diplomsko delo
Andjela Rajović (Author), Marko Robnik Šikonja (Mentor)

Abstract

Razumevanje parlamentarnega govora in širših političnih razprav je ključno za razumevanje političnih procesov in odločitev, ki vplivajo na družbo. Naloga obravnava problem strojne identifikacije in analize stališč poslancev in strank do različnih tematik s trirazredno klasifikacijo: ''za'', "proti", in "nevtralno". Analiza zajema primerjave stališč v srbskem parlamentu. Za analizo smo uporabili nabor ročno označenih podatkov, ki vsebuje 1019 učnih primerov. Ovrednotili smo več jezikovnih modelov, kot so XML-RoBERTa, BERTić, POLITICS, YugoGPT in Llama-3.1, ter primerjali njihove rezultate. Analiza potrjuje splošno znanje o političnih strankah in njihovih usmeritvah ter prikazujejo zmogljivost velikih jezikovnih modelov za analizo velikih zbirk besedil.

Keywords

prepoznavanje stališč;veliki jezikovni modeli;parlamentarni govor;univerzitetni študij;diplomske naloge;

Data

Language: Slovenian
Year of publishing:
Typology: 2.11 - Undergraduate Thesis
Organization: UL FRI - Faculty of Computer and Information Science
Publisher: [A. Rajović]
UDC: 004.85:81'322(043.2)
COBISS: 212592131 Link will open in a new window
Views: 143
Downloads: 42
Average score: 0 (0 votes)
Metadata: JSON JSON-RDF JSON-LD TURTLE N-TRIPLES XML RDFA MICRODATA DC-XML DC-RDF RDF

Other data

Secondary language: English
Secondary title: Stance detection in Serbian parliamentary speech
Secondary abstract: Understanding parliamentary discourse and broader political debates is essential for comprehending the political processes and decisions that impact society. The thesis addresses the challenge of machine learning-based identification and analysis of the stances of parliament members and their parties on various topics using a three-class classification: 'for,' 'against,' and 'neutral.' The analysis includes comparisons of stances in the Serbian parliament. We utilized a manually annotated dataset containing 1,019 examples for the analysis. We evaluated several language models, such as XML-RoBERTa, BERTić, POLITICS, YugoGPT, and Llama-3.1, and compared their performance. Our findings confirm the general knowledge of political parties and their orientations, demonstrating the capability of large language models to analyze large datasets.
Secondary keywords: stance detection;large language models;parliament speech;computer and information science;diploma;Računalniško jezikoslovje;Politično govorništvo;Računalništvo;Univerzitetna in visokošolska dela;
Type (COBISS): Bachelor thesis/paper
Study programme: 1000468
Embargo end date (OpenAIRE): 1970-01-01
Thesis comment: Univ. v Ljubljani, Fak. za računalništvo in informatiko
Pages: 1 spletni vir (1 datoteka PDF (57 str.))
ID: 25001832