Automatizacija procesa konstruiranja domenske ontologije iz polustrukturiranih dokumenata

doctoral thesis

• Title (english): Automating process of domain ontology construction from semistructured documents
• Author: Damir Jurić; MBZ: 267771
• Mentor: Marko Banek (mentor)
• Committee member: Zoran Skočir (predsjednik povjerenstva)
• Committee member: Šandor Dembitz (član povjerenstva)
• Committee member: Mladen Varga (član povjerenstva)
• Committee member: Maja Matijašević (član povjerenstva)
• Committee member: Boris Vrdoljak (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; (Department of Electrical Engineering Fundamentals and Measurements); Zagreb
• Defense date and country: 2011-02-28, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing
• Universal decimal classification (UDC): 004 - Computer science and technology. Computing. Data processing

Abstract

U disertaciji je razvijena metoda za konstruiranje domenske ontologije iz enciklopedijskog teksta. Metoda koristi enciklopediju Wikipediju jer ta enciklopedija predstavlja iscpan i djelomično strukturiran izvor koji je javno dostupan. Prema razvijenoj metodi najprije se izvodi algoritam za ekstrakciju pojmova koji će izgrađivati buduću domensku ontologiju iz članaka Wikipedije. Članci prolaze proces čišćenja i pripreme za obradu teksta te se konstruira njihov sažeti formalni zapis članka. Ekstrahiraju se rečenične strukture koje osim subjekta, predikata i objekta sadrže i objekte prijedložnih oznaka (struktura n‐arnog tipa). Relacije dobivene iz ovog postupka se generaliziraju algoritmom za ekstrakciju hijerarhijskih relacija iz članaka Wikipedije i WordNeta razvijenim za potrebe ove disertacije. U disertaciji je riješen i problem konzistentnog imenovanja relacija. Iskorišteni su semantički resursi poput baze glagola i baze značenja prijedloga da bi se za različite relacije koje postoje između događaja i aktera koji sudjeluju u događajima pronašli uniformni nazivi. Definiran je način zapisivanja ekstrahiranih n‐arnih relacija u standardni jezik za zapisivanje ontologija. Na kraju disertacije izvršena je evaluacija te je pokazano da je točnost ove metode usporediva s drugim metodama, koje su međutim po opsegu i specifičnosti ekstrahiranog znanja slabije.

Abstract (english)

This doctoral thesis proposes a method for automated construction of a domain ontology from encyclopedic text. The Wikipedia encyclopedia is used as source because it contains exhaustively written articles that are partially structured and publicly available. The method developed in this dissertation starts with extracting the concepts from Wikipedia articles that will populate the future domain ontology. The articles undergo a process of cleaning and preparation for grammatical text analysis and are transformed into their formal concise representations. Algorithms extract grammatical structures that apart from subjects, objects and verbs also contain prepositions and prepositional objects (n‐ary structure). Relations obtained from this process are generalized using the algorithm for taxonomy extraction from Wikipedia articles and WordNet, also developed in this dissertation. This dissertation also proposes a method for automated relation labeling using semantic resources, a verb database and a database of preposition meanings, in order to uniformly label relations that exist between events and the participants involved in those events. The extracted n‐ary relations are written in a standard ontology language. Evaluation results show that the method developed in this thesis has precision comparable with other methods, which are not so rich from the aspect of scope and specificity.

• Keywords: konstruiranje ontologija; domenske ontologije; ekstrakcija relacija; ekstrakcija taksonomije; enciklopedija Wikipedija; jezik OWL
• Keywords (english): ontology construction; domain ontologies; relation extraction; taxonomy extraction; Wikipedia; OWL
• Language: croatian
• URN:NBN: urn:nbn:hr:168:223158
• Project: Number: 036-0362027-1638; Title: Umrežena ekonomija; Leader: Zoran Skočir; Jurisdiction: Croatia; Funder: MZOS; Funding stream: ZP
• Study programme: Title: Computer Science; Study programme type: university; Study level: postgraduate; Academic / professional title: Doktor znanosti (Doktor znanosti)
• Catalog URL: http://lib.fer.hr/cgi-bin/koha/opac-detail.pl?biblionumber=34371
• Type of resource: Text
• Extent: 158 str. ; 30 cm
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2019-06-07 08:20:41