How far is the state of development of context sensitive translation memories?
Thread poster: Els Eerdekens

Els Eerdekens
French to Dutch
May 2, 2013

Dear all,

I have some questions about an article from Ziad Chama on this website:

1) He writes the following in his article: “Basically, these systems have been working on more or less the same principles for the past 15 years: source and target sentences are captured as pairs in a database. The pairs of sentences, or the so-called translation units, can be appended with additional information to create a specific context.”

I suppose the TMs still work the same way today?

2) In his article, he says the following: “If both target sentences are saved in the system, then these will be suggested while translating. The translator can decide in an interactive manner which of these suggestions is appropriate.”
Does such a TM, in which we have to choose between two suggested translations, already exists?

3) He also talks about the system STAR Transit where matching is done with the help of reference files.

What are reference files?

4) He also writes the following: “For more than ten years Translation Memory Systems have been splitting documents into segments, which usually correspond to a sentence. Current developments are moving towards storing the immediate environment of a translation unit as well in a segment. Occasionally, there are also discussions on whether it might be more useful to move away entirely from the classic segment.”

How is the current environment nowadays stored?
Does he mean the same if he talks about environment and context?
Does he refer to the ContextMatches of Trados Studio 2009, Déjà Vu and memoQ? Does Trados Studio 2011 still uses this ContextMatches?
Or are there any other developments?

5) And finally, he talks about the ideal situation for TM’s: “The ideal situation, therefore, the user should be able to configure his Translation Memory System in such a way that it saves larger or smaller semantic units depending on the requirement and type of text, namely, sections, paragraphs or individual sentences like before.”

Can users already configure the TM? I know you can change the alignment, in which you can take into consideration the context by aligning into paragraphs instead of single sentences.

Thanks a lot!


Direct link Reply with quote

David Haglund  Identity Verified
Local time: 07:40
Member (2012)
English to Swedish
Some replies Jul 22, 2013

1. Yes, this is the basis of most TMs I have heard of.

2. Yes, I have see this in both STAR Transit and Trados. If the TM has two 100% matches it will suggest both and you can choose one. This of course depends on the settings during the import, you can choose to force the use of the first 100% match or choose not to pretranslate and let the user decide interactively during the translation.

3. The principle of reference files is that you have a flexible TM. Reference files are basically previously translated projects (source and target files) and sometimes also aligned files. You use these for pretranslation and suzzy matching, just as you use a normal TM. But the advantages of reference files is the person doing the import can decide exactly what older projects to include. This makes the ref material more precise as you can choose to only use projects for the subject "monitors" and leave out projects having to do with "harddisks". You can also leave out projects you know are of inferior quality.

Since it is more likely that the reference files are more reliable when it comes to context, STAR Transit looks at the structure of the entire new document and compares it to the new document. It looks to see it the segment is a header, plain text or something else. It loks at the entire paragraph to see what segments surround the segment to be translated. If all things match up, the suggestion is more than 100% reliable. STAR Transit was first with this contextbased pretranslation principle, and Trados copied it, basically.

4-5. I don't know this. But I think context=environment here. I know that you have been able to segment on sentence or paragraph during import or while creating a new TM for a while, but that is not the same as storing semantic info.

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 »

How far is the state of development of context sensitive translation memories?

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 »
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 »

  • All of
  • Term search
  • Jobs
  • Forums
  • Multiple search