Use the opportunity to share your latest projects or ideas in a short (3-5 min) lightning talk. Submit your lightning talk by replying to this topic until Tuesday, 18 November, 23:59 UTC.
SWIB25 Lightning Talks
Semantic Web Agents Redux: From Knowledge Representation to AI-Driven Discovery (@jimfhahn)
Slides: semantic_web_agents_redux.pdf (410.9 KB)
Recordings: YouTube
The emergence of the Semantic Web in the early 2000s, followed by initiatives like the BIBFRAME ontology a decade later, introduced linked data as a transformative approach to knowledge representation and reasoning in libraries. Semantic Web theorists envisioned intelligent agents using structured RDF/OWL to take meaningful actions on behalf of users. However, early implementations struggled with the inherent trade-off between the structured precision of formal representations and the computational cost of reasoning, limiting their practical adoption.
Today, advances in generative AI, particularly Retrieval Augmented Generation (RAG), signal the possibility of a new class of Semantic Web agents that fuse the structured logic of linked data with the adaptability of LLM based reasoning. By integrating BIBFRAME based GraphQL APIs from the Share-VDE discovery system with chat interfaces, libraries can enhance search experiences, enabling more contextual and dynamic connections across collections.
This presentation explores a pilot implementation utilizing Share-VDE’s BIBFRAME APIs alongside generative AI-driven text generation, demonstrating how libraries can integrate LLM technologies to transform conventional catalog search into intelligent knowledge discovery.
Semantic metadata for software and AI models (@ljcastro)
Slides: 2025.11.19 - Metadata for SW and AI.pptx.pdf (1.3 MB)
Discussion: Semantic metadata for software and AI models
Recordings: YouTube
Semantic metadata for research artifacts makes it easier to realize the FAIR principles while also providing a good overview for humans and machines. Such metadata for scholarly articles is well-covered, with metadata about datasets also getting momentum. However, metadata for other research artifacts, e.g., software and Artificial Intelligence (AI) models, is not yet that frequently provided. We will present our approach to schema.org-based metadata for software (compatible with Software Management Plans) and AI models, including tools making easier for researchers to share metadata for their own research artifacts.
A domain-agnostic toolkit for community-curated SKOS vocabularies (@David_Linke)
Slides: 2025-11-19_SWIB2025_LightningTalk_Community-curated_SKOS_vocabularies_DLinke.pdf (987.9 KB)
Discussion: A domain-agnostic toolkit for community-curated SKOS vocabularies
Recordings: YouTube
Maintaining SKOS vocabularies collaboratively is hard. RDF/turtle syntax intimidates domain experts. Manual validation is error-prone. Tracking changes and discussions scattered across emails is chaotic. Result? Many communities avoid creating controlled vocabularies altogether.
We built an open-source toolkit that simplified the process: contributors edit terms in Excel spreadsheets, submit via GitHub pull requests, and automated workflows handle everything else – SKOS conversion, SHACL validation, documentation generation, and publishing with persistent URIs and content negotiation.
Why Excel? Familiar to everyone. Why GitHub? It provides version control with complete history tracking, structured discussions via issues/PRs, free hosting, automatic Zenodo publishing and increasingly, built-in AI assistance for contributors and maintainers. The toolkit (Python package + repository template) is domain-agnostic – originally developed for a catalysis vocabulary, but reusable by any community.
Published on Zenodo under: doi:10.5281/zenodo.17653394
JSON-LD context and framing for BIBFRAME data in the Bluecore Project (@edsu)
Slides: tbd
Discussion: JSON-LD context and framing for BIBFRAME data in the Bluecore Project
Recordings: YouTube
If there’s time I’d like to talk briefly about our use of JSON-LD context and framing for BIBFRAME data in the Bluecore Project, and make a case for shared community practices around BIBFRAME JSON-LD data.
LLM-Based Information Extraction to Support Scientific Literature Research and Publication Workflows (@Samy)
Slides: Samy-Ateia-SWIB25-Lightning-talks-slides.pdf (163.7 KB)
Discussion: LLM-Based Information Extraction to Support Scientific Literature Research and Publication Workflows
Recordings: YouTube
We evaluated the performance of open-weight and proprietary LLMs on the automatic extraction of key concepts from scientific texts. We focused on a specific example domain, namely business process management in computer science. The models answered expert-generated extraction questions for coding 122 papers from the Business Process Management Conference using the full text of the papers, and these answers were compared to our manual gold standard.
Open-weight models like Qwen-2.5 and Llama-3.3 seem competitive with proprietary models. Building on this work, in 2026 we plan on training specialized small models for cost-efficient local extraction and explore the workflow integration in Zotero. We additionally plan to explore embedding and retrieval for longer key concepts, such as the research question of the paper, on a centralized search service to improve related-work discovery.