Usporedba pristupa analizi sentimenta

master's thesis

• Title (english): Comparison of Sentiment Analysis Approaches
• Author: Dunja Vesinger
• Mentor: Marko Čupić (mentor)
• Committee member: Marko Čupić (predsjednik povjerenstva)
• Committee member: Vlado Glavinić (član povjerenstva)
• Committee member: Alan Jović (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2019-07-10, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Ovaj rad proučava različite pristupe rješavanja problema binarne analize sentimenta. Njezin je zadatak odrediti je li sentiment izražen u nekom tekstualnom dokumentu pozitivan ili negativan.

Uspoređeni su različiti modeli strojnog učenja koji koriste reprezentacije dokumenata tf-idf te oni koji se služe vektorskim reprezentacijama riječi kako bi predstavili dokumente. Performanse modela mjerene su na četiri različita skupa podataka koja su sastavljena od korisničkih recenzija. Pokazuje se da su na manjim skupovima podataka uspješniji tradicionalni pristupi koji koriste stroj potpornih vektora i reprezentacije tf-idf, dok se na većim skupovima podataka bolje pokazuju konvolucijske neuronske mreže koje koriste vektorske reprezentacije riječi.

Abstract (english)

This paper studies different approaches to binary sentiment analysis. Its task is to determine whether a textual document expresses positive or negative sentiment.

Different machine learning models were compared using tf-idf document representations as well as word embeddings. The model performance was measured using four different datasets compound of user reviews. It is shown that traditional approaches which use support vector machines and tf-idf representations are more successful on smaller datasets. Convolutional neural networks that use word embeddings performed better on larger datasets.

• Keywords: analiza sentimenta; obrada prirodnog jezika; strojno učenje; duboko učenje; stroj potpornih vektora; SVM; neuronske mreže
• Keywords (english): sentiment analysis; natural language processing; machine learning; deep learning; support vector machine; SVM; neural networks
• Language: croatian
• URN:NBN: urn:nbn:hr:168:361308
• 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: 2020-01-07 22:48:40