Because sign languages are the first language for those who are born deaf or who lost their hearing in early childhood, it is better to use sign languages rather than transcribed spoken language to provide important information to these people. We have been developing a sign language computer graphics generation system to provide information to deaf people, and in this paper, we present a translation method from spoken language to sign language that can be used in the system. In general, since the number of glosses used when transcribing sign language is limited, a single meaning is often expressed by a combination of multiple sign words, i.e., the word "library" is expressed in Japanese Sign Language with two words: "book" and "building." To merge these expressions into one token, we propose gloss pair encoding (GPE), which is inspired by bite pair encoding (BPE). This technique is expected to enable more accurate handling of expressions that have a single meaning in multiple sign words. We also show that it is effective as data augmentation on the sign language side in sign language translation, which has not been done much so far.
@inproceedings{miyazaki:24004:sign-lang:lrec,
author = {Miyazaki, Taro and Tan, Sihan and Uchida, Tsubasa and Kaneko, Hiroyuki},
title = {Sign Language Translation with Gloss Pair Encoding},
pages = {32--38},
editor = {Efthimiou, Eleni and Fotinea, Stavroula-Evita and Hanke, Thomas and Hochgesang, Julie A. and Mesch, Johanna and Schulder, Marc},
booktitle = {Proceedings of the {LREC-COLING} 2024 11th Workshop on the Representation and Processing of Sign Languages: Evaluation of Sign Language Resources},
maintitle = {2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation ({LREC-COLING} 2024)},
publisher = {{ELRA Language Resources Association (ELRA) and the International Committee on Computational Linguistics (ICCL)}},
address = {Torino, Italy},
day = {25},
month = may,
year = {2024},
isbn = {978-2-493814-30-2},
language = {english},
url = {https://www.sign-lang.uni-hamburg.de/lrec/pub/24004.pdf}
}
| Tan, Sihan