Hi, I think giving the option to hide in Matchboard the numbers of the hits (they're still available via TM tab) can improve its readability, I still find the Matchboard cluttered and hard to quickly find things.
Another suggestion would be the way it sorts stuff, I find hard to identify terms in there so, why not sorting the entries by the position they appear in the segment AND /OR by category (e.g. MT, AA, Fuzzy, Terms, Fragments, Virtual...) in blocks?
Michael, you were right, Big_MAMA (i.e. Total Recall TM) is blaming (it's a whole export from memoQ) so I designed a new memory handling (more suitable for CT) and I will use Big_MAMA just for fuzzy, no hits, this helped a lot with accuracy.
By the way, now that I'm not using Big_MAMA for fuzzies, I copied your Memory preferences and they're way better :)
But there everybody will have his/her own preferences. The sort by quality is a good step, but there is still room for improvement.
You're right, every user will have its own preferences so it must be really hard for Igor to take design decisions.
I think it also has to do with the tool we come from, e.g., the company I work for uses Trados packages, even if they give me a Trados license I cannot work in there just for the side-by-side view (and for other stuff of course) of segments, it's something silly but it affects my productivity, in spite of I translate faster on Trados than on memoQ (Trados can pick things from MT to autopick, and I think this is quite great! and the way it shows stuff in autopick its very good too, as they're marked as the kind of element they really are, fragments, autosuggest, concordance...).
The problem arises when I try to do stuff the way I used to do in memoQ (which has been my main tool for years), things, concepts, and routines are SO different in CT that we need to adapt instead of trying to imitate our own routines with other tools (and that's what I've been doing since I've met CT).
So I will again try to get used to the CT way before I give up, but will take my productivity into account at all moment.
@Victor: You might also want to try switching off Look up word stems (in Memory settings). I find it can get a bit too fuzzy (thus producing more not so useful noise) with it on.
I own licences for the latest versions of SDL Studio, memoQ and Déjà Vu, but always end up coming back to CafeTran as it helps me work much faster. Indeed, SDL Studio's auto-complete/suggest implementation is very good, however the rest of Studio's terminology system (adding/finding specific terms, etc.) is utter crap. For example, every time you add a term, the entire UI jumps all over the place and jiggles. Not to mention the whole program is riddled with bugs.
Don't give up on CT; I promise you that once it clicks, there will be no going back!
PS: I have 32GB of RAM (and 3 SSDs), but still have Switch to preliminary matching threshold set to 50000. I tried changing it to something really high, so Preliminary matching wouldn't be used, but found it took way too long for the Matchboard to populate itself when moving from segment to segment. I would recommend using low value here so it is used, as this will make moving from segment to segment much faster once it's ready.
By the way, when working with large TMs, you can also do: Memory > Export > Export preliminary matching memory. This is very cool as it allows you to export just what CT found to be useful to a smaller TMX, which you can then use instead of the big one. This isn't always a good idea as you lose several things (e.g. if you use CT for concordance on large TMs), but can be a very useful tool in other specific scenarios.
Also, if you are getting too much noise, try switching your individual TM settings:
Right-click on TM pane: Options > Matching type: Fuzzy (instead of Fuzzy & hits)
And about Total Recall, I would recommend not using it at all, until you are more familiar with the CT. I'd suggest starting with TMXs only, and later experimenting with it if you need it. I actually don't use TR at all these days, strictly TMX, and am very happy this way. However, I should add that I do all TMX collecting/maintenance/pretranslating with LogiTerm, but that's another story...