id	author	title	date	pages	extension	mime	words	sentences	flesch	summary	cache	txt
work_u2itmvi23nfwjgk64b2kurj7xy	Jason Lee	Fully Character-Level Neural Machine Translation without Explicit Segmentation	2017	15	.pdf	application/pdf	8051	802	65	Fully character-level neural machine translation without explicit Most existing machine translation systems operate at the level of words, relying on explicit segmentation to extract tokens. For one, they are unable to model rare, out-ofvocabulary words, making them limited in translating languages with rich morphology such as Czech, This demonstrates excellent parameter efficiency of character-level translation in a multilingual setting. empirically show that (1) we can train character-tocharacter NMT model without any explicit segmentation; and (2) we can share a single character-level instance, a character-level model may easily identify morphemes that are shared across different languages. Highway networks are shown to significantly improve the quality of a character-level language model when used and the translations from the subword-level baseline and our character-level model as bpe2char and char2char, respectively. bilingual character-level models can be trained in multilingual many-to-one translation task, where results show that the character-level model can assign	./cache/work_u2itmvi23nfwjgk64b2kurj7xy.pdf	./txt/work_u2itmvi23nfwjgk64b2kurj7xy.txt