20.500.12556/DKUM-97623 Analiza netekočnosti v govorjenem jeziku s strojnim učenjem magistrsko delo Disfluency analysis in spoken language with machine learning Magistrsko delo obravnava analizo govorne netekočnosti v govorjeni komunikaciji z uporabo metod strojnega učenja. Namen raziskave je bil preučiti pojavnost in značilnosti netekočnosti ter ovrednotiti možnost njihove samodejne prepoznave z uporabo akustičnih značilk. Delo se umešča na področje obdelave naravnega jezika in govorne tehnologije. Uporabljene so bile metode ekstrakcije značilk ter modeli, kot so klasifikator z naključnimi gozdovi, skriti modeli Markova in nevronske mreže. Rezultati kažejo, da je netekočnosti mogoče prepoznati z zadovoljivo natančnostjo, vendar je pri interpretaciji potrebna previdnost. This master's thesis explores speech disfluencies in spontaneous speech and their automatic detection using machine learning methods. The aim of the study was to analyze the occurrence of disfluencies and evaluate whether they can be successfully detected based on acoustic features. The research belongs to the field of natural language processing and speech technology. Feature extraction methods and models such as Random Forest Classifier, Hidden Markov Models and neural networks were applied. The results show that disfluencies can be detected with satisfactory accuracy; however, careful interpretation of the findings is required. govorne netekočnosti strojno učenje obdelava govora speech disfluencies machine learning speech processing true true false [L. Rantuša] Slovenski jezik Angleški jezik Magistrsko delo/naloga 2026-03-25 19:34:16 2026-05-08 13:48:23 2026-05-16 03:14:17 0000-00-00 00:00:00 2026 0 Maribor 0 Maribor 1 spletni vir (1 datoteka PDF (IX, 80 str.)) 0000-00-00 NiDoloceno NiDoloceno NiDoloceno 0000-00-00 0000-00-00 0000-00-00 004.934.2:004.85(043.2) 278308611 MAG_Rantusa_Lara_2026.pdf MAG_Rantusa_Lara_2026.pdf 1 BC118883D7CEEB221F2C8C9345DD275F b7b25851de42bfca915b1c651fe91a23971c98e50da28f7bcb67f9360d3051f7 6c5db252-287a-11f1-bbad-00155d0001a1 https://dk.um.si/Dokument.php?lang=slv&id=204796 Fakulteta za elektrotehniko, računalništvo in informatiko 0 0 0