This thesis outlines the analysis, evaluation, and integration of Generative Artificial Intelligence models designed for the automatic classification and semantic data extraction of digital assets within the THRON DAM platform. The project, developed during an internship at THRON S.p.A., was structured in two primary phases. The first phase focused on a comparative evaluation of multimodal AI models and services for vector content analysis. Concurrently, prompt engineering techniques were developed to translate user inputs into structured, effective prompts. This included asset optimization and reprocessing strategies designed to maximize analysis efficiency according to the selected LLM's capabilities. The second phase involved the technical implementation of the solution, which required developing backend components in Go and integrating them into an AWS-based serverless architecture. Ultimately, the selected model was successfully deployed within the platform's Automation Studio as a new, fully functional automation.
Il presente documento ha lo scopo di descrivere l'analisi, la valutazione e l'integrazione di modelli di Intelligenza Artificiale Generativa finalizzati alla classificazione automatica e all'estrazione di informazioni semantiche degli asset digitali all'interno della piattaforma THRON DAM. Il progetto, svolto durante il tirocinio presso THRON S.p.A. di Piazzola sul Brenta (PD), si è articolato in due fasi principali. La prima ha previsto una valutazione comparativa dei modelli e servizi AI multimodali per l'analisi vettoriale dei contenuti. Parallelamente, sono state studiate e implementate strategie di prompt engineering, con l'obiettivo di tradurre le richieste degli utenti in prompt strutturati ed efficaci, e strategie di ottimizzazione e rielaborazione degli asset con l'obiettivo di rendere l'analisi più efficiente ed efficace rispetto alle capacità del modello LLM. La seconda fase ha riguardato l'implementazione tecnica della soluzione tramite lo sviluppo di componenti backend in linguaggio Go e l'integrazione in un'architettura serverless basata su servizi AWS. Il modello selezionato è stato infine inserito all'interno della piattaforma Automation Studio come nuova automazione.
Classificazione semantica di contenuti digitali con Intelligenza Artificiale
ROCCO, MATTEO ALBERTO
2025/2026
Abstract
This thesis outlines the analysis, evaluation, and integration of Generative Artificial Intelligence models designed for the automatic classification and semantic data extraction of digital assets within the THRON DAM platform. The project, developed during an internship at THRON S.p.A., was structured in two primary phases. The first phase focused on a comparative evaluation of multimodal AI models and services for vector content analysis. Concurrently, prompt engineering techniques were developed to translate user inputs into structured, effective prompts. This included asset optimization and reprocessing strategies designed to maximize analysis efficiency according to the selected LLM's capabilities. The second phase involved the technical implementation of the solution, which required developing backend components in Go and integrating them into an AWS-based serverless architecture. Ultimately, the selected model was successfully deployed within the platform's Automation Studio as a new, fully functional automation.| File | Dimensione | Formato | |
|---|---|---|---|
|
Rocco_Matteo_Alberto_2042354.pdf
Accesso riservato
Dimensione
10.49 MB
Formato
Adobe PDF
|
10.49 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/20.500.12608/116517