Model za rerangiranje za hrvatski jezik temeljen na kontekstualiziranim višejezičnim jezičnim modelima

master's thesis

• Title (english): Learning-to-Rank Model for Croatian based on Contextualized Multilingual Language Models
• Author: Bruno Šarlija
• Mentor: Jan Šnajder (mentor)
• Committee member: Jan Šnajder (predsjednik povjerenstva)
• Committee member: Vlado Glavinić (član povjerenstva)
• Committee member: Ante Đerek (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2019-07-11, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Klasičan primjer uporabe modela za dohvat informacija je pronalazak odgovarajućih dokumenata za postavljen upit. U ovom radu implementirani su modeli za rerangiranje dokumenata iz skupa podataka na hrvatskom jeziku temeljeni na modelima za dohvat informacija. Također, implementirana je kompozicija prethodno spomenutih modela za rerangiranje i jezičnih modela. Dvije vrste korištenih jezičnih modela su kontekstualizirani jezični model i kontekstualizirani višejezični jezični model. Rezultati implementiranih modela evaluirani su srednjim recipročnim rangom i srednjom prosječnom preciznošću. Ovi rezultati uspoređeni su s rezultatima dobivenima tradicionalnim modelima za dohvat informacija.

Abstract (english)

Finding the appropriate documents for the requested query is the most common usage of information retrieval models. In this thesis, learning to rank models based on information retrieval models were implemented and used for ranking documents written in the Croatian language. Also, the composition of the previously mentioned learning to rank models and language models were implemented. The two types of language models were used are a contextualized language model and a contextualized multilingual language model. The results of the implemented models were evaluated by mean reciprocal rank and mean average accuracy. These results were compared with the results obtained with traditional information retrieval models.

• Keywords: dohvat informacija; model za rerangiranje; kontekstualizirani višejezični jezični model
• Keywords (english): information retrieval; learning to rank model; contextualized multilingual language model
• Language: croatian
• URN:NBN: urn:nbn:hr:168:277033
• 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: 2020-01-07 22:17:06