Interpretacija modela strojnog učenja koji predviđaju sentiment

master's thesis

• Title (english): Interpretation of Machine Learning Models Predicting Sentiment Analysis
• Author: Dorotea Protrka
• Mentor: Domagoj Jakobović (mentor)
• Committee member: Domagoj Jakobović (predsjednik povjerenstva)
• Committee member: Marko Đurasević (član povjerenstva)
• Committee member: Jan Šnajder (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2021-07-15, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

U vrijeme dominacije društvenih mreža u međuljudskoj komunikaciji, detekcija govora mržnje bitan je problem. Nakon interpretacije modela logističke regresije, čini se da, iako je pokazao dobre rezultate, ima prostora za poboljšanje. Kako bi se suočio s problemom pristranosti, GBT je obucen s istim značajkama. Rezultati su poboljšani. Drugi model, potpuno povezana neuronska mreža, osposobljena je pomoću vektoriziranih prikaza rijeci. Interpretacija modela dovodi do boljeg razumijevanja slabih strana modela.
Na temelju važnosti značajki, zaključujemo da je govor mržnje u tvitovima često rezultat političkog neslaganja i netolerancije.

Abstract (english)

At a time dominated by social networks, hate speech detection is an important
problem. Few different approaches were tested. After model interpretation, logistic
regression, even though it showed good results, appeared to have more space for improvement. To face the problem of bias, GBT was trained with the same features.
Results improved. Another model, fully connected neural network, was trained using
the word embeddings. Model interpretation lead to the better understanding of model’s
weak points.
Based on feature importance, hate speech in tweets is often a result of political disagreement and intolerance in general.

• Keywords: Interpretabilnost; analiza sentimenta; objašnjivi modeli; SHAP
• Keywords (english): Interpretability; Sentiment Analysis; Interpretable Models; SHAP
• Language: english
• URN:NBN: urn:nbn:hr:168:594704
• Study programme: Title: Computing; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra inženjer/inženjerka računarstva (magistar/magistra inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2023-01-13 13:35:03