Vol. 35:1 (2023) ► pp.1–33
How do translators select among competing (near-)synonyms in translation?
A corpus-based approach using random forest modelling
This article investigates how translators choose between multiple competing onomasiological variants to express (verbal) inchoativity in English-to-Dutch translations. Using a corpus-based multifactorial research design, we measure the impact of three well-known socio-cognitive mechanisms on the actual choice, namely the complexity principle, risk aversion, and cognate exposure. We apply the behavioural profile method, which allows us to operationalise these three explanatory mechanisms via ID-tags, and we then use conditional random forest modelling to determine the impact of each mechanism on the choice between four competing verbs of inchoativity. The results of our analyses show that the complexity principle plays a clear role in translated texts, as there is a significant preference for the active construction and for prototypical verbs in passive constructions. Genre-specific risk-averse behaviour as well as cognate avoidance were not observed.
Article outline
- 1.Introduction
- 2.Previous research: Near-synonymy and socio-cognitive explanations
- 2.1Complexity
- 2.2Risk aversion
- 2.3Cognate exposure
- 2.4Summing up
- 3.Methodology
- 3.1Lexeme selection
- 3.2Behavioural profiles
- 3.2.1Complexity
- 3.2.2Risk aversion
- 3.2.3Cognate exposure
- 3.3Statistical analysis
- 4.Results
- 4.1Random forest models for non-translated and translated Dutch
- 4.2In-depth analysis of the most influential features
- 4.2.1Voice
- 4.2.2Genre
- 4.2.3Cognate exposure
- 5.Discussion and conclusion
- Acknowledgements
- Notes
-
References
https://doi.org/10.1075/target.20095.deb