recently reviewed content that colleagues of mine had been working carefully on. It is about the well-being of future generations and will be published on the Service Manual shortly.
I started to play with the idea of how to consider the Welsh language when planning content. That is, before writing or finalising any English content.
One way to do this is to consider a specific subject area’s terminology and create new vocabularies.
In this example, it was used to:
- ensure consistency within an area where there is already a large amount of Welsh content
- consider unfamiliar or technical terms that may be an obstacle or cause difficulty for the user
Before translators get in touch, I’m aware that there are legislative translation memories and general ones available, and that many of them are sure to contain terms and translations about this subject area. But the purpose here is not just to feed the translation machine.
What I used:
- claude.ai: for browsing the websites and creating vocabulary lists
- Excel: to store the data
- Phrase translation software: in order to contribute to translation work in the software
The instructions
1. Use claude.ai to retrieve the information
I was dealing with public information, so I asked claude.ai for help to:
build a comprehensive bilingual terminology (English and Welsh) for the Well-being of Future Generations Act. I’ll use it as a basis to create a consistent translation in some related content
The English pages you will be referring to are:
https://www.gov.wales/well-being-of-future-generations-wales (single page)
https://www.gov.wales/well-being-future-generations (page and links in text)The corresponding pages in Welsh are:
https://www.llyw.cymru/llesiant-cenedlaethaur-dyfodol-cymru
https://www.llyw.cymru/llesiant-cenedlaethaur-dyfodol
Claude spent some time browsing before creating a comprehensive list that included:
- Welsh terms
- equivalent English terms
- categories for the terms
- a note about some terms
I didn’t ask for categories or notes but these were helpful.
2. Extracting information from claude.ai
Claude gave me these in markdown or pdf formats to be downloaded. I wanted a .csv file (or Excel version) for different purposes. I asked Claude to create these and it worked.
3. Check the contents
Care needs to be taken to verify the content of GenAI. It’s surprisingly good, but I don’t fully trust it, particularly with the Welsh language.
Once I got a version that I could install in an Excel file, it was possible to check the content and review what was produced. To do this in great detail would require cross-referencing each term. For this purpose, I sampled the terms that felt by intuition most likely to be wrong, and found that Claude had been right.
Once we have this file, it can be used as a basis for discussing terminology in Welsh and English and a reference cross with our own content. This can provide an opportunity to consider how to present terms from a niche area in a different context.
I haven’t reviewed it all yet, but here’s a version of the future generations well-being terminology on Google Sheets.
4. Customize the Excel file and load it from Phrase
To facilitate the translation and proofreading work that will take place in the translation software, the Excel file can be adapted to the correct format as a ‘Term base’ for use in the project.
Suitable use
This method could be suitable for:
- creating terminology that is shared across teams before starting to create content in any language
- sharing glossaries where people don’t have access to a translation memory or the ability to open .tmx files
- creating consistency in a project and discuss what to do with complex terms or jargon before writing content in any language
- identifying terms that may be useful to test with users
- exploring terms that are already used bilingually in a particular area, where the information is already public
- creating a new terms database
It may also be possible to audit English content with this method in order to identify technical or jargon terms that need to be considered before creating new Welsh content.
Why do this with GenAI
This didn’t take more than an hour’s work, which is amazing, really. The purpose is not just to save time of course, but to use my time more efficiently. And isn’t it better to use the time it would have taken to audit the terminology manually, or research terms whilst translating, in order to review and verify what GenAI created? In this case, I think that’s true.
And if the intention is to create an artefact that creates discussion amongst a team – then using GenAI is convenient. It facilitate us to use our expertise, not just letting GenAI do everything for us.
I’d like to know if anyone has tried something similar. Have you?