in a distributor’s or rights manager’s roster.
In CRM systems, customers have multiple entries because of one diacritic. Irē becomes Ire if the user didn’t have ē installed.
In museum heritage databases and webshops, items disappear because their place of origin changed names three times since the accession record was created.

Our little example was created to accompany a digital humanities publication, but it’s not just a “humanities” problem. It’s a cross-sector, multilingual, historical, bureaucratic, data problem.

And we’re all living in it.

Lessons We Took Away

Don’t fight ambiguity. Model it.
Linked data models (RDF, Wikibase) handle aliases and variants with elegance.
Small, local, curated vocabularies can scale conceptually to global systems.
Top-down standardization fails in diverse data ecosystems — context wins.

See It / Fork It / Repurpose It

You can explore the full Livonian Gazetteer here:

Web UI: https://reprexbase.eu/fu/Main_Page
RDF example: https://reprexbase.eu/fu/Special:EntityData/Q4429.ttl
Also check out our TextileBase project — same model, but for 19th-century Latvian shirts and skirts

If your stack includes messy location names, user-generated labels, non-English content, or legacy records — maybe this can help.

And if you feel like you’ve seen this movie before… you have.

It’s Data Sisyphus all over again.

👉 https://reprex.nl/post/2021-07-08-data-sisyphus/

Linked Open Datasets on Garments from the Latgale Region

Mon, 07 Apr 2025 17:00:00 +0100

Entry examples from Linked Open Datasets on Garments from the Latgale Region, from right to left: Q142, Q180, Q179, Q181.

The first published dataset, Linked Open Datasets on Garments from the Latgale Region contains data on Latvian traditional shirts and skirts from the Latgale region in Eastern Latvia. The The female and male shirts, and the skirts in the dataset are handmade and were worn in the 19th century. They represent both festive and daily wear of the local female and male peasants. The shirts are stored at the National History Museum of Latvia and the Ethnographic Open-Air Museum of Latvia. The data contain information on the locality of their origin, their approximate date of creation with various precisions, the materials they are made of, and the way of their fabrication, as well as their purpose of wearing (festive or daily wear) and wearer’s ethnicity and gender. They also include the name of the museum each shirt is stored at, supplemented with its unique inventory number. Data on some sample shirts also include a photo of the shirt.

Check out the properties (relations) in the data model: ListProperties
See every entry in the database: All items

Dataweek²⁴: Data-driven Compliance with the Corporate Social Responsibility Directive

Thu, 06 Jun 2024 16:22:00 +0100

All slides

This is a lightly edited version of the presentation at the Data-driven and Automated Compliance section of Dataweek²⁴ Event page.

It has often been said that ESG reporting is a data problem. If you want to fulfil the new requirements set by the Corporate Social Responsibility Directive, which is a legal act that changes European laws on financial accounting and its audit or assurance, you will encounter a very serious data linking and integration problem. If you have such sustainability bookkeeping, you must be able to factually link your financial accounts to your environmental and social accounts. In simple terms, if you expense the cost of 1 MW of electricity, then you cannot calculate the footprint of 0.98 MW in your sustainability report.

All slides

ESG introduces two related challenges, which are best addressed by explicit knowledge bases, enterprise graphs connected to open knowledge graphs, and data sharing spaces.

The company must be able to collect data and report on other companies and things that happen outside of the boundaries of the company or company group. Such a practice had been in place in some heavily regulated industries; for example, the food industry had to bear a qualified responsibility over the entire food supply chain or the nuclear industry for the lifecycle of the fossile fuels.
The company must also be able to curate trustworthy data about entirely new domains: data about the natural or ecological environment, such as data on biodiversity, recycling, water; and data about the social environment, for example, indicators about affected communities, workers in the supply chain or end users.

Joining a data sharing space is a good solution because the new data requirements are not one-time data upgrades but require a permanent data connection with the ecological and social environment. A company and its ERP system or its key performance management users cannot import new data, such as metadata of the EU Taxonomy regulation, and call it a day. Data curation must be an ongoing activity.

All slides

Most companies import relatively little data regularly, and therefore, they have little experience in data curation, i.e., the art of organising, annotating, and integrating data collected from various sources in a way that is presentable in the form of indicators or can be reused later. Perhaps their bookkeeping needs to import foreign exchange rates regularly if they export or import in their activities. Data curation will become an ongoing activity if they start to monitor and measure the environmental and social environment impacts of their actions.

So, if we agree that joining a data sharing space, i.e., an organisation that has pre-agreed terms and conditions on sharing and exchanging data, with pre-agreed terminology or vocabulary (“semantics”) and technology, the question is, what kind of data sharing organisation is the most appropriate?

The CSRD Directive has industry-agnostic elements, which must be fulfilled in every company, and industry-specific elements, which are being developed as we speak. For example, we work mainly with music and film production, both of which belong to the Media and Entertainment Group, which must apply the same industry-specific variations of the European Sustainability Development Standards. We think that the best is to create industry-specific data sharing spaces, such as the famous Data 4.0 for manufacturing, but allow them to be federated and to exploit further synergies: there are plenty of financial, economic, social or environmental data that are used by other existing data sharing spaces and it is not necessary to curate and produce them in, for example, a music or film-production oriented data sharing space.

Are for-profit and social enterprises technically ready to join data-sharing spaces? Not without help. While some large corporations have explicit knowledge bases and enterprise graphs, most smaller European enterprises do not necessarily have a distinct IT function. It is a simple relational database with a fixed schema if they manage databases. Bringing them on board requires simple systems and assistance.

All slides

Reprex is building a data sharing system, Reprexbase, which is built around Wikibase as a knowledge broker system. Wikibase is the open-source software that hosts Wikidata, the largest open knowledge graph in the world. We extend it with various ETL modules and an ecosystem of peer-reviewed statistical libraries to create scientifically correct impact indicators and benchmarks. These extensions are necessary not only because most enterprises are not familiar with working with linked data or graphs but also because Companies are usually familiar with creating financial indicators (for SMEs, simple accounting indicators, larger enterprises with a controlling function, and more complex indicators) but not with the creation of non-financial statistical indicators. The CSRD directive calls for reporting more than 200 indicators over five environmental and four social matter groups, which is more than a company would have on a balanced scorecard. Reliably producing so many indicators is no small feat.

Large and public companies are directly responsible for CSRD reporting, i.e., integrating financial, environmental and social accounting. They need to improve their bookkeeping or ERP systems because often they do not even organise into a database much of the contents of the current invoices (physical quantities, units), which would allow them to create the factual basis between energy cost, energy use and energy footprint, for example. However, they must ask their suppliers for further environmental and social data. Without readily applicable Digital Product Passports, this is a considerable cost, estimated at around 1500 euros per supplier. If you are a festival organiser or a film producer with hundreds of suppliers, this will quickly pay for a simple data sharing space; if you do it with a cluster in your industry, the costs can be significantly reduced.