id	author	title	date	pages	extension	mime	words	sentences	flesch	summary	cache	txt
work_nw3etn3etra45jwazg6txw2oly	Anoop Kunchukuttan	Leveraging Orthographic Similarity for Multilingual Neural Transliteration	2018	14	.pdf	application/pdf	8454	836	52	We address the task of joint training of transliteration models for multiple language pairs encoder-decoder model that maximizes parameter sharing across language pairs in order training multiple language pairs (referred to as multilingual transliteration henceforth). Multilingual transliteration can be seen as an instance of multi-task learning, where training each In this work, we explore multilingual transliteration involving orthographically similar languages. We propose that transliteration involving orthographically similar languages is a scenario where multilingual training can be very beneficial. output layer for target languages, but share the encoder, decoder and character embeddings across languages. The multilingual transliteration task involves learning transliteration models for l language pairs across all language pairs used to train the multilingual model is equivalent to the size of the bilingual target languages are covered, we use the trained multilingual model discussed in previous sections for the multilingual transliteration models can generalize well to language pairs not encountered during	./cache/work_nw3etn3etra45jwazg6txw2oly.pdf	./txt/work_nw3etn3etra45jwazg6txw2oly.txt