OK, so here's a question. Isn’t LookAhead supposed to look ahead? And yet, with it switched ON, if I connect a large TM to my project (with 2 million TUs), and click Ctrl+Enter, I still have to wait over a minute for the results to appear in the Translation Results pane. What's going on here?
It is supposed to be... but I'd be surprised if we did a lot of testing with that many TUs! It's fairly unusual for file-based TMs to be that large I think even though it's possible. I'm building one with 1.3 Million (don't have enough resources readily available to build anything larger) right now and will test tomorrow (or later today!) when it's finished updating.
In the meantime can you share any other information about where your TM is stored for example?
I used a TMX with around 1.3 million TUs that wound up creating a TM with approx. 900k TUs. So this is around half the size of your TM. I was originally intended to record the performance of this but there seems little point now I have played with it for a while.
Lookup for fuzzy matches and also repaired matches takes no more than a second or two. Using LookAhead makes no difference at all in this context, probably because the TM is local on my hard drive and the lookup speeds are already fast. There is no lag from having to access a resource over the internet or on a shared drive on a network somewhere.
I think, to answer your question you will need to provide more information about the following:
- where is your TM stored?
- what type of documents are you translating... is it heavily tagged?
- is it only this project causing these kind of delays or all your projects?
Sorry Paul, meant to respond sooner but am in the middle of something large and urgent. Looks like I was mistaken, since I had/have been playing around with the RAT (aka Rapid Add Term), so I had actually disabled LookAhead at the point where I was wondering why it wasn't working. Duh.
So … LookAhead appears it works quite well with my 2 million-TU TM. Not quite 1-2 secs, but pretty close. This is on a local SSD.
But this does raise another question: it would be super amazing if you could make it so the RAT could be used with LookAhead switched on, since I already can't live without my RAT, but also don't want to switch off LookAhead, since I like working with my big (2 million-TU) Mothership TM.
Why not work much more efficiently? Simply create a project TM out of your huge big mamma, then it will do what you expect. A project TM fulfills exactly the same job in your project as the big mamma, if you select "Search project and main TM" in your project settings. This way matches come fast (without LookAhead) and if you need concordance, it is found in the big mamma too. Another advantage - fragment matching utilizes the whole TM, not just the project TM.
This is how I work since most probably the very beginning of my project work - I bought my first NAS back in January 2010 or 2011 and since then use always project TMs. The main TM is on SSD, the PTM on NAS. Works pretty well.
Very interesting. Thanks Jerzy! Am just preparing a nice big job following your suggestion, and will see how it goes...
Of course this changes a lot, as when the job is finished, you MUST run either "Finalize" or at least "Update main TM" task, otherwise all the translations done within this project will remain just there...
Another advantage of the use of project TM I did not mention before is, that you update your main TM with a finished - meaning proofread and corrected - translation. This way you protect your main TM from unnecessary errors.