Fine-tuning Translation Memory Suggestions

By Edouard on November 22, 2011

A few weeks ago, I blogged about the recent changes to the Translation Memory server: new ranking algorithm, aberrations removal and pushing the 100% matches to the top.

This morning I added a setting to fine-tune the translation memory matching, based on ranking. At the bottom of the translation interface, in the toolbar, you will find a new selector:

It lets you choose the minimum level of accuracy of the suggestions pulled from the Translation Memory server.

It also lets you control machine translation suggestions. If you select a minimum match of 4, and all suggestions fall under 4, WebTranslateIt will suggest machine translations from Bing Translator and Google Translate, should you have set up the API tokens in the project settings.

How should you use it?

All projects are different, so play with it. I usually lower the ranking for new projects where the Translation Memory is almost empty to 2.

If you project’s translations is mostly complete, you might want to raise the level to 7 or 10.

Note that this is a user setting, so it only affects your own suggestions.

About WebTranslateIt’s ranking

Segment ranking is loosely equivalent to the percentage matching (50%-100%) you can find on traditional translation tools.

WebTranslateIt only pulls good suggestions from the Translation Memory server. Segments with a matching of 2 are OK results, 4 are a good results. 10 are excellent matches and “Complete Match” means the exact same segment was pulled from the Translation Memory.

I hope you will find this improvement useful. Thank you for using WebTranslateIt.