Primerjava metod jezikovnih tehnologij za odkrivanje lažnih novic

magistrsko delo

Nejc Lovrenčič (Author), Borko Bošković (Mentor), Janez Brest (Co-mentor)

Abstract

Socialna omrežja in tradicionalni viri novic imajo velik vpliv na razmišljanje ter dejanja posameznikov v družbi. Napačna ali izmišljena dejstva in lažne novice lahko zato povzročijo veliko škodo. V sklopu magistrskega dela smo primerjali metode Naivni Bayes, logistično regresijo, nevronsko mrežo z dolgim kratkoročnim spominom in graf konvolucijsko nevronsko mrežo za odkrivanje lažnih novic. S preučitvijo sorodne literature in primerjavo metod smo ugotovili, da je težko prepoznati lažne novice zgolj s klasifikacijo besedila. Pri klasifikaciji novic na dva razreda se je najbolje izkazal logistična regresija, pri klasifikaciji na šest razredov pa nevronska mreža LSTM.

Keywords

jezikovne tehnologije;nevronske mreže;lažne novice;klasifikacija besedila;magistrske naloge;

Data

Language:	Slovenian
Year of publishing:	2022
Typology:	2.09 - Master's Thesis
Organization:	UM FERI - Faculty of Electrical Engineering and Computer Science
Publisher:	[N. Lovrenčič]
UDC:	004.8:519.76(043.2)
COBISS:	110205187
Views:	177
Downloads:	40
Average score:	0 (0 votes)
Metadata:

Other data

Secondary language:	English
Secondary title:	Comparison of natural language processing methods for detecting fake news
Secondary abstract:	Social networks and traditional news sources have a great influence on individuals' thinking and actions. False and made-up facts, as well as fake news, can therefore cause a lot of damage. As a part of the master's thesis, we compared Naive Bayes, logistic regression, long short-term memory neural network, and graph convolutional network for news classification into two and six classes. By studying related literature and executing method comparisons, we have figured out that it is difficult to identify fake news simply by using text classification methods. When classifying the news into two classes, we achieved the best results using logistic regression, while LSTM neural network proved to be the best when classifying news into the six classes.
Secondary keywords:	natural language processing;neural networks;fake news;text classification;
Type (COBISS):	Master's thesis/paper
Thesis comment:	Univ. v Mariboru, Fak. za elektrotehniko, računalništvo in informatiko, Računalništvo in informacijske tehnologije
Pages:	1 spletni vir (1 datoteka PDF (IX, 40 f.))
ID:	14960290