It'd be nice if AutoSuggest could suggest plural form
Thread poster: Jacques DP

Jacques DP  Identity Verified
Switzerland
Local time: 05:17
Member (2003)
English to French
Mar 8, 2015

Hi there,

Everytime I have a glossary match on a term which is in plural form in source text (but singular form in the glossary), I am offered the singular form and can't use it.

For example my glossary has:

dictionary attack ---- attaque par dictionnaire

but my source text has:

Dictionary attacks are carried out using programs...

So I start translating and write (^ is the insertion point):

Les a^

At this point auto-suggest offers

attaque par dictionnaire

But if I accept it I get

Les attaque par dictionnaire^

and so I would need to go back to the end of the second word

Les attaque^ par dictionnaire

in order to add an "s" (mark of plural) and then go back to end of line.

Les attaques par dictionnaire^

But since I am a professional linguist and writer I type fast, and that means doing all this is not worth it and it's faster to write it anew. (And I lose something because I need to make the decision as well.)

It would be a huge gain for me if Studio could, with a minimal grammatical understanding, realize that we have the plural form of the glossary match in source text and offer the plural form of its glossary translation.

I understand that it would require language-specific knowledge.

But is it really extraordinary to expect minimal language-specific knowledge from a computer-assisted translation tool? I think not!

By the way do you know of another CAT tool that would be able to do that?

Thank you.


[Edited at 2015-03-08 11:13 GMT]


Direct link Reply with quote
 

SDL Community  Identity Verified
United Kingdom
Local time: 05:17
English
Well... Mar 8, 2015

... you're asking the CAT to carry out machine translation if you don't tell it that adding the 's' makes this word plural? Of course this is not beyond the capability of a computer to do this but it's not always so straightforward. Even in English you have plural ('s' or 'es'), irregular plurals ('en'), funnies ('ies'), those that are nouns and actually contract when made plural, and probably more I can't even think about. Then we have every other supported langauge that are probably a lot more complex than the English.

The easy solution is to add the plural form into a termbase and don't rely on autosuggest dictionaries which are derived from things you did in the past.

Then of course you could also use machine translation at subsegment level, there are two options for this:

SDL Trados Studio MT AutoSuggest (uses whatever MT engine you are using)

Google Translate AutoSuggest (only uses Google and doesn't require a Google account)

So to answer your question "But is it really extraordinary to expect minimal language-specific knowledge from a computer-assisted translation tool?". Yes I think it is. You can help the computer, and various tools have means and ways of doing this, but to expect the computer to just do it... I think only if it's using machine translation.

Regards

Paul
SDL Community Support


Direct link Reply with quote
 

Meta Arkadia
Local time: 10:17
English to Indonesian
+ ...
Nope Mar 8, 2015

Jacques DP wrote:
By the way do you know of another CAT tool that would be able to do that?

But it can be done for single words.



I think it is based on the Hunspell spelling dictionary. Just open one (both the .dic and the .aff), and you'll see what I mean. Rules. But that's for single words only. And my CAT tool doesn't use this. CafeTran's autosuggest function is based on what's in your project, and in your resources, with priority to the project file. So if you entered attaques par dictionnaire once, it'll be autosuggested.

Cheers,

Hans

[Edited at 2015-03-08 13:28 GMT]


Direct link Reply with quote
 

Jacques DP  Identity Verified
Switzerland
Local time: 05:17
Member (2003)
English to French
TOPIC STARTER
Tricks Mar 8, 2015

Hi Paul,

Thank you for your answer. Let me elaborate slightly on this.

The reason why I posted this is because I find myself confronted to this exact same problem (plural form of a glossary match) many, many times a day.

Whenever one realizes that the same small problem happens again and again, and one needs to take care of it manually by doing the same small mechanical routine, it's natural to look for a way in which the computer might help, don't you think?

I thank you for your links. I tested the Google one. It's interesting, and I will keep experimenting with it a little, but I think it doesn't hit the target often enough that I may put it as first AutoSuggest provider and use it productively (I can only use the first suggestion of AutoSuggest -- reading the various suggestions, and moving in the list to choose one is much too long). Just like segment-level MT, it is noise to me and slows me down.

What I contemplated is not MT but a language-specific knowledge very similar to the language-specific knowledge of French grammar that MS Word has, and which allows it to tell me that "les attaque" has a grammar mistake because "attaque" needs an "s" in order to agree with "les".

Studio could have language-specific modules (different modules for various languages), with "tricks". Tricks would be very limited and specific compared to MT but they would almost always hit the target, and therefore be useful instead of being noise (not a sound theoretical difference but an important practical one).

Aren't you already doing such tricks to adapt numbers or dates? Is it not already language-specific? I'm not sure where this code is now, but maybe it could reside in a language-specific module, along with my tricks?

The "Glossary Agreement" trick might do the following. The English and French modules would have a basic rule/exceptions-based knowledge of plural in English and French respectively. When a fuzzy glossary match is found ("attacks" in text, "attack" in glossary), Studio would check if that difference is the mark of plural (TRUE in this instance). If so, the French Tricks module would pluralize the glossary translation so that it is offered first in the autosuggest list. By doing this, you don't even need to know whether the terms in the glossary are nouns or verbs, singular or plural.

You think it's a hack? Wrong, it's a trick It's a hack but it would save me hours and bring me joy as it would increase the ratio of enjoyable activity (writing) to tedious mechanical repetitive edits. It's exactly what you have done with localization of date, measures, etc. You just need to add more tricks.

Now back to work...

Kind regards

Jacques


Direct link Reply with quote
 

Jacques DP  Identity Verified
Switzerland
Local time: 05:17
Member (2003)
English to French
TOPIC STARTER
Kind of Mar 8, 2015

Hi Hans,

Thanks for your input.
If it doesn't come first in autosuggest list it is useless for me.
I need the correct form as the first form.

Adding as autosuggest whatever you have already typed in the project seems potentially extremely useful. At least you don't repeat the same edits many times in the same project.

Jacques


Direct link Reply with quote
 

SDL Community  Identity Verified
United Kingdom
Local time: 05:17
English
In the meantime... Mar 8, 2015

... can't you add them to your termbase as you work and then you only do it once.

Regards

Paul


Direct link Reply with quote
 

Jacques DP  Identity Verified
Switzerland
Local time: 05:17
Member (2003)
English to French
TOPIC STARTER
No, because Mar 8, 2015

just one of the glossaries I currently use has 630K entries (IATE EN>FR). Of course when the exact same thing comes back again and again I will deal with it by adding the plural form in the glossary myself. But I cannot do it preventively, as it would be a waste of time. Thanks.

[Edited at 2015-03-08 17:33 GMT]


Direct link Reply with quote
 

Jacques DP  Identity Verified
Switzerland
Local time: 05:17
Member (2003)
English to French
TOPIC STARTER
In the meantime? Mar 8, 2015

I noted "in the meantime". So I guess there is hope!

Direct link Reply with quote
 

SDL Community  Identity Verified
United Kingdom
Local time: 05:17
English
No it's not... Mar 8, 2015

Jacques DP wrote:

It's exactly what you have done with localization of date, measures, etc. You just need to add more tricks.



... as here we just use what is already available in the windows operating system.

Yes there is always hope, but this is pretty unlikely to be something coming from SDL. If you want something like this then we have exposed all the APIs to make it possible for you to develop it. If you don't know how to develop it why not get together with other enthusiastic users who think this is a worthwhile solution and chip in for a developer to do it? If SDL did it there would have a good business case for it so let's see what that is and how many people are interested? I'd be happy to help you find a developer.

You could also consider what Hans suggested with the dictionaries. This is quite an interesting idea as you could take the existing dictionary file and then translate every term in there. Now you have a termbase. You could also just import this monolingual file into the AutoText list and it would work out of the box exactly as Hans has shown for his favoured tool.

Regards

Paul
SDL Community Support


Direct link Reply with quote
 

Meta Arkadia
Local time: 10:17
English to Indonesian
+ ...
Nothing of the kind Mar 9, 2015

SDL Community wrote:
This is quite an interesting idea as you could take the existing dictionary file and then translate every term in there.

The example I provided comes from Hunspell (I think), and it's a "system-wide" Mac autosuggest. System-wide, but it works for Cocoa apps only, not for say Word:mac or CafeTran, unfortunately. And it's what I call "static," just like Trados' autosuggest. CafeTran's auto-complete, on the other hand, is "dynamic." It's based on your project (highest priority), and your resources. More chance the first suggestion is the one you want. And you don't have to add anything anywhere. It'll help Jacques a bit better than static solutions, but it isn't perfect.
You could probably write a rule that offers attaques after les as the first suggestion, but that won't help you a bit in Dictionary attacks are quite a nuissance. Unless you know of a way to implement the whole French (and Dutch, please, and...) syntax into the rules, making human translators even more redundant than they are now (according to some).

Cheers,

Hans


Direct link Reply with quote
 

Meta Arkadia
Local time: 10:17
English to Indonesian
+ ...
Nervous breakdown Mar 9, 2015

SDL Community wrote:
Now you have a termbase. You could also just import this monolingual file into the AutoText list and it would work out of the box exactly as Hans has shown for his favoured tool.

No, it wouldn't. That again would be static. It would just as easily suggest attaque de nerves as attaque par dictionaire. I wouldn't do that to Jacques, and neither would CafeTran. CafeTran's auto-completion offers suggestions based on:

Words that you type in the target box
Words in your User.dic, your personal list of words with unknown spelling
Words in your Translation Memory
Words in the target segments of your translation project
Words at the target side of your glossary
(copied from the CafeTran Wiki)

So even if you use IATE as a resource, it won't provide suggestions from it as a first choice, unless the other - context sensitive - resources don't come up with anything.

CafeTran doesn't use a specific database for auto-completion, so you don't have to create one. It's already there. Intelligently.

Now if you use auto-assemble, and take good care of the settings of your resources, you don't have to typ at all. That is, if you let CafeTran insert those most likely matches. Depending on your language pair, this may create a mess, because of a different word order. I use auto-assemble - I depend on it - and if the word order is wrong, I just start typing, deleting the CafeTran's suggested translation with the first keystroke. And for me, that's where auto-completion comes in. And it won't suggest attaque de nerves instead of attaque par dictionaire in an IT text on hacking. And not only because I don't translate into French.

Cheers,

Hans


Direct link Reply with quote
 

SDL Community  Identity Verified
United Kingdom
Local time: 05:17
English
Studio is not static as you suggest either... Mar 9, 2015

Meta Arkadia wrote:

The example I provided comes from Hunspell (I think), and it's a "system-wide" Mac autosuggest. System-wide, but it works for Cocoa apps only, not for say Word:mac or CafeTran, unfortunately. And it's what I call "static," just like Trados' autosuggest. CafeTran's auto-complete, on the other hand, is "dynamic." It's based on your project (highest priority), and your resources. More chance the first suggestion is the one you want. And you don't have to add anything anywhere. It'll help Jacques a bit better than static solutions, but it isn't perfect.



... at least AutoSuggest Dictionaries and termbases are not. Autotext is obviously static as it's monolingual and whatever is in the source is irrelevant. But if you were to take the Hunspell dictionary and translate it, then you wouild have a dynamic resource you could use.

I hope you recover soon.

Regards

Paul
SDL Community Support


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 »

It'd be nice if AutoSuggest could suggest plural form

Advanced search







Protemos translation business management system
Create your account in minutes, and start working! 3-month trial for agencies, and free for freelancers!

The system lets you keep client/vendor database, with contacts and rates, manage projects and assign jobs to vendors, issue invoices, track payments, store and manage project files, generate business reports on turnover profit per client/manager etc.

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