Uporaba umetne inteligence za povzemanje in analitično obravnavo videoposnetkov

Pavčnik, Luka (author), Rožanc, Igor (mentor)

Diplomska naloga obravnava problem učinkovitega povzemanja in analize videoposnetkov z uporabo umetne inteligence. Izziv predstavlja hitro pridobivanje ključnih informacij iz dolgih videoposnetkov, kar je posebej pomembno v izobraževalnem in raziskovalnem kontekstu. Pristop temelji na integraciji naprednih modelov umetne inteligence kot sta Whisper za pretvorbo govora v besedilo in GPT za analizo vsebine. Razvita je bila spletna aplikacija, ki omogoča samodejno generiranje povzetkov, ključnih točk in odgovarjanje na vprašanja o vsebini videoposnetkov. Na izbranem videoposnetku so rezultati pokazali, da aplikacija uspešno tvori jasne in natančne informacije. Ključni prispevek je implementacija rešitve za obdelavo daljših videoposnetkov in izboljšanje kakovosti podnapisov še zlasti za slovenske vsebine, kar znatno razširja uporabnost aplikacije za različne tipe videoposnetkov v različnih jezikovnih okoljih.

• Language: Slovenian
• Keywords: umetna inteligenca, veliki jezikovni modeli, openAI, whisper, GPT
• Work type: Bachelor thesis/paper
• Typology: 2.11 - Undergraduate Thesis
• Organization: FRI - Faculty of Computer and Information Science
• Year: 2024
• PID: 20.500.12556/RUL-162286
• COBISS.SI-ID: 214136579
• Publication date in RUL: 20.09.2024

PAVČNIK, Luka, 2024, Uporaba umetne inteligence za povzemanje in analitično obravnavo videoposnetkov [online]. Bachelor’s thesis. [Accessed 17 May 2025]. Retrieved from: https://repozitorij.uni-lj.si/IzpisGradiva.php?lang=eng&id=162286

Secondary language

• Language: English
• Title: Using artificial inteligence for video summarization and analytical processing

Abstract

This thesis addresses the challenge of efficiently summarizing and analyzing video content using artificial intelligence. The solution lies in quick extraction of key information from lengthy videos, which is particularly crucial in educational and research contexts. The approach integrates advanced AI models such as Whisper for speech-to-text conversion and GPT for content analysis. A web application was developed that enables automatic generation of summaries, key points, and answers to questions about video content. Results from a selected video demonstrated that the application successfully produces clear and accurate information. The key contribution is the implementation of a solution for processing longer videos and improving subtitle quality, especially for Slovenian content, significantly expanding the application's usability across different language environments and video types.

• Keywords: artificial intelligence, large language models, openAI, whisper, GPT