Milestone 1: Building the Lemma Bank
The first phase of the LiITA: Linking Italian project focusses on building a Lemma Bank from existing Italian lemma sets that will be meticulously selected and compared. The Lemma Bank is available as Linked Open Data, adhering to the widely accepted vocabulary outlined in the OntoLex-Lemon model for describing lexical resources:
- Browse the Lemma Bank: https://liita-lod.github.io/query-interface/
- View: http://liita.it/data/id/lemma/LemmaBank
- Download: https://github.com/LiITA-LOD/LiITA_LemmaBank
Milestone 2: Linking the Resources
Access the data through our SPARQL endpoint.
The second phase revolves around integrating a set of freely available Italian linguistic resources with the Knowledge Base.
Lexical Resources
CompL-it
Computational Lexicon for Italian. See http://hdl.handle.net/20.500.11752/ILC-1007 for more details on the resource. View: (https://dspace-clarin-it.ilc.cnr.it/repository/xmlui/handle/20.500.11752/ILC-1007), download ttl (https://dspace-clarin-it.ilc.cnr.it/repository/xmlui/bitstream/handle/20.500.11752/ILC-1007/complit.ttl.gz?sequence=1&isAllowed=y).
Vocabolario della Lingua Parmigiana
A bilingual lexicon having Italian entries and the corresponding translations in Parmigiano, edited by Umberto Pavarini and Gruppo di Lavoro Memento Mori. View: (https://liita.it/data/id/DialettoParmigiano/lemma/LemmaBank.html), download ttl (linkhttps://github.com/LiITA-LOD/LocalVarieties/tree/main/Parmigiano).
Sicilian-Italian Lexicon
A Sicilian-Italian lexicon extracted from wikizziunariu.
View: https://liita.it/data/id/LexicalResources/DialettoSiciliano/Wikizziunariu.html
Download ttl: https://github.com/LiITA-LOD/LocalVarieties/tree/main/Siciliano.
Sentix
Sentix is an affective lexicon for the Italian language, incorporating 63,660 entries, with associated polarity scores (ranging from -1 to +1) and categorical polarity classifications (Positive, Neutral, Negative).
ELIta
Emotion Lexicon for Italian, where words are annotated for basic emotions with scores from 0 to 1 and the emotion dimensions from 1 to 9.
Textual Resources
Luigi Pirandello’s Novellas
This corpus contains a series of short stories by Luigi Pirandello. The texts have been tokenised, pos-tagged and lemmatised specifically for linking to the LiITA Knowledge Base.
View: https://liita.it/data/id/corpora/Pirandello/id/corpus.html
Milestone 3: Developing Tools
The project’s third phase brings forth the development of a tool that empowers resource providers to automatically connect their data to the Lemma Bank. Coupled with the project’s commitment to utilising established vocabularies for knowledge representation as LOD, this tool promotes an open-ended approach. As a result, the knowledge base becomes readily extensible and adaptable for future enrichment and expansion.
NL2SPARQL
https://github.com/tonazzog/nl2sparql
A Retrieval-Augmented Generation system for querying LiITA in natural language: semantic retrieval over few-shot examples guides an LLM to generate SPARQL queries, with syntax validation, execution, and automatic error correction. Includes an agentic (ReAct-based) variant and an MCP server for integration with external clients.
PRISMA
https://github.com/tonazzog/prisma-per-liita
A natural-language interface to LiITA where an LLM is used only to classify user intent, while SPARQL generation is fully deterministic and template-based, ensuring transparent and structurally correct queries.
MoSAIC
https://github.com/tonazzog/mosaic-liita
A block-based system that assembles SPARQL queries over LiITA from reusable pattern components, offering both a fully deterministic rule-based mode and an agentic mode where an LLM decomposes the query into structured operations.