Title: Syllable-to-Syllable and Word-to-Word Transducers for Burmese Dialect Translation
Abstract: Weighted Finite State Transducers (WFST) can be very efficient to implement Burmese dialects translation. We illustrate this on two Burmese dialect language pairs, Burmese-Beik and Burmese-Rakhine. In this study, we examine syllable and word segmentation schemes and their effect on alignment and transducing between dialect language pairs. We performed alignments with Anymalign, fastalign, pialign, Hieralign, eflomal and GIZA ++ approaches and implemented WFST based machine translation system with OpenFst library. From the overall results, syllable segmentation achieved higher BLEU and chrF scores for Burmese-Rakhine and Rakhine-Burmese translations. However, word segmentation achieved better translation performance for Burmese-Beik and Beik-Burmese translation directions. Alignment techniques fast align, Hieralign, eflomal and GIZA ++ are working well for low-resource Burmese dialects.
Publication Year: 2022
Publication Date: 2022-11-05
Language: en
Type: article
Indexed In: ['crossref']
Access and Citation
AI Researcher Chatbot
Get quick answers to your questions about the article from our AI researcher chatbot