Alberto Catania
Design and Development of a RAG-Based Knowledge System Using Vector Databases for Marketing and Business Intelligence.
Rel. Paolo Garza. Politecnico di Torino, Corso di laurea magistrale in Ingegneria Informatica (Computer Engineering), 2026
Abstract
Retrieval-Augmented Generation (RAG) is a standard paradigm for interacting with corporate knowledge bases via Large Language Models (LLMs). Applying traditional query-time RAG to unstructured community data (such as social media threads) introduces latency bottlenecks, context saturation, and "attention exhaustion". This thesis presents the design and implementation of a Hybrid Intelligence System that connects structured corporate frameworks, such as marketing personas, with unstructured community discourse for real-time business intelligence. The proposed architecture shifts the computational burden from query-time to ingestion-time. A Two-Step Agentic Pipeline decouples metric extraction from strategic reasoning, allowing the system to classify user pain points and psychological sub-personas without context dilution.
To handle the hierarchical nature of online discussions, a Conversation-Window Chunking strategy fuses child comments with parent context in O(1) time, avoiding the "lost in the middle" phenomenon without recursive parsing
Relatori
Anno Accademico
Tipo di pubblicazione
Numero di pagine
Informazioni aggiuntive
Corso di laurea
Classe di laurea
Aziende collaboratrici
URI
![]() |
Modifica (riservato agli operatori) |
