AI-BRIDGES Open Forum: June meeting
Hosting Jason Evans & Adrián Cuadrón Cortés
In May we met physically at the AI-BRIDGES Symposium and resumed our online meeting on June 26th at 15:00.
Meeting agenda included:
* Presentation by Jason Evans, Open Data Manager & AI Lead, at the National Library of Wales.
* Presentation by Adrián Cuadrón Cortés, a researcher working at the HiTZ Center (University of the Basque Country) on the ECHOLOT project.
* Update on the AI-BRIGDES project: including the AI-BRIDGES Symposium in May, an update on the project’s development, and a brief discussion of Wikimania, which will be hosted this year in Paris during July 21-25.
Below is the recording, as well as details of the presentation and an AI summary.
As we will be meeting in July at Wikimania, and August is a summer vacation, our next online meeting will be September 25ht.
Stay cool and enjoy the summer,
Shani.
————————————————–
Meeting recording
Details of Jason’s presentation
Title: “Transforming Data with Wikidata and AI and the National Library of Wales”
Short blurb: Jason Evans, Wikimedian, Open Data Manager and AI lead at the National Library of Wales will discuss efforts to enrich traditional library Metadata using AI tools and Wikidata. His talk will explore the challenges and opportunities of employing AI in a trusted knowledge institution and the importance of Open crowd sourced data in enriching digital infrastructure. He will show practical examples of how AI can be used to enrich data, and improve access to knowledge.”
Details of Adrián’s presentation
Title: “Preserving Cultural Heritage in the Digital Knowledge Space: Entity Recognition and Linking with culturally aware LLMs”
Short blurb: Integrating cultural heritage collections into the digital knowledge space requires accurately identifying and linking entities across multilingual historical documents. In the ECHOLOT project, we explore the use of Large Language Models (LLMs) to recognize, disambiguate, and reconcile entities with open knowledge bases such as Wikidata. Focusing on Basque (Euskara), a low-resource language, we investigate how culturally aware models such as Latxa, developed by the HiTZ Center, can improve entity linking by leveraging stronger linguistic and cultural knowledge.
Short Bio: I studied Computer Engineering and later completed a Master’s in Language Analysis and Processing. I am currently a researcher working at the HiTZ Center (EHU) on the Echolot project, where my work focuses on entity recognition and entity reconciliation. In previous projects, I have also worked extensively on Retrieval-Augmented Generation (RAG) systems.
AI Summary
AI and Linked Data Projects
AI Collaboration and Data Reconciliation
Shani shared additional context on ECHOLOT, the project connected to Adrian’s work, emphasizing the potential collaboration with AI-BRIDGES to avoid duplicating efforts. Jason asked about challenges in AI reconciliation, to which Adrian responded that errors often stem from incorrect database links or Wikidata IDs, and highlighted the importance of defining prompts accurately for the model. David provided additional background on the ECHOLOT project, explaining its goal to enable GLAM institutions to upload and enrich data for the Wikimedia ecosystem and European aggregators, while noting challenges in entity linking, particularly with name variants and scoring thresholds.
David reported positive results using Wikidata embeddings for intelligent open refine, though the current implementation is limited to entities with Wikipedia articles. Adrián noted challenges with entity duplication in Wikidata that could affect evaluation accuracy. Shani announced that the AI Bridges pipeline development work is being paused over the summer to avoid duplication of efforts with ECHOLOT, with plans to resume in September using a new feedback tool for collaborative design input. The AI-BRIDGES team is also developing a Synthetic Reasoning Corpus and will report more closer to the end of the year.
Wikibase Large Language Model Integration
Philippe shared a new MCP for connecting LLMs to Wikibase data, which was recently relaunched at the AI-BRIDGES Symposium. For questions on that, please write: [email protected].
Shani updated the group on the Symposium recordings and mentioned a mini-strategic meeting held with about 40 Wikimedians and adjacent organizations.
AI Events at Wiki AI Fringe
Sam and Shani discussed upcoming AI-related events at Wiki AI Fringe and Wikimania. Sam explained that there will be a pre-conference day on Tuesday with various tool builders and workshops, followed by a fringe festival with hacking sessions, unconference sessions, and demonstrations from Wednesday through Friday. Shani mentioned that there is a full program of AI-related activities at the main Wikimania venue, including a panel on AI-BRIDGES and a State of Wiki & AI session with Jimmy. Sam also shared information about the Public AI Network’s annual seminar series, which focuses on the practicalities of hosting public AI services (register here if interested: https://publicai.network/
The next open meeting will take place in September due to Wikimania & summer vacation, with exciting discussions planned for the rest of the year.