Implementacija i optimizacija arhitekturalnog obrasca "dohvatom pojačano generiranje"

master's thesis

• Title (english): Implementation and optimization of "Retrieval-Augmented Generation" architectural pattern
• Author: Franjo Mindek
• Mentor: Mario Brčić (mentor)
• Committee member: Mario Brčić (predsjednik povjerenstva)
• Committee member: Mirjana Domazet-Lošo (član povjerenstva)
• Committee member: Nikica Hlupić (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2024-07-05, Croatia

Abstract

Veliki jezični modeli (engl. Large Language models, LLM) dobili su značajnu pozornost u modernim aplikacijama, međutim, često su ograničeni statičkom prirodom podataka na kojima su trenirani, što dovodi do netočnosti i zastarjelih odgovora. Kao odgovor na ove izazove, arhitekturni obrazac dohvatom pojačano generiranje (engl. Retrieval-Augmented Generation, RAG) pojavljuje se kao obećavajuće rješenje. Ovaj diplomski rad ima za cilj istražiti implementaciju i optimizaciju RAG arhitekturnog obrasca. Diplomski rad je usmjeren na dva RAG temeljna koncepta: gutanje i dohvaćanje podataka, za koje su implementirane različite strategije. Svaka implementacija je provedena kroz okvir za testiranje, koji implementaciju ocjenjuje po definiranim metrikama.

Abstract (english)

Large Language Models (LLMs) which have gained significant traction in modern applications are often constrained by the static nature of
their training data, leading to inaccuracies and outdated responses.
In response to these problems, the Retrieval-Augmented Generation (RAG) architectural pattern emerges as a promising solution.
This thesis aims to delve into the implementation and optimization of the RAG architectural pattern. This is done by focusing on two different RAG core concepts: data ingestion and retrieval, for which different strategies were implemented. Each of the implementations was run through the testing environment which scored it based on predefined metrics.

• Keywords: RAG; LLM; semantičko pretraživanje; vektorske baze podataka; NLP
• Keywords (english): RAG; LLM; semantic search; vector databases; NLP
• Language: english
• URN:NBN: urn:nbn:hr:168:706350
• Study programme: Title: Computing; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra inženjer/inženjerka računarstva (magistar/magistra inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2024-11-07 05:40:57