Pre-translation: a weird problem
Thread poster: Elif Baykara

Elif Baykara  Identity Verified
Turkey
Local time: 07:23
Member (2015)
German to Turkish
+ ...
Oct 23, 2016

Hello, all!

I have a weird problem regarding pre-translation. I have searched through the fora and followed the relevant steps proposed therein.

I created a project of 3 Word files (173 segments) as usual. I added the TM I always use with such projects. And I performed an analysis of the project (all segments were empty). I attached the screenshot of the analysis below: there are some matches, about 45% of all segments is matched either as 100%, 101% or 95-99%.

Then I pre-translated the file for exact matches: no change at all.

I am puzzled. This is not an urgent question, I will complete the project without pretranslation anyway. But I would like to hear your opinions on the subject.

Best,
Elif

Screenshot 2016-10-23 15.51.10


Direct link Reply with quote
 

Anthony Green  Identity Verified
Italy
Local time: 06:23
Italian to English
what is puzzling you? Oct 23, 2016

Hello Elif
I don't understand what you are puzzled by as it seems fairly normal. Can you explain?


Direct link Reply with quote
 

Elif Baykara  Identity Verified
Turkey
Local time: 07:23
Member (2015)
German to Turkish
+ ...
TOPIC STARTER
None of the segments were pre-translated. Oct 23, 2016

Probably I am missing something very basic, but since I have some exact matches (e.g., 35 segments are matched by 100%), I expect them to be pre-translated. At least this is what happened before in the previous projects where I created the project, in the same way, using the same TM.

Elif


Direct link Reply with quote
 

Mirko Mainardi  Identity Verified
Italy
Local time: 06:23
Member
English to Italian
+ ...
"Good match" Oct 23, 2016

Out of curiosity, what happens if you pre-translate the documents using the "Good match" option?

Direct link Reply with quote
 

Anthony Green  Identity Verified
Italy
Local time: 06:23
Italian to English
I see Oct 23, 2016

Elif Baykara wrote:

since I have some exact matches (e.g., 35 segments are matched by 100%), I expect them to be pre-translated.
Elif

Oh yes, I understand. Yes that's strange.
If you go into one of the segments with a match (100% or otherwise), do you see the red field in the Translation Results pane corresponding to the match?


Direct link Reply with quote
 

Elif Baykara  Identity Verified
Turkey
Local time: 07:23
Member (2015)
German to Turkish
+ ...
TOPIC STARTER
Now I did this: Oct 23, 2016

As a final check, I did the following:

1) I created a new project with the same 3 Word files. This time, it gave 179 segments in total (it was 173).

2) I used the same TM as before.

3) I analyzed the project: I got pretty much the same result.

4) I pre-translated for exact matches: none of the segments were changed (all segments are still empty).

5) I pre-translated for good matches: none of the segments were changed (all segments are still empty).

6) I pre-translated for any match: 6 of the segments were pre-translated (all other segments are still empty). Pretranslated segments contain either a number or a single word.


Direct link Reply with quote
 

Elif Baykara  Identity Verified
Turkey
Local time: 07:23
Member (2015)
German to Turkish
+ ...
TOPIC STARTER
I can't see the 100% matches at all. Oct 23, 2016

Anthony Green wrote:

Elif Baykara wrote:

since I have some exact matches (e.g., 35 segments are matched by 100%), I expect them to be pre-translated.
Elif

Oh yes, I understand. Yes that's strange.
If you go into one of the segments with a match (100% or otherwise), do you see the red field in the Translation Results pane corresponding to the match?


I filtered the files for fuzzy, 100% and 101% matches. There are some with 99% or 66%, for example. But no, I don't see any 100% or 101% matches.


Direct link Reply with quote
 

Kuba Kościelniak  Identity Verified
Poland
Local time: 06:23
Member (2013)
English to Polish
+ ...

MODERATOR
Homogeneity? Oct 23, 2016

I have noticed that for your analysis you have Homogeneity turned on — maybe it lists high-percent matches within the document itself, and not from the TM? Try to re-run the analysis with Homogeneity field unchecked and let us now about the results.

Greetings,

K.


Direct link Reply with quote
 

Mirko Mainardi  Identity Verified
Italy
Local time: 06:23
Member
English to Italian
+ ...
Good point! Oct 23, 2016

Kuba Kościelniak wrote:

I have noticed that for your analysis you have Homogeneity turned on — maybe it lists high-percent matches within the document itself, and not from the TM? Try to re-run the analysis with Homogeneity field unchecked and let us now about the results.


That might well be the solution to this mystery


Direct link Reply with quote
 

Elif Baykara  Identity Verified
Turkey
Local time: 07:23
Member (2015)
German to Turkish
+ ...
TOPIC STARTER
@Kuba Oct 23, 2016

I performed two analyses, with and without homogeneity on. However, I have already started with the translation and I have 23 confirmed segments.

I pre-translated using this updated TM and all 23 segments were pre-translated. Comparing the two analyses also confirms that you are right.

I believe that I should not postpone a training on memoQ anymore I have a faint idea of homogeneity, but I have never thought that it could be that misleading.

Thank you all for your efforts


Screenshot 2016-10-23 18.26.02


Screenshot 2016-10-23 18.27.59


Direct link Reply with quote
 

Philippe Etienne  Identity Verified
Spain
Local time: 06:23
Member
English to French
Homogeneity Oct 23, 2016

Mirko Mainardi wrote:

Kuba Kościelniak wrote:

I have noticed that for your analysis you have Homogeneity turned on — maybe it lists high-percent matches within the document itself, and not from the TM? Try to re-run the analysis with Homogeneity field unchecked and let us now about the results.


That might well be the solution to this mystery

When Homogeneity is on, there are no repeats, instead the "repeats" are counted as 100 and 101%.
From your analyses, 687+139+2 = 507+321

Why:
MemoQ scans an unknown segment that is repeated later: counts it as normal, ie 0% if there is no match in the TM.
MemoQ later scans the same segment: with Homogeneity On, MemoQ says that it's already "virtually" translated (from the scan above), so it's a 100% (or 101%). With Homogeneity Off, MemoQ counts it as a repeat.

Homogeneity counts "internal fuzzies" in a text:
Homogeneity Off, TM empty
"I have a dog" 0% match
"I have a frog" 0% match
Homogeneity On, TM empty
"I have a dog" 0% match
"I have a frog" 85% match (or something)
Therefore the discrepancy in fuzzy bands too in your analyses with and without Homogeneity.

Homogeneity On = Lower weighted wordcount

Philippe

Edit for errors in the English

[Edited at 2016-10-23 17:16 GMT]


Direct link Reply with quote
 

Elif Baykara  Identity Verified
Turkey
Local time: 07:23
Member (2015)
German to Turkish
+ ...
TOPIC STARTER
Thank you Philippe! Oct 23, 2016

You have explained it just the way I could understand.

I have only a few clients for which I have agreed to apply a discount for fuzzy matches. I use these analyses mainly to assess the projects so that I know what is waiting for me. So they serve just as a guide for me.

I am glad that I noticed this problem/feature relatively early and on a short project. In this way, I can make more accurate assessments and I will be more happy in general

Best,
Elif


Direct link Reply with quote
 

Niina Lahokoski  Identity Verified
Finland
Local time: 07:23
Member (2008)
English to Finnish
+ ...
Try checking "perform fragment assembling" Oct 23, 2016

Did you previously use the "copy source to target" function? That might prevent segments being pre-translated sometimes. The segments need to be empty for it to work.
Is "Perform Segment assembling" (or something like that) checked?

Apparently it needs to be checked in order for Pre-Translation to work. I learned this from MemoQ support when I had the same problem once.

By the way, the above setting is one example of user-unfriendly design. The name tells the user nothing at all about what the setting does and what it is needed for. It definitely would never have occurred to me to check this option just to get Pre-Translation to work. "Fragment assembling" doesn't sound like something to do with TM matches. I would have guessed it to be related to machine translation.

[Edited at 2016-10-23 19:09 GMT]


Direct link Reply with quote
 

LEXpert  Identity Verified
United States
Local time: 23:23
Member (2008)
Croatian to English
+ ...
Are the pre-translated segments confirmed before re-running the analysis? Oct 23, 2016

So you ran an analysis, then pretranslation, then re-ran the analysis?

Just running the pretranslation won't change the analysis unless you confirm the pre-translated segments, otherwise nothing in your TM is changing.


Direct link Reply with quote
 


To report site rules violations or get help, contact a site moderator:


You can also contact site staff by submitting a support request »

Pre-translation: a weird problem

Advanced search






SDL MultiTerm 2017
Guarantee a unified, consistent and high-quality translation with terminology software by the industry leaders.

SDL MultiTerm 2017 allows translators to create one central location to store and manage multilingual terminology, and with SDL MultiTerm Extract 2017 you can automatically create term lists from your existing documentation to save time.

More info »
TM-Town
Manage your TMs and Terms ... and boost your translation business

Are you ready for something fresh in the industry? TM-Town is a unique new site for you -- the freelance translator -- to store, manage and share translation memories (TMs) and glossaries...and potentially meet new clients on the basis of your prior work.

More info »



Forums
  • All of ProZ.com
  • Term search
  • Jobs
  • Forums
  • Multiple search