BEGIN:VCALENDAR VERSION:2.0 PRODID:-//pretalx//talks.osgeo.org//foss4g-europe-2026//talk//SAADFE BEGIN:VTIMEZONE TZID:EET BEGIN:STANDARD DTSTART:20001029T050000 RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=10 TZNAME:EET TZOFFSETFROM:+0300 TZOFFSETTO:+0200 END:STANDARD BEGIN:DAYLIGHT DTSTART:20000326T040000 RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=3 TZNAME:EEST TZOFFSETFROM:+0200 TZOFFSETTO:+0300 END:DAYLIGHT END:VTIMEZONE BEGIN:VEVENT UID:pretalx-foss4g-europe-2026-SAADFE@talks.osgeo.org DTSTART;TZID=EET:20260629T150000 DTEND;TZID=EET:20260629T153000 DESCRIPTION:Metadata are a foundational component of any data infrastructur e\, as they enable dataset discovery\, evaluation\, and reuse. Metadata cr eation and maintenance\, however\, is typically a costly\, inconsistent\, and largely manual process. In the European context\, this bottleneck is visible in major data sharing initiatives such as the INSPIRE Directive an d the Common European Data Spaces\, where high-quality\, interoperable met adata are a legal requirement as well as a precondition for data market pa rticipation.\nTwo main challenges affect the creation and maintenance of g eospatial metadata. The first is schema heterogeneity\, since existing sch emas (including ISO 19115\, DCAT\, GeoDCAT-AP\, Dublin Core\, INSPIRE prof iles) show partially overlapping semantics but incompatible serialisations . Migration between schemas is typically performed through hand-crafted tr ansformations encoding explicit field-to-field mappings: brittle\, schema- specific artefacts requiring specialist maintenance. The second challenge is content inconsistency: even within a single schema\, records produced b y different operators (including within the same organisation) may exhibit inconsistent terminology\, structure\, and level of detail in free-text f ields – which undermines catalogue-level discoverability.\nGenerative Artificial Intelligence (GenAI) and Large Language Models (LLMs) in partic ular offer promising capabilities to address these challenges. AI-assist ed metadata management (acquisition\, cleaning\, verification and maintena nce) has been already explored in literature [1]\, but to our knowledge no t in the case of geospatial datasets. Existing work includes for example A I support for digitizing libraries\, classifying research resources\, and create website content. Whatever the purpose\, research has focused on how to instruct LLMs to produce structured outputs for high-quality metadata [2] or to address the limitations of metadata schemas\, particularly free- text attributes whose definitions are often self-referential or provide in sufficient contextual meaning for GenAI to process [3].\nThis work extends such literature by proposing an open-source\, AI-powered geospatial metad ata editor. The core design principle is schema agnosticism: output schema s are defined as declarative configuration files\, and any new schema can be registered without modifying the application code. Field extraction and generation are driven by attribute type classification\, independently of field names or schema-specific logic\, with GeoDCAT-AP 3.0.0 as the curre nt default.\nThe tool recognises three different types of metadata attribu tes and tackles them differently. Type A attributes\, such as title\, keyw ords\, and description\, are characterised by a free-text nature and as su ch\, they are suitable for automatic generation using LLMs and prompts wit h contextual information. Type B are technical attributes such as spatial resolution\, bounding box\, file extension\, etc. that can be automaticall y extracted from the dataset by standard functions. Finally\, Type C are o rganisational attributes\, such as publisher\, contact point and licensing : these can be usually reused across several records within the same organ isation.\nInput handling is multimodal. For extraction of Type B attribute s\, the tool accepts native geospatial datasets processed via GDAL/OGR\, o r the URL to the OGC service serving the dataset. As a third option\, usef ul for large datasets\, the textual output of the gdalinfo or ogrinfo insp ection utilities can be provided directly. Type A attributes are generated using accompanying documentation on the relevant dataset\, such as techni cal reports and scientific publications\, which are indexed as corpus cont ent. The generation phase implements Retrieval-Augmented Generation (RAG): semantically oriented queries are issued against the vector corpus for ea ch free-text attribute\, results are ranked by relevance\, and a context p assage is assembled to inform generation. The LLM component is any OpenAI- compatible inference endpoint\, supporting both cloud-hosted and fully loc al deployments.\nThe tool supports three modes of operation through a unif ied processing pipeline\, with differences arising solely from the inputs provided. In the creation mode\, the user supplies a geospatial dataset a nd supporting documentation. Technical (Type B) attributes are populated d eterministically from the dataset\; publisher’s information (Type C attr ibutes) is parsed from the structured publisher document provided as input . Finally\, by querying the assembled corpus and including the Type B and C metadata attributes\, the LLM generates free-text fields (Type A) such a s title\, description\, keywords\, and provenance. A shared\, versioned pr ompt template encodes conventions – expected content sections\, field or der\, and level of specificity – applied consistently across all generat ed metadata records. This template functions as a content harmonisation in strument: metadata produced by different operators for different datasets converge on a common descriptive structure\, improving catalogue-level dis coverability without requiring ad-hoc normalisation. The use of a template and of different prompting techniques to reduce model hallucination and i mprove harmonisation across datasets has been investigated in an article c urrently under review [4].\nIn the schema migration mode\, the user addit ionally supplies an existing metadata record in any supported structured f ormat. The legacy record is parsed into field candidates and indexed in th e retrieval corpus. During generation\, the LLM receives the legacy metada ta as contextual input\; the mapping from source schema to target schema f ields emerges from the model semantic knowledge of both standards rather t han from explicit transformation rules\, generalising to schema pairs for which no hand-crafted converter exists.\nFinally\, in the enrichment mode\ , the user supplies a partial record or publisher documentation declaring organisational fields such as publisher\, license\, and contact point\; t hese are incorporated at the extraction time\, while remaining free-text f ields are generated and harmonised through the prompt template.\nIn all th ree modes\, the LLM is invoked only during the generation phase and only u pon explicit user action\, preserving a human-in-the-loop validation step before the export.\nSource code was submitted for intellectual property an d security screening for release under the open-source European Union Publ ic License (EUPL) according to the organisation policy. Schema definitions and prompt templates are expressed as human-readable declarative files\, versioned and customisable without modifying application code. \nThe empir ical evaluation in [4] is currently bounded by the institutional homogenei ty of JRC-produced metadata\; future research will address this limitation by incorporating records from Member State authorities engaged in the INS PIRE GeoDCAT-AP Pilot [5]\, enabling a more representative assessment of t he tool's harmonisation capacity across heterogeneous producer communities . DTSTAMP:20260605T020624Z LOCATION:A01 SUMMARY:An Open-Source AI-Powered Geospatial Metadata Editor for Schema-Agn ostic Generation\, Migration\, and Content Harmonisation - Margherita Di L eo URL:https://talks.osgeo.org/foss4g-europe-2026/talk/SAADFE/ END:VEVENT END:VCALENDAR