Analysis of Transfer Learning for Named Entity Recognition in South-Slavic Languages
Nikola Ivačič, Thi Hong Hanh Tran, Boshko Koloski, Senja Pollak, Matthew Purver · 2023
This paper focuses on Named Entity Recognition for South-Slavic languages using pretrained multilingual neural network models.We investigate whether the performance of the models for a target language can be improved by using data from closely related languages.The results show that this is not the case for the Slovene language, while for Croatian and Serbian, the results are better in selected crosslingual settings.The most significant performance improvement is observed for the Serbian language, which has the smallest corpora, showing the potential of the method in lessresourced settings.