Multi-word units in neural machine translation. Why the tip of the iceberg remains problematic
(2024) Recent Advances in Multiword Units in Machine Translation and Translation Technology — ISBN: [9789027217905], p. 2-17, published
Files
AbNMT.pdf
Restricted Access - Adobe PDF
- 404.72 KB
Details
- Authors
- Abstract
- Neural machine translation (NMT) has recently made significant progress in improving the quality of the texts it produces. New features of NMT include the fluidity of translations and the successful handling of multi-word units. In this paper we first report the results of an automated evaluation of the percentage of phraseology in the translations produced by Google Translate and DeepL. A corpus-based approach makes it possible to estimate that both NMT systems succeed in producing an average percentage of phraseology that is quite reasonable and sometimes even higher than in natural language production by native speakers. However, a closer look at some problematic cases shows that the ability of NMT systems to treat phraseological units can be deceptive, as they are often unable to cope with contextual complexity and low-frequency idioms.
- Affiliations
Citations
Colson, J.-P. (2024). Multi-word units in neural machine translation. Why the tip of the iceberg remains problematic. In Johanna Monti, Gloria Corpas Pastor, Ruslan Mitkov & Carlos Manuel Hidalgo-Ternero (Eds.) (ed.), Recent Advances in Multiword Units in Machine Translation and Translation Technology (p. p. 2-17). John Benjamins. https://doi.org/10.1075/cilt.366