diplomsko delo
Martin Božič (Author), Marko Robnik Šikonja (Mentor)

Abstract

Med najpogostejše napake pri pisanju besedil v slovenščini sodi postavljanje vejic. V diplomski nalogi se bomo osredotočili na postavljanje vejic s pomočjo globokih nevronskih mrež. Predstavili bomo dve arhitekturi, eno na podlagi nevronskih mrež s celicami GRU in drugo z vnaprej naučenim jezikovnim modelom tipa BERT. Pri uporabi jezikovnega modela tipa BERT opazimo boljšo klasifikacijsko točnost. Vzrok za to je boljša in kompleksnejša arhitektura modela ter proces učenja, ki izpopolnjuje model z obširnim jezikovnim znanjem. Z uporabo večjezičnega modela BERT, naučenega na 104 jezikih in le manjšo množico slovenskih besedil, pridobimo rešitev, ki je primerljiva z rešitvijo, ki smo jo pridobili z uporabo trojezičnega, slovensko-hrvaško-angleškega modela BERT.

Keywords

globoke nevronske mreže;mreže GRU;model BERT;postavljanje vejic;računalništvo in informatika;univerzitetni študij;diplomske naloge;

Data

Language: Slovenian
Year of publishing:
Typology: 2.11 - Undergraduate Thesis
Organization: UL FRI - Faculty of Computer and Information Science
Publisher: [M. Božič]
UDC: 004.85:003.086(043.2)
COBISS: 27670787 Link will open in a new window
Views: 1060
Downloads: 485
Average score: 0 (0 votes)
Metadata: JSON JSON-RDF JSON-LD TURTLE N-TRIPLES XML RDFA MICRODATA DC-XML DC-RDF RDF

Other data

Secondary language: English
Secondary title: Deep neural networks for comma placement in Slovene
Secondary abstract: Comma placement is the most frequent orthological mistake in Slovene. The thesis focuses on comma placement using deep neural networks. We present two architectures, one based on neural networks with GRU cells and another using a pre-learned BERT language model. Using a pre-learned BERT language model, we get better classification accuracy. The reason for this is better and more complex architecture and the learning process, which fine-tuned a pretrained model with substantial language knowladge. With the multilingual BERT, trained on 104 languages with only a small amount of Slovene texts, we achieve comparable results to Slovene-Croatian-English BERT model, trained with much more Slovene texts.
Secondary keywords: deep neural networks;GRU networks;BERT model;comma placement;computer and information science;diploma;
Type (COBISS): Bachelor thesis/paper
Study programme: 1000468
Embargo end date (OpenAIRE): 1970-01-01
Thesis comment: Univ. v Ljubljani, Fak. za računalništvo in informatiko
Pages: 25 str.
ID: 12021428