Semantic Data Discovery over the SLAIF Catalogue
Overview
This service enables semantic search and discovery over the AI Factory's catalogue of services and assets, built on the semantic catalogue from Task 3.3. Users can post natural language queries via dedicated APIs and a user-facing search interface, or use structured faceted search to discover relevant assets---filtering by domain, service type, provider, license, technology readiness level, and other dimensions derived from the AI Factory OWL ontology---without requiring knowledge of SPARQL or the underlying data structure. Unlike keyword search, which relies on exact matches, the service analyses the semantics of users' queries against structured ontological descriptions of catalogue entries, leveraging a formal OWL ontology that describes service offerings, their domains, target audiences, prerequisites, licensing, provenance, and assets. The service will combine a triple store for RDF storage and SPARQL querying with a search engine for ranked full-text and faceted search, and an LLM-assisted layer for translating natural language queries into structured queries grounded in the ontology. Machine learning and natural language processing techniques, including query analysis and semantic similarity matching, are incorporated over the well-structured metadata descriptions maintained in the semantic catalogue. Metadata will be aligned with European interoperability standards, including EOSC Resource/Provider Profiles and MLDCAT-AP, ensuring interoperability with the broader European AI and open science ecosystem.
Outcome
A search and discovery interface returning ranked lists of SLAIF assets (datasets, AI models, workflows, PoC solutions) with machine-readable metadata, accessible via web interface and APIs.
Service Card
| Category | Infrastructure and data-related services |
| Domain | Data Infrastructure, Semantic Search |
| Vertical | Cross-domain |
| Target users | Software Solution Providers, Data Scientists, SMEs, Public Sector |
| Service type | Service |
| Outputs | Ranked catalogue entries with structured metadata, faceted browsing views, SPARQL query results. |
| Requirements | Server node for triple store and search engine; LLM API access for query translation. |
| Limitations | Discovery quality depends on the coverage and completeness of ontological annotations in the semantic catalogue. |
| Availability | Q3/2027 |
| Licensing | CC-BY 4.0 for metadata extracted by the discovery service; Apache 2.0 for the software |
| Visual content | |
| Visual caption |