For small and medium-sized enterprises (SMEs), participating in public tenders has become increasingly complex: information is scattered across multiple documents, published in heterogeneous formats, and often lacks timely notifications, making it hard to locate and use in time. This thesis introduces OFFLINETender, a strategic utility engineered according to privacy-by-design principles: thanks to Retrieval-Augmented Generation (RAG), answers are produced exclusively from the indexed document context, preventing the disclosure of sensitive information. By integrating Large Language Models (LLM) with RAG, OFFLINETender automatically extracts key metadata from public procurement documents and, through a simple and intuitive interface, enables users not only to view this data but also to query it via a chatbot, all while guaranteeing the full confidentiality of the information processed.
Per le piccole e medie imprese (PMI) partecipare ai bandi di gara pubblici è diventato sempre più complesso: le informazioni sono frammentate su documenti diversi, pubblicate in formati eterogenei e spesso prive di notifiche tempestive, rendendo difficile reperirle e utilizzarle in tempo utile. Questa tesi presenta OFFLINETender, una utility strategica progettata secondo i principi del privacy-by-design: grazie all’impiego del Retrieval-Augmented Generation (RAG), le risposte vengono generate esclusivamente a partire dal contesto documentale indicizzato, evitando la divulgazione di informazioni sensibili. Integrando Large Language Models (LLM) con RAG, OFFLINETender estrae automaticamente i metadati chiave dai documenti di appalto pubblico e, tramite un’interfaccia semplice e intuitiva, consente non solo di visualizzarli, ma anche di interrogarli mediante un chatbot, garantendo al contempo la completa riservatezza dei dati trattati.
OFFLINETender: Estrazione intelligente di dati strutturati da documenti di gare e appalti pubblici tramite Large Language Model e Retrieval-Augmented Generation
PIRON, MATTEO
2024/2025
Abstract
For small and medium-sized enterprises (SMEs), participating in public tenders has become increasingly complex: information is scattered across multiple documents, published in heterogeneous formats, and often lacks timely notifications, making it hard to locate and use in time. This thesis introduces OFFLINETender, a strategic utility engineered according to privacy-by-design principles: thanks to Retrieval-Augmented Generation (RAG), answers are produced exclusively from the indexed document context, preventing the disclosure of sensitive information. By integrating Large Language Models (LLM) with RAG, OFFLINETender automatically extracts key metadata from public procurement documents and, through a simple and intuitive interface, enables users not only to view this data but also to query it via a chatbot, all while guaranteeing the full confidentiality of the information processed.| File | Dimensione | Formato | |
|---|---|---|---|
|
TESI_MATTEO_PIRON.pdf
Accesso riservato
Dimensione
2.83 MB
Formato
Adobe PDF
|
2.83 MB | Adobe PDF |
The text of this website © Università degli studi di Padova. Full Text are published under a non-exclusive license. Metadata are under a CC0 License
https://hdl.handle.net/20.500.12608/93199