Start a new topic

Request: Sane order of matches on matchboard

 Hi,
I'd like the matches displayed on the matchboard to be displayed in a sane order.

I'd like to have the most likely results (exact glossary matches, exact memory matchs, prefix glossary matches) at the top and the speculative ones (fuzzy and virtual memory matches [what's the difference?]) hidden away out of sight at the bottom.




2 people like this idea

After finally getting "Sort by appearance in source" as option is there any chance to get a sane order soon? This would include or mean:

Exact matches

Fuzzy matches

Terms

Fragments & segments

Machine Translation

or maybe:

Exact matches

Fuzzy matches

Terms

Machine Translation

Fragments & segments


I must confess I forgot AA above. I would set it between Fuzzy matches and Terms, perhaps.

> after dozens of speculative 'hits' (whatever that means).


It depends on your TM settings (e.g. "Prefix matching" option on for the given TM). You should be able to see the type of the match (e.g. Exact, Virtual, Prefix) in the Matchboard. A Matchboard screenshot of the issue would be clearer.

This still doesn't work as it should.

I'm set to 'Sort by quality', but I have exact matches from the project TM (priority set to high) right right at the bottom the (very long), after dozens of speculative 'hits' (whatever that means).


Also one comment on the fuzzy algorithm – from the frequency with which it decides that common words are matches for something completely spurious, I assume this looks only at positives, and takes no account of negatives.

I'll illustrate what I mean by an example:
My TM tab currently lists the following match (indeed it's the top match):
DE: DIese
EN: processing


The TM is one I use for data protection/privacy policy translation, so a large proportion of segments contain the term 'processing' and indeed 'diese', but there's no actual correlation between the two.


My impression (which may be mistaken) is that CT takes no account of the large number of segments which contain 'diese' but do not contain 'processing', which should cause it to discard this as a match.


Would it be possible to correct this behaviour – i.e. allow the presence of even modest numbers of non-matches to negate the presence of large numbers of matches. I appreciate that this would slow things down a little, but it would vastly improve the results.


Just re-animating this conversation:


The matchboard is unflexible and does not reflect the usual order. This would be:


Exact matches

Fuzzy matches

Terms

Fragments & segments

Machine Translation


or maybe:


Exact matches

Fuzzy matches

Terms

Machine Translation

Fragments & segments


Any chances to have a rather conveniently ordered Matchboard soon? Not to forget such a trivial thing as "Sort by appearance in source".  

Then if you use any of the recent CafeTran 2017 updates, you should have the expected Matchboard sorting - the exact fragment matches at the top (just below full segment exact and fuzzy matches). Note that with source duplicate matches, CafeTran groups them all, along with the exact fragment duplicate.

Hi Igor,
sort matches alphabetically is already unclicked on my system.

 

Right-click at the Matchboard and uncheck Sort matches alphabetically. Then CafeTran sorts by quality with source duplicate grouping.


Virtual matches are fuzzy fragments with higher probability of the accurate match. You can control this option via Edit > Preferences > Memory tab > Subsegment to Virtual match threshold.

Login to post a comment