Pages in topic:   [1 2 3] >
MemoQ delivering totally wrong analysis
Thread poster: Daniel Arnold

Daniel Arnold
Germany
Local time: 04:37
Member (2013)
German to English
May 19

Dear Colleagues,

MemoQ seems on the rise and many of my clients have switched from other CAT tools to MemoQ. I quite like working with it I have to say, I really appreciate its QA functions and that I dont have to send return packages anymore, but I have a pretty troubling problem with the way memoQ analyses the documents. What I mean is the analysis that my customer, the translation agency, runs on their memoQ and that forms the basis for what I am paid for the translation (you know the list of new words, repetions, fuzzies etc.)

I have recently had not one but a number of projects in memoQ where the analysis was just completely wrong. There were segments counted as "80%" that in reality had nothing in common with the TM than a few "is" or "here" or alike. The sentence had 18 words, three were "real" matches like the ones I have just described, and the other 15 words were no match at all. Yet memoQ tells me it's 80%. I even had segments were I was told it's a 60% match and there wasn't a single word that would have matched.

This worries me because those of us who work a lot for agencies get paid on the basis of the analysis the agencies run with their memoQ. And if it's wrong I have more work than the analysis says and don't get paid a fair rate for the job.

Did anyone have similar experiences ? What did you do about it?

Thanks, Dan


Kodi Dotterer
 

Kay-Viktor Stegemann
Germany
Local time: 04:37
Member (2016)
English to German
Reproducible? May 19

Can you reproduce such a sentence without violating an NDA?

 

Thomas T. Frost  Identity Verified
Member (2014)
Danish to English
+ ...
Different possibilities May 19

They could have included a translation memory that isn't attached to your project.

Or there is some sort of error in the original analysis, which may have been run in different software. I recently had a case like that. Some segments in the TM had the source filled in as target, and had been counted as 100%.

When there is more than marginal difference between the result of my own analysis in MemoQ and the client's, I ask for an explanation. There is usually a reason. I don't blindly accept an incorrect analysis.


 

Daniel Arnold
Germany
Local time: 04:37
Member (2013)
German to English
TOPIC STARTER
Example May 19

Here is an example of a segment that MemoQ says is a 62% match. If you look at it, it's clear that it's not anywhere near that.

 

Daniel Arnold
Germany
Local time: 04:37
Member (2013)
German to English
TOPIC STARTER
Pic May 19

jukcg8vkgogyqa8wkwx5.jpg

 

Epameinondas Soufleros  Identity Verified
Greece
Local time: 05:37
Member (2008)
English to Greek
+ ...
Totally wrong May 19

This happens here too. Plus, the total reported by the count function is different from the one reported by the analysis function, and has been since memoQ 2015...

As if these were not enough, in my most recent big project, consisting of a large bilingual Excel file, the count showed 15,000 fewer words before I selected the option to segment text inside cells. And I based my quote on it...

Conclusion: don't trust memoQ for your word counts. The quality of this app has decreased markedly in the past few years.

PS: I can see in your screenshot that your LSC is broken, too. The default setting for LSC hits (the orange ones) is to only show hits of more than 3 words and more than 16 characters. But memoQ totally ignores the setting and lists a long list of tiny, single-word fragments... Again, this behaviour is present both in version 7 (2015) and version 8.

[Edited at 2018-05-19 22:41 GMT]


 

Epameinondas Soufleros  Identity Verified
Greece
Local time: 05:37
Member (2008)
English to Greek
+ ...
More like a 38% match May 19

Your new sentence has 21 words.
The one in your TM has 7 words in common with the new one: one-third.
So, 62% must be the difference, not the similarity!

Possibly a careless programmer reversed the calculation somehow... If this is confirmed, it must be the most serious bug I have ever seen in a CAT tool, and it speaks volumes of the apparent lack of QA in Kilgray's DevOps pipeline.

[Edited at 2018-05-19 22:57 GMT]


 

Georgie Scott  Identity Verified
France
Local time: 04:37
Member (2009)
French to English
+ ...
Same experience May 20

I see this sort of thing all the time. Luckily the client I work with on MemoQ mostly only asks for discounts on repetitions and 100%/101% matches.

 

Lian Pang  Identity Verified
Netherlands
Local time: 04:37
Member (Mar 2018)
English to Chinese
+ ...
Only for reference May 20

TM can only be seen as a source of reference. In different contexts, the same word or texts can have totally different meanings. Like " clear the room" in a military context and in a household context.

I am reluctant to work for agencies or clients who ask for a discount for fuzzy matches.

The word "reluctant" is a bit of an understatement. I just reject them.


 

Epameinondas Soufleros  Identity Verified
Greece
Local time: 05:37
Member (2008)
English to Greek
+ ...
Not only 'for reference' May 20

Translation memories are a valuable asset if they are appropriately set up.

Using 'big mama' TMs will be of very little value.

But using targeted TMs, based on:
  • a specific subject matter

  • a specific end client, or

  • a combination of translation agency and end client (which is an 'account' from the perspective of the freelance translator)

will be of immense value.


But let's not continue this discussion here, as it is off topic.


 

Daniel Arnold
Germany
Local time: 04:37
Member (2013)
German to English
TOPIC STARTER
Not the only one May 20

Seems I am not the only one with this experience. I do allow my agency clients for the usual discounts, but I will have to review this in light of what memoQ is doing here. It also distorts the view on your work progress and the workload of a job. I must say I am quite annoyed by this and would love to hear from someone at Kilgray here in response to this thread.

 

Epameinondas Soufleros  Identity Verified
Greece
Local time: 05:37
Member (2008)
English to Greek
+ ...
One fix coming May 20

As I complained recently about the discrepancy between the total in the count and the total in the analysis, they had a look and they said they fixed it.

Let's see if they include it in the next build of memoQ 8. I will stay with memoQ 2015, though.


 

Christophe Delaunay  Identity Verified
Italy
Local time: 04:37
Spanish to French
+ ...
Here is the only way to go by May 21

Georgie Scott wrote:

I see this sort of thing all the time. Luckily the client I work with on MemoQ mostly only asks for discounts on repetitions and 100%/101% matches.


and it should be the ONLY "discount" we accept (if we do). More than that means we agree to shoot ourselves in the foot!


 

Daniel Arnold
Germany
Local time: 04:37
Member (2013)
German to English
TOPIC STARTER
Second example May 22

Just for the sake of documentation in the case someone from Kilgray looks at this thread, here is another example. Different job and different client. Completely wrong fuzzy percentage.

 

Daniel Arnold
Germany
Local time: 04:37
Member (2013)
German to English
TOPIC STARTER
Pic May 22

Pic

 
Pages in topic:   [1 2 3] >


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


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

MemoQ delivering totally wrong analysis

Advanced search







SDL Trados Studio 2019 Freelance
The leading translation software used by over 250,000 translators.

SDL Trados Studio 2019 has evolved to bring translators a brand new experience. Designed with user experience at its core, Studio 2019 transforms how new users get up and running and helps experienced users make the most of the powerful features.

More info »
Déjà Vu X3
Try it, Love it

Find out why Déjà Vu is today the most flexible, customizable and user-friendly tool on the market. See the brand new features in action: *Completely redesigned user interface *Live Preview *Inline spell checking *Inline

More info »



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