Chinese character decomposition for neural MT with multi-word expressions
Han, LifengORCID: 0000-0002-3221-2185, Jones, Gareth J.F.ORCID: 0000-0003-2923-8365, Smeaton, Alan F.ORCID: 0000-0003-1028-8389 and Bolzoni, Paolo
(2021)
Chinese character decomposition for neural MT with multi-word expressions.
In: 23rd Nordic Conference on Computational Linguistics (NoDaLiDa 2021), 31 May- 2 June 2021, Reykjavik, Iceland (Online).
Chinese character decomposition has been used as a feature to enhance Machine Translation (MT) models, combining rad- icals into character and word level mod- els. Recent work has investigated ideo- graph or stroke level embedding. How- ever, questions remain about different de- composition levels of Chinese character representations, radical and strokes, best suited for MT. To investigate the impact of Chinese decomposition embedding in detail, i.e., radical, stroke, and intermedi- ate levels, and how well these decomposi- tions represent the meaning of the original character sequences, we carry out analy- sis with both automated and human evalu- ation of MT. Furthermore, we investigate if the combination of decomposed Mul- tiword Expressions (MWEs) can enhance the model learning. MWE integration into MT has seen more than a decade of explo- ration. However, decomposed MWEs has not previously been explored.
Proceedings of the 23rd Nordic Conference on Computational Linguistics (NoDaLiDa).
.
Linköping University Electronic Press, Sweden via Association for Computational Linguistics (ACL).
Publisher:
Linköping University Electronic Press, Sweden via Association for Computational Linguistics (ACL)
Science Foundation Ireland SFI Research Centres Programme (Grant 13/RC/2106) and co-funded under the European Regional Development Fund., Science Foundation Ireland under grant number SFI/12/RC/2289 (Insight Centre)
ID Code:
25742
Deposited On:
30 Apr 2021 15:18 by
Lifeng Han
. Last Modified 05 Jan 2022 17:24