Can I set penalty to a TM (all TUs of the TM)?
6 comments
-
Permanently deleted user Yes, you can set penalties for text differences (wording, spelling) and/or for formatting differences (inline tags).
You can set this up independently for each TM/termbase., but also in the default configuration to be applied to all.
Independently: Select the TM and click “Search settings”
Default configuration : “Settings” > “Translation Memories”
-
Mike Holland Exactly how many percentage points are subtracted at each level of the setting?
-
Stephan Böhmig Your question goes straight to the point. I'd love to be able to tell you "2%" less, but I am unable:
The mechanism that compares two segments in order to calculate a percentage is quite complex. It looks at word positions, segment lengths, languages, character differences, word/token differences, grammatical adjustments and much more. After comparison, the system obtains a list of differences (e.g. a comma not present in the other segment), each with information of "importance" and an individual percentage: "house" and "houses" will give a < 100% similarity. The system then combines all those differences into one single match percentage.
The algorithm is driven by many parameters, among which the settings described by Annelie. Some parameters are themselves calculated and others are preset by us.
When you move the text or formatting differences settings up or down, the individual differences' percentages are adjusted. For example setting the "text differences" to "Highest" will change 99% to 98%, 98% to 96%, 97% to 94% and so on. "Lowest" will change all values below 99% to 99%. Once this is done, all the differences will be combined together using other parameters.
It is therefore not possible to give a simple answer other than: Try out which settings are best for you.
ASIAN LANGUAGES: For Japanese, Chinese or Korean it is highly recommended to set the "text difference" penalty to "Highest". This will substantially reduce the amount of noise in results and lead to best performance.
-
Richard Issac Delgado Jr. What is a TM?
-
Permanently deleted user TM is the abbreviation for translation memory.
-
Sales Hello Richard,
Indeed, TM stands for Translation Memory.
Please find hereafter a link for detailed information : http://www.wordbee.com/translation-management/translation-memory/