prikaz prve stranice dokumenta Strojno učenje za analizu sentimenta temeljenu na aspektima
No public access
undergraduate thesis
Strojno učenje za analizu sentimenta temeljenu na aspektima
Zagreb: University of Zagreb, Faculty of Electrical Engineering and Computing, 2020. urn:nbn:hr:168:693800
Zec, Mario
University of Zagreb
Faculty of Electrical Engineering and Computing

Institutional repository: FER Repository

Cite this document

Zec, M. (2020). Strojno učenje za analizu sentimenta temeljenu na aspektima (Undergraduate thesis). Zagreb: University of Zagreb, Faculty of Electrical Engineering and Computing. Retrieved from https://urn.nsk.hr/urn:nbn:hr:168:693800

Zec, Mario. "Strojno učenje za analizu sentimenta temeljenu na aspektima." Undergraduate thesis, University of Zagreb, Faculty of Electrical Engineering and Computing, 2020. https://urn.nsk.hr/urn:nbn:hr:168:693800

Zec, Mario. "Strojno učenje za analizu sentimenta temeljenu na aspektima." Undergraduate thesis, University of Zagreb, Faculty of Electrical Engineering and Computing, 2020. https://urn.nsk.hr/urn:nbn:hr:168:693800

Zec, M. (2020). 'Strojno učenje za analizu sentimenta temeljenu na aspektima', Undergraduate thesis, University of Zagreb, Faculty of Electrical Engineering and Computing, accessed 18 November 2024, https://urn.nsk.hr/urn:nbn:hr:168:693800

Zec M. Strojno učenje za analizu sentimenta temeljenu na aspektima [Undergraduate thesis]. Zagreb: University of Zagreb, Faculty of Electrical Engineering and Computing; 2020 [cited 2024 November 18] Available at: https://urn.nsk.hr/urn:nbn:hr:168:693800

M. Zec, "Strojno učenje za analizu sentimenta temeljenu na aspektima", Undergraduate thesis, University of Zagreb, Faculty of Electrical Engineering and Computing, Zagreb, 2020. Available at: https://urn.nsk.hr/urn:nbn:hr:168:693800

Metadata
TitleStrojno učenje za analizu sentimenta temeljenu na aspektima
Title (english)Machine Learning for Aspect-Based Sentiment Analysis
AuthorMario Zec
MentorJan Šnajder (mentor)
Committee memberJan Šnajder (predsjednik povjerenstva)
Committee memberDomagoj Alagić (član povjerenstva)
Committee memberZoran Medić (član povjerenstva)
GranterUniversity of Zagreb
Faculty of Electrical Engineering and Computing
Zagreb
Defense date and country2020-07-06, Croatia
Scientific / art field,
discipline and subdiscipline		TECHNICAL SCIENCES
Computing
Abstract
Tema ovoga rada je analiza sentimenta temeljena na aspektima. Korišten je skup podataka s natjecanja Semeval iz 2015. godine za zadatak 12. Prva tri modela ovoga rada omogućila su nam bolji uvid u sami problem i dodatne informacije o skupu podataka. U njma su isprobani pristupi koji koriste koeficijent Sørensen-Dice, te jednoslojni i dvoslojni modeli modela CRF. Konačno rješenje temeljili smo na pretpostavci da svakome aspektu mora biti pripisan jedan dio rečenice. Ako to nije slučaj, onda dolazimo do problema s označavanjem skupa podataka jer ako jedna jednostavna rečenica sadrži dva različita aspekta onda nije jasno koji aspekt se odnosi na koji dio rečenice. Analizom skupa podataka i algoritamskim rastavljanjem rečenica uvjerili smo se da je moguće samo veznicima i interpunkcijskim znakovima rastvljati rečenice s uspješnošću od 98%, gdje broj dijelova rastavljene rečenice odgovara broju aspekata iz skupa podataka. Temeljem ove ideje izgrađen je model koji je rastavljao rečenice algoritamski te koristio strojno učenje za predviđanje aspekata na temelju dijelova rečenica. Uspješnost modela je zadovoljavajuća, ali ostaje velik prostor za poboljšanje modela.
Abstract (english)
This thesis is about aspect-based sentiment analysis (ABSA). The dataset of Semeval 2015 for task 12 was used. First three models gave us better view of the task and additional informations about dataset. Approaches that were used are: Sørensen-Dice coefficient and one and two layers CRF models. The fourth model was based on an assumption that every aspect belongs to one part of the sentence. If that is not true then we come to the problem that annotations are fuzzy. Based on dataset analysis and algorithmic partitioning of sentences we get that we can partition 98% of sentences correct using conjunctions and punctuation marks so that number of aspects is equal to the number of sentence parts. Based on this idea we build a model that algorithmically partitions sentences and then using machine learning predicts aspects based on sentence parts. Success of this model is satisfactory but there is room for improvements.
Keywords
Keywords (english)
Languagecroatian
URN:NBNurn:nbn:hr:168:693800
Study programmeTitle: Computing
Study programme type: university
Study level: undergraduate
Academic / professional title: sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva (sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva)
Type of resourceText
File originBorn digital
Access conditionsClosed access
Terms of useLicence In copyright icon
Public note
Created on2023-01-14 10:18:55