Tuesday, September 27, 2016

memoQ fuzzy match blues

In the past, I've criticized SDL and its programs for offering as "fuzzy matches" sentences that were far from helpful.

But the worst I've seen from SDL doesn't even compare with some absurd fuzzy matches that memoQ is suggesting. See for example:


I cannot understand how the matching engine in memoQ is suggesting as a 98% match for "Legal Entity" the translation for "CALL 1-800-555-5555".
Posted by Riccardo at 6:19 PM
Labels: , ,

5 comments:

  1. gergelyvWednesday, September 28, 2016

    Hello Riccardo,

    Have you tried repairing the TM? Are you using the latest build? This is a bug, probably an old one we have already fixed, leaving corrupted data in your TM. In memoQ, anything above 94% normally means there is no difference in text, only in numbers, tags, formatting, capitalization, etc.

    Gergely Vandor
    Kilgray

    ReplyDelete
  2. RiccardoWednesday, September 28, 2016

    Hi,
    Unfortunately this is a memory on a server, so I don't have access for operations such as repairs.
    I am using the latest build of memoQ 2015... I have no idea what build the server software is, unfortunately, as it is not under my control.
    But thank for the tip: I'll ask my customer to reindex the memory.

    ReplyDelete
  3. nikThursday, October 20, 2016

    I used the version 2015 and i think i have meet again this bug in past on this translation memory. Have you see in official site sometinhg?

    ReplyDelete
  4. Traductor translationFriday, January 06, 2017

    Hi! what is your view on google translate? do you think machine translation will be perfect, and in that case, when?

    ReplyDelete
    Replies
    1. RiccardoSaturday, January 07, 2017

      I doubt very much that machine translation will ever be perfect

      Delete

Thank you for your comment!

Unfortunately, comment spam has grown to the point that all comments need to be moderated. All legitimate comments will be published as soon as possible.

Subscribe to: Post Comments (Atom)