Institutional Repository of Vilnius University
LLM naudojimas interaktyvioms audio knygoms kurti /
Abstract
dc:descriptionThis bachelor’s thesis examines the use of large language models (LLMs) for the creation of interactive and personalized audiobooks. The aim of the thesis is to develop a prototype interactive audiobook system that generates a narrative in real time based on user preferences and to evaluate the impact of personalization on user engagement and the listening experience. The theoretical part of the thesis analyzes interactive audiobooks, large language models, and personalization technologies. It discusses transformer-based models such as GPT-4 and LLaMA, their application to narrative generation, and the principles of adaptive storytelling. It also examines user profiling methods, dynamic prompt generation, and real-time content adaptation based on user behavior and preferences. In the practical part of the work, a modular system was designed and implemented, consisting of speech recognition (STT), user profile management, LLM-based story generation, and text-to-speech (TTS) components. The system allows the user to select the narrative genre, pace, and dialogue intensity, and these parameters are dynamically integrated into the LLM’s system prompts. The generated text is converted into a Lithuanian audio narrative using a neural TTS voice. The experimental section involved a study with 30 participants, comparing personalized and non-personalized versions of the system. User engagement was assessed using objective metrics (session duration, story completion rates, and number of interactions) as well as subjective survey metrics (engagement, enjoyment, and relevance). The results of the study showed that personalized storytelling increases user engagement, improves the listening experience, and provides a greater sense of control and relevance. The results of this work indicate that LLM-based personalized interactive audiobooks have the potential to become a new form of adaptive digital storytelling, enabling the creation of a dynamic and personalized user experience.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2026
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Tomkevič, Rafal,
Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- lit
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD308338232&prefLang=en_US