On Tue, Jul 21, 2009 at 1:00 PM, Peter Klügl<[email protected]> wrote: [cut] > > The implementation of JAPE should be faster than TextMarker, but I don't > know how fast the bridge is. In contrast to JAPE, TextMarker is still in an > early project phase. JAPE is well documented and tested, TextMarker not > (yet). >
I wrote my own bridge for JAPE since the one in GATE 4.0 was compatible only with IBM-UIMA. It's a custom, very custom, ad targeted to our type sytem. This bridge map to JAPE and back to UIMA a lot of annotations, so there are a lot of objects to be garbage collected for each document. For comparison a 3-thread-pipeline with jape-uima-bridge inside poresses 80/90000 docs per hour. My docs are small: < 5kb. The same pipeline without jape goes to 140k doc/h. Cheers, R. -- Roberto Franchini http://www.celi.it http://www.blogmeter.it http://www.memesphere.it Tel +39-011-6600814 jabber:[email protected] skype:ro.franchini
