Postupci odabira značajki i prikaza dokumenta za klasifikaciju teksta

master's thesis

• Title (english): Feature Selection and Document Representation Methods for Text Classification
• Author: Mihael Šafarić
• Mentor: Jan Šnajder (mentor)
• Committee member: Jan Šnajder (predsjednik povjerenstva)
• Committee member: Slobodan Ribarić (član povjerenstva)
• Committee member: Siniša Šegvić (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2015-07-14, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

S porastom količine digitalnih informacija raste i potreba za metodama klasifikacije teksta koje su učinkovitinije i u mogućnosti obraditi velike količine podataka. Iz tog razloga pozornost se sve više okreće metodama temeljenim na strojnom učenju. U ovom radu isprobane su metode koje koriste reprezentaciju dokumenta kao vreće riječi koje u prvom koraku rade odabir najznačajnijih značajki. Uz to, isprobane su i metode temeljene na neuronskim mrežama koje za pojedini dokument grade njegovu vektorsku reprezentaciju.Korištenjem tih metoda ostvarili su se bolji rezultati nego korištenjem metoda koje koriste reprezentaciju dokumenta kao vreće riječi. Svi eksperimenti provedeni su nad različitim zbirkama tekstova na hrvatskom i engleskom jeziku.

Abstract (english)

With growing amount of online information, there is a growing need for text classification methods that are more efficient and capable to process large amount of data. Therefore, there is increased attention to the methods based on machine learning. This paper experiments with methods that use bag-of-words document representation and feature selection methods. In addition, this paper experiments with a neural network based methods. These methods build vector representation of each document and the results achieved with these methods are better than the results achieved using methods that use bag-of-words document representation. All of the experiments are performed over a few different document collections in Croatian and English.

• Keywords: strojno učenje; obrada prirodnog jezika; klasifikacija teksta; odabir značajki; word2vec; doc2vec; reprezentacija dokumenta
• Keywords (english): machine learning; natural language processing; text classification; feature selection; word2vec; doc2vec; document representation
• Language: croatian
• URN:NBN: urn:nbn:hr:168:755362
• 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: 2018-02-06 12:10:06