Model strojnog prevođenja temeljen na konvolucijskim neuronskim mrežama

master's thesis

• Title (english): Machine Translation Model Based on Convolutional Neural Networks
• Author: Mihael Nikić
• Mentor: Jan Šnajder (mentor)
• Committee member: Jan Šnajder (predsjednik povjerenstva)
• Committee member: Marko Čupić (član povjerenstva)
• Committee member: Bojana Dalbelo-Bašić (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2018-07-17, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Strojno prevođenje jedan je od središnjih i komercijalno najisplativijih zadataka obrade prirodnog jezika. Najnoviji modeli temeljeni na dubokom strojnom učenju, odnosno modeli neuronskog strojnog prevođenja, doveli su do značajnih napredaka u točnosti sustava strojnog prevođenja. Iako se modeli neuronskog strojnog prevođenja obično temelje na povratnim neuronskim mrežama, u okviru ovog rada izgrađen je model neuronskog strojnog prevođenja koji se u potpunosti temelji na konvolucijskim neuronskim mrežama. Nad izgrađenim modelom provedeno je eksperimentalno vrednovanje postupaka te je obavljena analiza dobivenih rezultata.

Abstract (english)

Machine translation is one of the central and most commercially successful tasks of natural language processing. The latest models based on deep machine learning, i.e., neural machine translation models, have led to significant advances in the accuracy of the machine translation system. While neural machine translation models are usually based on recurrent neural networks, in this thesis, a model of neural machine translation was built that is entirely based on convolutional neural networks. The constructed model underwent an experimental process evaluation and an analysis of the obtained results was performed.

• Keywords: strojno prevođenje; neuronsko strojno prevođenje; konvolucijske neuronske mreže; povratne neuronske mreže; propusno linearne ćelije; rezidualne veze; mehanizam pažnje; slijed-u-slijed modeli; čitač-pisač struktura
• Keywords (english): machine translation; neural machine translation; convolutional neural networks; recurrent neural networks; gated linear units; residual connections; attention mechanism; sequence-to-sequence models; encoder-decoder structure
• Language: croatian
• URN:NBN: urn:nbn:hr:168:860784
• 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: 2019-03-13 23:23:28