Implementacija jezičnog asistenta temeljenog na umjetnoj inteligenciji za brodarsku industriju

master's thesis

• Title (english): Implementation of an artificial intelligence-based chatbot for the shipping industry
• Author: Ivan Špoljar
• Mentor: Marina Bagić Babac (mentor)
• Committee member: Marina Bagić Babac (predsjednik povjerenstva)
• Committee member: Gordan Gledec (član povjerenstva)
• Committee member: Boris Vrdoljak (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2024-07-11, Croatia

Abstract

Ovaj diplomski rad izrađen je kako bi se objasnili osnovni pojmovi i koncepti potrebni za razumijevanje kako radi jednostavan chatbot, da bi se demonstrirao njegov razvoj te da bi se usporedile performanse nekoliko odabranih velikih jezičnih modela (Large Language Models). Chatbot je razvijen slijedeći tehniku generiranja proširenog dohvaćanjem (Retrieval-Augmented Generation). Korišteni veliki jezični modeli su preuzeti sa Hugging Face-a. Modeli korišteni za računanje smještenja teksta (text embedding) su “nomic-embed-text-v1.5-Q5_K_M” i “snowflake-arctic-embed-m-long--Q5_K_M”. Modeli korišteni za generiranje teksta su “Meta-Llama-3-8B-Instruct-Q4_K_M” i “mistral-7b-openorca-oasst_top1_2023-08-25-v2.Q4_K_M”. Ova aplikacija sastoji se od stranice za unos novog znanja i stranice za pozivanje chatbot-a koja također sadrži kontejner za korištene izvore.

Abstract (english)

This paper was made to explain the essential terms and concepts needed to understand how a simple chatbot works, to demonstrate its development, and to compare performance of few selected Large Language Models. The chatbot was developed following the technique of Retrieval-Augmented Generation. The Large Language Models used are downloaded from Hugging Face. The models used for calculating text embeddings are “nomic-embed-text-v1.5-Q5_K_M” and “snowflake-arctic-embed-m-long--Q5_K_M”. The models used for text generation are “Meta-Llama-3-8B-Instruct-Q4_K_M” and “mistral-7b-openorca-oasst_top1_2023-08-25-v2.Q4_K_M”. This application consists of a page for inserting new knowledge and a page for calling a chatbot which also contains a container for used sources.

• Keywords: Chatbot; RAG; LLM; vektori; Semantic Kernel; Kernel Memory; LLamaSharp
• Keywords (english): Chatbot; RAG; LLM; vectors; Semantic Kernel; Kernel Memory; LLamaSharp
• Language: english
• URN:NBN: urn:nbn:hr:168:861629
• 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: 2024-10-12 22:18:09