Adapting an English Corpus and a Question Answering System for Slovene

Developing effective question answering (QA) models for less-resourced languages like Slovene is challenging due to the lack of proper training data. Modern machine translation tools can address this issue, but this presents another challenge: the given answers must be found in their exact form with...

Whakaahuatanga katoa

I tiakina i:
Ngā taipitopito rārangi puna kōrero
Ngā kaituhi matua: Uroš Šmajdek, Matjaž Zupanič, Maj Zirkelbach, Meta Jazbinšek
Hōputu: Tuhinga
Reo:Ingarihi
I whakaputaina: University of Ljubljana Press (Založba Univerze v Ljubljani) 2023-09-01
Rangatū:Slovenščina 2.0: Empirične, aplikativne in interdisciplinarne raziskave
Ngā marau:
Urunga tuihono:https://journals.uni-lj.si/slovenscina2/article/view/12064
Tags: Tāpirihia he Tūtohu
No Tags, Be the first to tag this record!
_version_ 1839629941627420672
author Uroš Šmajdek
Matjaž Zupanič
Maj Zirkelbach
Meta Jazbinšek
author_facet Uroš Šmajdek
Matjaž Zupanič
Maj Zirkelbach
Meta Jazbinšek
author_sort Uroš Šmajdek
collection DOAJ
description Developing effective question answering (QA) models for less-resourced languages like Slovene is challenging due to the lack of proper training data. Modern machine translation tools can address this issue, but this presents another challenge: the given answers must be found in their exact form within the given context since the model is trained to locate answers and not generate them. To address this challenge, we propose a method that embeds the answers within the context before translation and evaluate its effectiveness on the SQuAD 2.0 dataset translated using both eTranslation and Google Cloud translator. The results show that by employing our method we can reduce the rate at which answers were not found in the context from 56% to 7%. We then assess the translated datasets using various transformer-based QA models, examining the differences between the datasets and model configurations. To ensure that our models produce realistic results, we test them on a small subset of the original data that was human-translated. The results indicate that the primary advantages of using machine-translated data lie in refining smaller multilingual and monolingual models. For instance, the multilingual CroSloEngual BERT model fine-tuned and tested on Slovene data achieved nearly equivalent performance to one fine-tuned and tested on English data, with 70.2% and 73.3% questions answered, respectively. While larger models, such as RemBERT, achieved comparable results, correctly answering questions in 77.9% of cases when fine-tuned and tested on Slovene compared to 81.1% on English, fine-tuning with English and testing with Slovene data also yielded similar performance.
format Article
id doaj-art-e82081c8122a4ecabc00a6fd7a167b2a
institution Matheson Library
issn 2335-2736
language English
publishDate 2023-09-01
publisher University of Ljubljana Press (Založba Univerze v Ljubljani)
record_format Article
series Slovenščina 2.0: Empirične, aplikativne in interdisciplinarne raziskave
spelling doaj-art-e82081c8122a4ecabc00a6fd7a167b2a2025-07-14T12:57:03ZengUniversity of Ljubljana Press (Založba Univerze v Ljubljani)Slovenščina 2.0: Empirične, aplikativne in interdisciplinarne raziskave2335-27362023-09-0111110.4312/slo2.0.2023.1.247-27418451Adapting an English Corpus and a Question Answering System for SloveneUroš Šmajdek0Matjaž Zupanič1Maj Zirkelbach2Meta Jazbinšek3University of Ljubljana, Faculty of Computer and Information Science, SloveniaUniversity of Ljubljana, Faculty of Computer and Information Science, SloveniaUniversity of Ljubljana, Faculty of Computer and Information Science, SloveniaUniversity of Ljubljana, Faculty of Arts, SloveniaDeveloping effective question answering (QA) models for less-resourced languages like Slovene is challenging due to the lack of proper training data. Modern machine translation tools can address this issue, but this presents another challenge: the given answers must be found in their exact form within the given context since the model is trained to locate answers and not generate them. To address this challenge, we propose a method that embeds the answers within the context before translation and evaluate its effectiveness on the SQuAD 2.0 dataset translated using both eTranslation and Google Cloud translator. The results show that by employing our method we can reduce the rate at which answers were not found in the context from 56% to 7%. We then assess the translated datasets using various transformer-based QA models, examining the differences between the datasets and model configurations. To ensure that our models produce realistic results, we test them on a small subset of the original data that was human-translated. The results indicate that the primary advantages of using machine-translated data lie in refining smaller multilingual and monolingual models. For instance, the multilingual CroSloEngual BERT model fine-tuned and tested on Slovene data achieved nearly equivalent performance to one fine-tuned and tested on English data, with 70.2% and 73.3% questions answered, respectively. While larger models, such as RemBERT, achieved comparable results, correctly answering questions in 77.9% of cases when fine-tuned and tested on Slovene compared to 81.1% on English, fine-tuning with English and testing with Slovene data also yielded similar performance. https://journals.uni-lj.si/slovenscina2/article/view/12064question answeringmachine translationmultilingual models
spellingShingle Uroš Šmajdek
Matjaž Zupanič
Maj Zirkelbach
Meta Jazbinšek
Adapting an English Corpus and a Question Answering System for Slovene
Slovenščina 2.0: Empirične, aplikativne in interdisciplinarne raziskave
question answering
machine translation
multilingual models
title Adapting an English Corpus and a Question Answering System for Slovene
title_full Adapting an English Corpus and a Question Answering System for Slovene
title_fullStr Adapting an English Corpus and a Question Answering System for Slovene
title_full_unstemmed Adapting an English Corpus and a Question Answering System for Slovene
title_short Adapting an English Corpus and a Question Answering System for Slovene
title_sort adapting an english corpus and a question answering system for slovene
topic question answering
machine translation
multilingual models
url https://journals.uni-lj.si/slovenscina2/article/view/12064
work_keys_str_mv AT urossmajdek adaptinganenglishcorpusandaquestionansweringsystemforslovene
AT matjazzupanic adaptinganenglishcorpusandaquestionansweringsystemforslovene
AT majzirkelbach adaptinganenglishcorpusandaquestionansweringsystemforslovene
AT metajazbinsek adaptinganenglishcorpusandaquestionansweringsystemforslovene