emjotde / eserix Goto Github PK

Hi there,

I have found a strange behaviour while debugging the SUMMA MT module. Eserix depends on capitalization being correct. Is this a bug? Example below. This is a problem because we do not have correct capitalisation in ASR output. But we could get the punctuation module to uppercase all words following an inserted full stop and leave eserix alone.

abmayne@sif:~/experiments/summa/hackathon17/summa-mt$ /mnt/skaol0/lexi/experiments/summa/hackathon17/summa-mt/eserix/build/bin/eserix -r /mnt/skaol0/lexi/experiments/summa/hackathon17/summa-mt/eserix/srx/rules.srx -l de -t
(INPUT) this is a test with a long sentence . And another test . He He he . this is a
(OUTPUT - correct) this is a test with a long sentence .
(OUTPUT - correct) And another test .
(OUTPUT - incorrect) He He he . this is a

(INPUT) this is a test with a long sentence . And another test . He He he . This is a
(OUTPUT - correct) this is a test with a long sentence .
(OUTPUT - correct) And another test .
(OUTPUT - correct) He He he .
(OUTPUT - correct) This is a

Hi Marcin, we have experienced some problems with building eserix. The solution was to add "std::" before "string&" in regexp.hpp lines 83, 91
Cheers,
Marcin