Why don't people use character-level machine translation?
This work addresses the gap between theoretical claims and practical usage in machine translation, highlighting incremental insights for researchers and practitioners in NLP.
The paper critically assesses character-level machine translation, finding that despite literature claims of comparability, character-level systems fail to match subword-based counterparts in competitive setups, showing no better domain robustness or morphological generalization, though they exhibit robustness to source noise and stable translation quality with larger beam sizes.
We present a literature and empirical survey that critically assesses the state of the art in character-level modeling for machine translation (MT). Despite evidence in the literature that character-level systems are comparable with subword systems, they are virtually never used in competitive setups in WMT competitions. We empirically show that even with recent modeling innovations in character-level natural language processing, character-level MT systems still struggle to match their subword-based counterparts. Character-level MT systems show neither better domain robustness, nor better morphological generalization, despite being often so motivated. However, we are able to show robustness towards source side noise and that translation quality does not degrade with increasing beam size at decoding time.