Prepoznavanje govora mržnje korištenjem strojnog učenja

undergraduate thesis

• Title (english): Identifying Hate Speech Using Machine Learning
• Author: Marija Maršić
• Mentor: Marin Šilić (mentor)
• Committee member: Marin Šilić (predsjednik povjerenstva)
• Committee member: Helena Čeović (član povjerenstva)
• Committee member: Adrian Satja Kurdija (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2023-06-29, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Opisuju se modeli i arhitekture mreža pogodni za modeliranje prirodnog jezika. Navedeni su i opisani neki od modela često korištenih za prepoznavanje govora mržnje u tekstu. Zatim je ispitana uspješnost prepoznavanja govora mržnje koristeći gubitak, preciznost i F1 vrijednost na 4 različita scenarija: skup podataka na engleskom jeziku provučen kroz model BERT, skup podataka na engleskom jeziku provučen kroz model mBERT, skup podataka na španjolskom preveden na engleski i provučen kroz model BERT i skup podataka na španjolskom jeziku provučen kroz model mBERT. Dobiveni rezultati su tablično i grafički prikazani te analizirani. Razjašnjeni su ključni dijelovi programskog koda korišteni za dobivanje i prikaz rezultata ovog rada.

Abstract (english)

Models and network architectures suitable for natural language processing are presented and described. Some of the models frequently used for hate speech detection in text are listed and described. Then, the performance of hate speech detection was evaluated using loss, precision, and F1 value in four different scenarios: hate speech detection in English dataset using the BERT model, hate speech detection in English dataset using the mBERT model, hate speech detection in Spanish dataset translated into English using the BERT model, hate speech detection in Spanish language dataset using the mBERT model. The obtained results were presented in tables and graphs and analyzed. The important parts of the source code used to obtain and display the results of this study are explained.

• Keywords: BERT; mBERT; gubitak; binarna unakrsna entropija; preciznost; F1; trening; validacija; prenaučenost
• Keywords (english): BERT; mBERT; loss; binary cross entropy; accuracy; F1; training; validation; overfitting
• Language: croatian
• URN:NBN: urn:nbn:hr:168:724225
• Study programme: Title: Computing; Study programme type: university; Study level: undergraduate; Academic / professional title: sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva (sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2023-10-11 09:43:43