Mobile menu

Why is terminology recognition so dreadfully poor?
Thread poster: Astrid Elke Witte

Astrid Elke Witte  Identity Verified
Germany
Local time: 02:01
Member (2002)
German to English
+ ...
Mar 16, 2008

I find it makes utter nonsense of the concept of building up terminology termbases if the software is designed in such a way that only SOME of the terminology is recognised, SOME of the time, ON AN ENTIRELY RANDOM basis.

No, I do not have duplicate or incomplete entries.

Most certainly, I have recently reorganised all my termbases.

My question is, why is software that is so expensive so erratic?


Astrid


Direct link Reply with quote
 

Stanislav Pokorny  Identity Verified
Czech Republic
Local time: 02:01
English to Czech
+ ...
Fuzzy searching Mar 16, 2008

Hi Astrid, have you turned on the "Fuzzy terminology recognition" option?

Direct link Reply with quote
 

Astrid Elke Witte  Identity Verified
Germany
Local time: 02:01
Member (2002)
German to English
+ ...
TOPIC STARTER
No, but it is not fuzzy terminology recognition that I am looking for Mar 16, 2008

I had meant when the identical word that I have put into the termbase, and in the identical form, occurs again and is not recognised.

Astrid


Direct link Reply with quote
 
lexical  Identity Verified
Spain
Local time: 02:01
Portuguese to English
cynical reply Mar 16, 2008

The cynical reply, Astrid, is that they want your money more than your satisfaction. The "Traduce" software has paid for itself many thousands of times over, but they still want 700+ euros a pop for it, and the more they develop it and bring out new versions, the less reliable it becomes. Complexity is the enemy of reliability in this case.

You may come to feel like me - that it's time to drop Traduce, and work in another, less exploitative CAT tool. You simply tell clients that you do not work in Traduce. The only clients that insist on a specific CAT tool are the ones who want to chisel away at your fees.


Direct link Reply with quote
 

ViktoriaG  Identity Verified
Canada
Local time: 20:01
English to French
+ ...
Bluntly put Mar 16, 2008

lexical wrote:

The only clients that insist on a specific CAT tool are the ones who want to chisel away at your fees.


It's a blunt remark - but it accurately reflects the truth (although some agencies insist on the use of certain CAT tools only because they, too, have become slaves to them, like many freelancers, and wouldn't be able to reuse older material if they switched to something else or left the choice of CAT tool up to the translator). Made me laugh out loud!


Direct link Reply with quote
 
Daniel García
English to Spanish
+ ...
Maybe you could add some example? Mar 17, 2008


I find it makes utter nonsense of the concept of building up terminology termbases if the software is designed in such a way that only SOME of the terminology is recognised, SOME of the time, ON AN ENTIRELY RANDOM basis.


In my case, working with English as the source language, the term recognition seems to work reasonably well. Maybe it does not work so well with other languages as source.

Nevertheless, I think it would be interesting to see some examples of those random suggestions you are getting.

Daniel


Direct link Reply with quote
 

Astrid Elke Witte  Identity Verified
Germany
Local time: 02:01
Member (2002)
German to English
+ ...
TOPIC STARTER
I am not getting any "random suggestions" Mar 17, 2008

Examples? The example is that, with German as the source language and English as the target language, many hundreds of words in my termbases are only being recognised when I open some TUs and not when I open others (the same words, without any variation).

In particular, the longer a TU is, the less chance there is of words being recognised. In an extremely long TU, on average around 30% of the words in the termbases are recognised. That occurs when the TUs consist of legal German, in the case of which the sentences are 100 to 200 words long. When I then move on to the next TU of only 30-40 words, the same words will be recognised again.

However, that was not my problem yesterday. I was doing a non-legal German into English translation, with sentences of average length, and words in my termbases were simply not being recognised - or being recognised in one TU and not the next. In this case, I also mean words, as well as TUs, of only normal length.

Astrid


Direct link Reply with quote
 

Heinrich Pesch  Identity Verified
Finland
Local time: 03:01
Member (2003)
Finnish to German
+ ...
You might try SDLX Mar 17, 2008

I have never used Multiterm, but in SDLX term recognition seems to work quite well.
And it is easy to import own glossaries.

Regards
Heinrich


Direct link Reply with quote
 

Astrid Elke Witte  Identity Verified
Germany
Local time: 02:01
Member (2002)
German to English
+ ...
TOPIC STARTER
That sounds like a good idea, Heinrich! Mar 17, 2008

Since I received SDLX together with Trados, I may as well try it out. I have never looked at it yet.

Thanks for the tip!

Best regards,

Astrid


Direct link Reply with quote
 
torb  Identity Verified
Norway
Local time: 02:01
English to Norwegian
+ ...
Minimum match value Mar 18, 2008

I think the problem here is the minimum match value found in Workbench... This setting tells how much match (degree of similarity) you have to have on a translation unit to find it and list it...

The problem here is the matching of translation units. You have no real method of searching for a single word. You can set this value as low as 30% to get more hits (don't remember the standard setting anymore) - but you will still have problems. So, in the best case, it will list the finding if it has 30% (or more) similarity to the term you are searching for. It makes it very difficult to search for a single word in a long sentence, right? The longer the sentence (or translation unit, which maybe would be the correct term) is, the more difficult it will be to find single words. This will explain why you find words some times, and other times not.

Another setting is the Maximum number of hits, which tells how many hits should be shown. The standard setting is very low, maybe for lowering the time it takes to return the search result, and should be raised to get more hits.

I personally think the whole concordance search is a joke. In extreme cases I have exported the TM to a .tmx, and used Notepad to search for the term I am looking for. Notepad is a good companion to Trados, which in itself should be alarming.

So.... what you should do to get the most out of the situation:

In Trados Workbench, go to Tools - Translation memory options
Set "Minimum match value %" to 30
Set "Maximum number of hits" to 50 (or the number you like)

I hope this helps somewhat.

Best regards,

Tor

Astrid Elke Johnson wrote:

Examples? The example is that, with German as the source language and English as the target language, many hundreds of words in my termbases are only being recognised when I open some TUs and not when I open others (the same words, without any variation).

In particular, the longer a TU is, the less chance there is of words being recognised. In an extremely long TU, on average around 30% of the words in the termbases are recognised. That occurs when the TUs consist of legal German, in the case of which the sentences are 100 to 200 words long. When I then move on to the next TU of only 30-40 words, the same words will be recognised again.

However, that was not my problem yesterday. I was doing a non-legal German into English translation, with sentences of average length, and words in my termbases were simply not being recognised - or being recognised in one TU and not the next. In this case, I also mean words, as well as TUs, of only normal length.

Astrid




[Edited at 2008-03-18 15:51]


Direct link Reply with quote
 

Astrid Elke Witte  Identity Verified
Germany
Local time: 02:01
Member (2002)
German to English
+ ...
TOPIC STARTER
I meant words that are a 100% match Mar 18, 2008

Hi Tor,

Thank you for your comments. I did originally have the match setting pretty high (at one time I tried to keep it at 90%, in order not to get rubbish matches, which frustrated me), I later turned it down to 65% to catch fuzzy matches with slightly different word endings, and now I have it lower still. However, that is not the issue. If words are ABSOLUTELY IDENTICAL to what is in the Multiterm termbase, then the match is 100%.

I do not think that we are talking about searching through sentences for words. It is the Multiterm dictionary that is supposed to be searched. The concordance search, through previous translation units, is a different issue. Therefore, I do not see why the sentence length should make a difference. I am complaining about identical words to those entered in the Multiterm dictionary, i.e. 100% match words, not being recognised.

The subject of hits also concerns the concordance search, and not the subject of recognising words in the Multiterm dictionary.

I have the search depth set to the maximum value of 999.

Best regards,

Astrid


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 »

Why is terminology recognition so dreadfully poor?

Advanced search


Translation news related to SDL Trados





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 »
LSP.expert
You’re a freelance translator? LSP.expert helps you manage your daily translation jobs. It’s easy, fast and secure.

How about you start tracking translation jobs and sending invoices in minutes? You can also manage your clients and generate reports about your business activities. So you always keep a clear view on your planning, AND you get a free 30 day trial period!

More info »



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