Metode računalne i forenzičke lingvistike za utvrđivanje autorstva

master's thesis

• Title (english): Computational and Forensic Linguistics Methods for Authorship Attribution
• Author: Marijana Šincek
• Mentor: Božo Bekavac (mentor)
• Committee member: Ivana Simeon (predsjednik povjerenstva)
• Committee member: Matea Filko (član povjerenstva)
• Granter: University of Zagreb; Faculty of Humanities and Social Sciences; (Department of Linguistics); Zagreb
• Defense date and country: 2022, Croatia
• Scientific / art field, discipline and subdiscipline: HUMANISTIC SCIENCES; Philology; General Linguistics

Abstract

Računalna je lingvistika grana lingvistike koja se od ostalih lingvističkih grana i metodologija razlikuje upotrebom računala, koja se mogu koristiti za opis jezičnih jedinica na raznim razinama jezika (Tadić, 2003). Forenzička lingvistika predstavlja primjenu lingvistike u domeni prava i pravnih pitanja, gdje se lingvistička znanja i vještine koriste za analizu pravnih tekstova, pravnog govora, ali i za pružanje dokaza istražiteljima (Coulthard i sur., 2017). Cilj istraživanja je ispitati mogućnosti povezivanja metoda računalne i forenzičke lingvistike te proučiti kako se one mogu koristiti za utvrđivanje autorstva tekstova koje je napisao Antun Gustav Matoš, kao i provjeriti koje metode i analize pružaju najindikativnije rezultate. U sklopu istraživanja analizirano je 26 tekstova pet autora, Antuna Gustava Matoša, Milana Begovića, Jure Turića, Josipa Kozarca i Joze Ivakića. Provedena je analiza stilometrijskih značajki teksta svih pet autora korištenjem analize varijance i Jaccardovog koeficijenta sličnosti pomoću korpusa i korpusnih alata. Rezultati istraživanja pokazali su da najlošije rezultate za utvrđivanje autorstva pružaju stilometrijske mjere dužine riječi i rečenica u tekstu, dok omjer različnica i pojavnica, koji pruža podatke o leksičkoj obogaćenosti teksta, ima veću uspješnost. Najveću uspješnost ostvarilo je mjerenje Jaccardovog koeficijenta sličnosti lema i n-grama u poznatim i provjeravanim tekstovima.

Abstract (english)

Computational linguistics is a branch of linguistics differentiated from other linguistic branches and methodologies by the use of computers, used for description of linguistic units on all levels of language (Tadić, 2003). Forensic linguistics denotes the application of linguistics when it comes to the law and legal matters, where linguistic knowledge and skills are used for analysis of written legal texts, legal language, as well as to provide evidence to investigators (Coulthard et al., 2017). The aim of this research was to examine the possibilities of combining the methods of computational and forensic linguistics, how they can be used for authorship attribution of texts written by Antun Gustav Matoš, as well as to explore which methods and analyses provide most indicative results. 26 texts from five authors, Antun Gustav Matoš, Milan Begović, Jure Turić, Josip Kozarac and Joza Ivakić were included in this research. An analysis of stylometric properties of the texts by all five authors was conducted using analysis of variance and Jaccard similarity index by using corpora and corpus tools. The results of the research showed that the worst results for authorship attribution were obtained using word and sentence length, followed by a more successful type-token ratio, which provided data on the lexical richness of the texts. The most successful was the Jaccard similarity index of the lemmas and n-grams used in known and questioned text samples.

• Keywords: računalna lingvistika; forenzička lingvistika; metodologija; korpus; utvrđivanje autorstva
• Keywords (english): computational linguistics; forensic linguistics; methodology; corpus; authorship attribution
• Language: croatian
• URN:NBN: urn:nbn:hr:131:137692
• Study programme: Title: Linguistics (double major); specializations in: General Linguistics, Comparative Linguistics, Computational Linguistics, Aplied Linguistics and Cognitive Linguistics, Cognitive; Course: General Linguistics; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra lingvistike (magistar/magistra lingvistike)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Open access
• Created on: 2022-07-13 17:10:24