Machine translation outputs in Spanish–Chinese legal translation: comparisons of different legal genres and domestic vs. non-domestic systems, error analysis and implications for post-editing
Although some studies have discussed the role of machine translation (MT) systems in legal texts, very few have addressed the Spanish–Chinese language pair. This article compares the outputs of Spanish legal texts of different genres translated into Chinese using various MT systems such as DeepL Tr...
Saved in:
| Main Author: | |
|---|---|
| Format: | Article |
| Language: | Aragonese |
| Published: |
Escola d'Administració Pública de Catalunya
2025-06-01
|
| Series: | Revista de Llengua i Dret - Journal of Language and Law |
| Subjects: | |
| Online Access: | https://revistes.eapc.gencat.cat/index.php/rld/article/view/4368 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Although some studies have discussed the role of machine translation (MT) systems in legal texts, very few have addressed the Spanish–Chinese language pair. This article compares the outputs of Spanish legal texts of different genres translated into Chinese using various MT systems such as DeepL Translator, Google Translator, GPT-4, Baidu Translator, Youdao Translator, and Tencent Translator. The aim is to identify whether there is a quality difference between Chinese domestic and non-domestic MT systems and to investigate whether MT raw outputs for legal texts across different genres exhibit distinct types of translation error. Three types of Spanish legal text were selected for analysis: an administrative approval authorisation, a maritime judgement, and a legislative text. The target texts produced by the six MT systems were evaluated using automated metrics as well as human evaluations focusing on adequacy and fluency. Additionally, the present paper follows the framework of the Dynamic Quality Framework (DQF) and the Multidimensional Quality Metrics (MQM) typology for MT, as proposed by Lommel and Melby (2018), to categorise and analyse different error types. Based on this analysis, translation strategies for post-editing (PE) are proposed.
|
|---|---|
| ISSN: | 0212-5056 2013-1453 |