Pretraživanje Java programskog koda predstavljenog dubokim značajkama na osnovi prirodnim jezikom zadane funkcionalnosti

master's thesis

• Title (english): Java Code Deep-Searching Based on the Functionality Description Given in Natural Language
• Author: Mario Pavić
• Mentor: Marin Šilić (mentor)
• Committee member: Marin Šilić (predsjednik povjerenstva)
• Committee member: Siniša Srbljić (član povjerenstva)
• Committee member: Goran Delač (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2019-07-08, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Duboko učenje je grana strojnog učenja čije metode nalaze primjenu u važnim područjima umjetne inteligencije poput računalnog vida, obrade prirodnog teksta i tako dalje. Duboko učenje prilikom pretrage programskih kodova na osnovi prirodnim jezikom zadane funkcije se koristi kako bi se programski kodovi i tekst napisan prirodnim jezikom, a koji predstavlja opis zadane funkcije, pretvorili u iste objekte, to jest smjestili u isti vektorski prostor i time dva različita objekta pretvorio u isti objekt. U ovom radu je uspješno implementiran sustav koji vektorizira programske kodove i vektorizira prirodnim jezikom zadane funkcije te ta dva vektora stavlja u isti vektorski prostor i omogućuje pretragu Java programskih kodova predstavljenog dubokim značajkama na osnovi prirodnim jezikom zadane funkcije.

Abstract (english)

Deep learning is a branch of machine learning whose methods are applied in important areas of artificial intelligence such as computer vision, natural text processing, and so on. Deep learning for searching program codes based on the functionality description given in natural language is used to convert the program codes and text written in the natural language, which represents the functionality description, into the same objects. Basically transforms two different objects in the same object and placing it in the same vector space. In this paper, a successfully implementation of the system that vectorize program codes and vectorize the natural language functionality description is provided. System places these two vectors in the same vector space and allows the search of Java code codes represented by the functionality description given in natural language.

• Keywords: Duboko učenje; Java; Pretraga kodova; Prirodni jezik; PyTorch
• Keywords (english): Deep learning; Java; Natural language; Code search; PytTorch
• Language: croatian
• URN:NBN: urn:nbn:hr:168:133928
• 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 21:01:49