Ponedeljkov seminar računalništva in informatike - Arhiv

V ponedeljek, 20. aprila 2026, bo ob 16:00 uri izvedeno
predavanje v okviru PONEDELJKOVEGA SEMINARJA RAČUNALNIŠTVA IN INFORMATIKE
Oddelkov za Informacijske znanosti in tehnologije UP FAMNIT in UP IAM.

ČAS/PROSTOR: 20. april 2026 ob 16.00 v FAMNIT-VP3.

---------------------------------------------
PREDAVATELJ: Andrija PAVIĆEVIĆ
---------------------------------------------

Andrija Pavićević comes from Montenegro. He has finished his Bachelor's studies at the Faculty of Mathematics and Natural Sciences in Montenegro (course Computer Science). He is currently studying a Master's in Data Science at UP FAMNIT, finishing his thesis.

-----------------------------------------------------------------------------------------------------------------------------------------------------------------
NASLOV: Development of a centralized hyperlocal event recommendation system with chat-bot integration for tourists
-----------------------------------------------------------------------------------------------------------------------------------------------------------------

POVZETEK:

Being a tourist in today’s digital world may seem easy — but is it really? The Internet is full of information, yet for someone who wants to relax and enjoy their stay, browsing multiple sources in search of relevant events can be discouraging and dull. This thesis presents the development and evaluation of a centralized web application designed to help international users discover local events more efficiently. The application includes both dummy event data and submissions from local event coordinators. It also features a chatbot interface, built using RASA, which enables users to interact in natural language (currently in English) and prompt about events. Additionally, the system integrates a recommendation engine that leverages user preferences — such as likes and interactions — using collaborative filtering, matrix factorization, and hybrid models. A user study was conducted to compare the application’s performance with existing solutions, measuring efficiency, recommendation accuracy, and overall usability. Results indicate that the proposed system improves user experience and shortens search time. As for the recommendations, they might not be great, but there is much room for improvement. This research highlights the potential of unifying various data sources with intelligent, conversational interfaces to enhance the event discovery experience for travelers.

Seminar bo potekal v angleškem jeziku v predavalnici FAMNIT-VP3 s pričetkom ob 16:00 uri.

Vabljeni!

V ponedeljek, 13. aprila 2026, bo ob 16:00 uri izvedeno
predavanje v okviru PONEDELJKOVEGA SEMINARJA RAČUNALNIŠTVA IN INFORMATIKE
Oddelkov za Informacijske znanosti in tehnologije UP FAMNIT in UP IAM.

ČAS/PROSTOR: 13. april 2026 ob 16.00 v FAMNIT-VP3.

-------------------------------------------------
PREDAVATELJICA: Kaja PRAPROTNIK
-------------------------------------------------

Kaja Praprotnik is a PhD student at UP FAMNIT under the supervision of Prof. Jernej Vičič. Her research focuses on natural language processing with a strong focus on artificial intelligence.

-------------------------------------------------------------------------------------------------------------------------------
NASLOV: Fine-Tuning Small Language Models for Sign Language Gloss-to-German Translation
-------------------------------------------------------------------------------------------------------------------------------

POVZETEK:

Sign language is to some people the first language and in some cases even the only language they know. Writing in spoken language might present some challenges in everyday life. To prevent further isolation for people who already struggle to live in a world designed for hearing people, the translation from sign language to spoken language is necessary.
Without compromising user privacy, the translation system should enable accessibility to essential applications, such as healthcare communication systems. 
Large language models are because of their computational needs usually hosted in the cloud, which can pose a risk of leakage of sensitive personal data. 
Small language models can run locally on the user's device, which protects privacy of the sensitive data.
Our research focuses on fine-tuning a small language model, for translating written sign language into spoken language.

Seminar bo potekal v angleškem jeziku v predavalnici FAMNIT-VP3 s pričetkom ob 16:00 uri.

Vabljeni!