Sustav za automatizirani pregled literature

master's thesis

• Title (english): System for Automated Literature Review
• Author: Filip Bacinger
• Mentor: Ivica Botički (mentor)
• Committee member: Ivica Botički (predsjednik povjerenstva)
• Committee member: Tomislav Jagušt (član povjerenstva)
• Committee member: Danijel Mlinarić (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2022-07-14, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Sustavni pregled literature je važan oblik znanstvenog rada temeljen na pretraživanju velikog obujma literature, što čini proces vremenski intenzivnim. Ubrzavanje izrade pregleda literature pomoću automatiziranih sustava moglo bi doprinijeti bržem razvoju novog istraživanja.
U radu je izrađena web-aplikacija koja omogućuje definiranje teme pregleda literature, automatsko pretraživanje izvora znanstvenih radova, automatsko označavanje radova (uz minimalan broj ručno označenih radova), pregled radova i izvoz rezultata u obliku dijagrama i svih pozitivno označenih radova.
Za obradu teksta korištene su metode tokenizacije, uklanjanja stop-riječi, vađenja korijena (Lancaster algoritam) i računanja TF-IDF faktora. Obrađeni tekstovi koriste se u kombiniranom modelu strojnog učenja izrađenom od dvadeset i četiri (24) pojedinačna modela. Kombinirani model je podešen da prioritizira točnu klasifikaciju pozitivnih primjera i ostvaruje ukupnu točnost preko 80\%.

Abstract (english)

Systematic literature review is an important form of scientific work based on reviewing a large volume of literature, which makes the process time-intensive. Speeding up the creation of literature reviews with the help of automated systems could contribute to a faster development of new research.
In this thesis, a web application is created which allows defining a subject for the literature review, automated querying of scientific article sources, automated article labelling (with a minimal number of manually labelled articles), viewing articles and exporting results in the form of a diagram and all positively labelled articles.
Methods of tokenisation, stop-word removal, stemming (Lancaster algorithm) and calculating TF-IDF factors were used for text processing. Processed texts are used in a combined model of machine learning which comprises of twenty-four (24) individual models. The combined model is adjusted to prioritize correctly classifying positive articles and has an overall accuracy of over 80\%.

• Keywords: pregled literature; web-aplikacija; automatizacija; obrada prirodnog jezika; strojno učenje; klasifikacija
• Keywords (english): literature review; web application; automatization; natural language processing; machine learning; classification
• Language: croatian
• URN:NBN: urn:nbn:hr:168:206769
• 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: 2023-01-14 08:21:07