Forum: XTM Cloud support
Topic: Yet another terrible discovery
Poster: Samuel Murray
Hello everyone
So, when I'm asked to do translations in XTM, there are often no fuzzy matches, so I just translate segment by segment, without worrying about checking fuzzies. Or, I'm often asked to proofread translations that were done in XTM.
When proofreading, I am often struck by how regularly it appears as if translators simply didn't edit fuzzy matches. The fuzzy matches are confirmed, but they weren't actually edited (or they were incompletely edited).
Later, I discovered that there is a feature in XTM (that is enabled by default) whereby a segment gets confirmed as soon as you click somewhere else (you can disable this behaviour in the settings). So if a translator puts his cursor in a fuzzy matched segment, don't do anything, and then click in another segment, XTM considers the segment to have been edited and confirmed.
I now discovered another cause for what I've been seeing.
Today I'm tasked with translating a job in XTM that has about 50 high fuzzy matches, so I decided to edit the fuzzy matches first. I filtered the "unedited fuzzy matches", then changed the settings so that the segment status is not changed automatically (i.e. I have to press Alt+; to confirm a segment) and so that repeated segments are NOT updated:
Then I went ahead and edited the fuzzy matches and used Alt+; to confirm them and used Alt+right to move to the next segment each time. So, in my of thinking, therefore, if a fuzzy match did have a repetition, then that repetition would NOT be updated automatically and (it goes without saying) the unedited repetition's status would not be changed to e.g. Confirmed.
Fortunately, I double-check my translations in Wordfast Classic, and that is when I discovered a horrible truth: yes, XTM did not update the content of the repetitions, but XTM *did* mark repetitions as confirmed. So, a repetition of a fuzzy match gets "confirmed" when the first instance of that repetition gets confirmed, even if only the first instance of the fuzzy match had been edited. Why????????
So, despite that the setting is called "Update text [u]and status[/u] of repeated segments in the current file", and you select "Do not update segments", it *does* update the statuses of those segments, even though the content is not updated.
The result of this behaviour was that I had several confirmed segments (which I did not confirm myself) which contained unedited fuzzy matches (because I had edited only the first instance of each repetition).
I have long stopped being upset by sloppy translations in XTM when it is clear that the "sloppiness" is the result of an unedited fuzzy match or the result of an incompletely edited segment that was autopropagated, because I know that the translators are not at fault. Today, I too was not sloppy -- I was being extra careful -- and still XTM managed to screw it up.
So, I then had to re-check all confirmed segments very carefully (and this time, the fuzzy match comparison pane did not help, because the unedited fuzzy matches were now considered exact matches with the TM). I'm very fortunate that there were only 50 segments and that checking them took only an extra half an hour of my time... and here I am giving discounts for fuzzy matches.
What is the lesson to be learnt here? Never check pre-translated fuzzy matches in XTM. Translate such segments from scratch, and refuse to give discounts for fuzzy matches in XTM because they actually cost MORE time than just translating from scratch. And if the client is unhappy because they're not getting consistency with the TM, well, then we should educate such clients that that is what you get from a CAT tool that is designed to produce faulty translations.
When clients choose XTM, they want to benefit from discounts for fuzzy matches, but they require that the translator use a tool that is booby-trapped, and then the translator's reputation suffers when (through no fault of his own) his translation ends up with confirmed unedited fuzzy matched segments.
Samuel
Topic: Yet another terrible discovery
Poster: Samuel Murray
Hello everyone
So, when I'm asked to do translations in XTM, there are often no fuzzy matches, so I just translate segment by segment, without worrying about checking fuzzies. Or, I'm often asked to proofread translations that were done in XTM.
When proofreading, I am often struck by how regularly it appears as if translators simply didn't edit fuzzy matches. The fuzzy matches are confirmed, but they weren't actually edited (or they were incompletely edited).
Later, I discovered that there is a feature in XTM (that is enabled by default) whereby a segment gets confirmed as soon as you click somewhere else (you can disable this behaviour in the settings). So if a translator puts his cursor in a fuzzy matched segment, don't do anything, and then click in another segment, XTM considers the segment to have been edited and confirmed.
I now discovered another cause for what I've been seeing.
Today I'm tasked with translating a job in XTM that has about 50 high fuzzy matches, so I decided to edit the fuzzy matches first. I filtered the "unedited fuzzy matches", then changed the settings so that the segment status is not changed automatically (i.e. I have to press Alt+; to confirm a segment) and so that repeated segments are NOT updated:
Then I went ahead and edited the fuzzy matches and used Alt+; to confirm them and used Alt+right to move to the next segment each time. So, in my of thinking, therefore, if a fuzzy match did have a repetition, then that repetition would NOT be updated automatically and (it goes without saying) the unedited repetition's status would not be changed to e.g. Confirmed.
Fortunately, I double-check my translations in Wordfast Classic, and that is when I discovered a horrible truth: yes, XTM did not update the content of the repetitions, but XTM *did* mark repetitions as confirmed. So, a repetition of a fuzzy match gets "confirmed" when the first instance of that repetition gets confirmed, even if only the first instance of the fuzzy match had been edited. Why????????
So, despite that the setting is called "Update text [u]and status[/u] of repeated segments in the current file", and you select "Do not update segments", it *does* update the statuses of those segments, even though the content is not updated.
The result of this behaviour was that I had several confirmed segments (which I did not confirm myself) which contained unedited fuzzy matches (because I had edited only the first instance of each repetition).
I have long stopped being upset by sloppy translations in XTM when it is clear that the "sloppiness" is the result of an unedited fuzzy match or the result of an incompletely edited segment that was autopropagated, because I know that the translators are not at fault. Today, I too was not sloppy -- I was being extra careful -- and still XTM managed to screw it up.
So, I then had to re-check all confirmed segments very carefully (and this time, the fuzzy match comparison pane did not help, because the unedited fuzzy matches were now considered exact matches with the TM). I'm very fortunate that there were only 50 segments and that checking them took only an extra half an hour of my time... and here I am giving discounts for fuzzy matches.
What is the lesson to be learnt here? Never check pre-translated fuzzy matches in XTM. Translate such segments from scratch, and refuse to give discounts for fuzzy matches in XTM because they actually cost MORE time than just translating from scratch. And if the client is unhappy because they're not getting consistency with the TM, well, then we should educate such clients that that is what you get from a CAT tool that is designed to produce faulty translations.
When clients choose XTM, they want to benefit from discounts for fuzzy matches, but they require that the translator use a tool that is booby-trapped, and then the translator's reputation suffers when (through no fault of his own) his translation ends up with confirmed unedited fuzzy matched segments.
Samuel