How to control homograph matches from Multiterm in Studio

This is a question that straddles MT and Studio, not sure which subforum is better...

So say you have homographs in your MT, like down (feathers) and down (adverb). Currently my entries look like this

down (feathers); ID 10505; target(s): clòimhteach, chlòimhteach, clòimhtich, chlòimhtich (they're inflectional forms)

down (adverb); ID 4817; target(s): sìos 

I do require both of them in the same termbase, and both occur in my TM but of course the adverb is more common that the feathers. However, Studio has this irritating habit that when it has a close match (for example if the TM contains a translation for "Move your fingers up on the touchscreen") to select the feathers, rather than the more common adverb i.e. say the new string is "Move your fingers down on the touchscreen", it will tweak the string but invariable with the word for the feathers.

It can't be based on occurrences, or it would pick the adverb. It's not the ID either, as the adverb has a lower ID than the feathers. My guess is that it defaults to picking the first alphabetical option but that's rather ham-fisted and ideally I'd like to find a way of making it suggest the homograph of MY choice, not its choice. 

Any ideas?

emoji
Parents
  •  

    Perhaps the solution here is to use AI?  In Trados Studio 2022 you have AO Professional, and in the upcoming 2024 version you'll have Trados Copilot - AI Assistant.  Both of these can use your termbase to make sure your translations are using the correct terminology.  But since it's AI it can also make sure that the correct form is used even if it's not actually written that way in your termbase.

    Perhaps worth playing with a little?

    Paul Filkin | RWS Group

    ________________________
    Design your own training!

    You've done the courses and still need to go a little further, or still not clear? 
    Tell us what you need in our Community Solutions Hub

    emoji
  • Sorry, but anything to do with AI and under-resourced languages won't fly. Ever. One word: data paucity. You invariably end up with stuff so bad you spend more time proofreading than translation would have taken in the first place.

    I feel that AI is being used as the lazy answer to almost everything at the moment, including stuff that could be more reliably fixed by some clever thinking. It's just kicking the can down the daisy chain in a very frustrating way because the answer I get from AI devs about fixing something is almost invariably 'send more data and maybe it'll work' (or permutations of that) because nobody really seems to know how or why AI does what it does.

    emoji
  •  

    One word: data paucity.

    One word?

    Just kidding... I do understand where you're coming from with this but I would be very interested to see what you thought if you were to give this a try.  I'm using the Trados Studio 2024 version of this as it is much improved (and will be available very soon) and hope you might see this does have some benefit, even for under-resourced languages.  I don't think this is a lazy answer to things... rather I think this is a very clever solution that will only be using your terminology to check whether the translated content uses the correct form of a word/phrase.

    I certainly wouldn't say never... and in fact I really hope this isn't the case as I see AI as one way of saving many endangered languages from extinction and ensuring we have a means of using them and teaching them in the future.  Anyway... have a look see whether you think this could be something that might have some value in helping you to solve your homonym problems?

    Paul Filkin | RWS Group

    ________________________
    Design your own training!

    You've done the courses and still need to go a little further, or still not clear? 
    Tell us what you need in our Community Solutions Hub

    emoji
Reply
  •  

    One word: data paucity.

    One word?

    Just kidding... I do understand where you're coming from with this but I would be very interested to see what you thought if you were to give this a try.  I'm using the Trados Studio 2024 version of this as it is much improved (and will be available very soon) and hope you might see this does have some benefit, even for under-resourced languages.  I don't think this is a lazy answer to things... rather I think this is a very clever solution that will only be using your terminology to check whether the translated content uses the correct form of a word/phrase.

    I certainly wouldn't say never... and in fact I really hope this isn't the case as I see AI as one way of saving many endangered languages from extinction and ensuring we have a means of using them and teaching them in the future.  Anyway... have a look see whether you think this could be something that might have some value in helping you to solve your homonym problems?

    Paul Filkin | RWS Group

    ________________________
    Design your own training!

    You've done the courses and still need to go a little further, or still not clear? 
    Tell us what you need in our Community Solutions Hub

    emoji
Children