Article type
Year
Abstract
Background:
Health in my Language (HimL) was an EU-funded, three-year project. It aimed to address the need for reliable and affordable translation of public health content into different languages via fully automatic machine-translation (MT) systems, focusing on translation from English into Czech, German, Polish, and Romanian. Recent advances in MT were used to produce health adapted MT engines. Health information from Cochrane and NHS24 served as the test cases to evaluate the obtained translations.
Objectives:
To show the effectiveness and translator acceptability of HimL MT when used as part of a post-editing workflow to translate Cochrane Plain Language Summaries (PLS), compared to Cochrane's standard translation workflow.
Methods:
10 randomly selected PLS were translated twice into all HimL languages:
1) Post-editing HimL MT; and
2) Using Cochrane’s standard workflow: translating from scratch with access to Google MT, translation memory and glossary.
Experienced Cochrane domain experts worked on the tasks. The tasks were set up in MateCat translation software, which measures the time needed to complete the translations.
After completion of the tasks, translators completed a survey to measure their acceptability of the post-editing process and HimL MT.
Results:
Post-editing of HimL MT was clearly quicker than Cochrane's standard workflow for Czech, German, and Romanian. For Polish, post-editing and standard workflow took about the same time.
9 out of 10 translators rated HimL MT for the purpose of post-editing as acceptable and/or helpful. 7 out of 10 found post-editing quicker, and 6 out of 10 preferred post-editing over the standard workflow.
Conclusions:
The results provide a clear case for using domain adapted, state of the art MT as a basis for post-editing to speed up and facilitate translation of Cochrane PLS in certain languages. Some languages may require more research.
Most translators seem to find post-editing of high-quality MT acceptable, useful and preferable to not having it.
Some translators may not want to or be able to adapt to post-editing as compared to traditional translation workflows.
Patient or healthcare consumer involvement: N/A
Health in my Language (HimL) was an EU-funded, three-year project. It aimed to address the need for reliable and affordable translation of public health content into different languages via fully automatic machine-translation (MT) systems, focusing on translation from English into Czech, German, Polish, and Romanian. Recent advances in MT were used to produce health adapted MT engines. Health information from Cochrane and NHS24 served as the test cases to evaluate the obtained translations.
Objectives:
To show the effectiveness and translator acceptability of HimL MT when used as part of a post-editing workflow to translate Cochrane Plain Language Summaries (PLS), compared to Cochrane's standard translation workflow.
Methods:
10 randomly selected PLS were translated twice into all HimL languages:
1) Post-editing HimL MT; and
2) Using Cochrane’s standard workflow: translating from scratch with access to Google MT, translation memory and glossary.
Experienced Cochrane domain experts worked on the tasks. The tasks were set up in MateCat translation software, which measures the time needed to complete the translations.
After completion of the tasks, translators completed a survey to measure their acceptability of the post-editing process and HimL MT.
Results:
Post-editing of HimL MT was clearly quicker than Cochrane's standard workflow for Czech, German, and Romanian. For Polish, post-editing and standard workflow took about the same time.
9 out of 10 translators rated HimL MT for the purpose of post-editing as acceptable and/or helpful. 7 out of 10 found post-editing quicker, and 6 out of 10 preferred post-editing over the standard workflow.
Conclusions:
The results provide a clear case for using domain adapted, state of the art MT as a basis for post-editing to speed up and facilitate translation of Cochrane PLS in certain languages. Some languages may require more research.
Most translators seem to find post-editing of high-quality MT acceptable, useful and preferable to not having it.
Some translators may not want to or be able to adapt to post-editing as compared to traditional translation workflows.
Patient or healthcare consumer involvement: N/A