What does the "character-width difference penalty" do?

Question

Hi there, 
 I recently noticed the appearance of a character-width penalty for Asian source languages. 
 
 I tried to search the Studio documentation and release notes for info about what this does, but can't find anything. 
 Would anybody be able to explain what this penalty does and when best (not) to use it? 
 
 Thanks, Neeltje

Former Member · Accepted Answer

Please let me explain how this option works on Japanese source language. 
 
 Both ｱﾒﾘｶ and アメリカ mean "America". 
 If the option "Get matches for character-width diferences (Asian source languages)" is disabled, no TM match is found. 
 
 If the option "Get matches for character-width diferences (Asian source languages)" is enabled, TM match is found with -1% penalty. 
 
 If the value of penalty is 0, it will get 100% match or CM.

Paul Filkin · Answer

Neeltje Stuurman 
 
 Studio was enhanced a while ago by improving stemming for Western languages and also by making improvements to the recognition of halfwidth and fullwidth characters (Japanese particularly). With this enhancement it became important to be able to identify, or penalise TM matches that had a different character width to the source. 
 I can only hazard a guess as to why this would be important as translation into a European target language for example would probably not care since the character widths would not apply for single byte languages. But if you were translating into another double-byte language then it could.

Daniel Brockmann · Answer

Hi Neeltje Stuurman - we had long debates whether we should have any penalty at all, since when the text is the same in Japanese and just the character width is different, you don't really need to have any penalty. We settled on 1% in the end, though, as it might be good for such segments to be flagged as very high fuzzy matches so that a translator would at least take a quick look during translation even if the text is identical. BTW this applies to all source/target combinations - e.g. when you go from Japanese to English and the difference in the Japanese segment is just around the width, you will see a 99% match and can then do a quick check to confirm that the text is indeed the same, or if you need to alter the translation in any way (unlikely!). Or you go to 0% and just trust the TM engine to do its magic. I hope this clears things up at least somewhat. Thanks, Daniel

Trados Studio > 1. Trados Studio

What does the "character-width difference penalty" do?

Top Replies