Paul Dixon - 2013-12-19

Seems like something is not right before running the fstrmepsilon command.
Trying looking at the outputs of fstcompile and arpa2fst with fstinfo. You
could also
pipe the the intermediate fsts to fstconnect and look for fsts that become
substantially smaller.
I think one of the intermediate scripts might be removing an essential
transition that allows for
complete paths to the final state.

On 19 December 2013 13:59, Florent florentmn@users.sf.net wrote:

Hello,

I'm trying to work with the following LM and use it in kaldi :
http://files.kde.org/accessibility/Simon/lm/

I also have my own 144k dictionary

I use the following scriptlet from voxforge recipe :

cat $lmdir/lm.arpa | \ grep -v ' ' | \ grep -v ' ' | \ grep -v ' ' | \ arpa2fst - | fstprint | \ utils/remove_oovs.pl $tmpdir/oovs.txt | \ utils/eps2disambig.pl | utils/s2eps.pl | fstcompile --isymbols=$test/words.txt \ --osymbols=$test/words.txt --keep_isymbols=false --keep_osymbols=false | \ fstrmepsilon > $test/G.fst

Evrything goes fine until last step :
Output of fstcompile is 524MB, but output of fstrmepsilon is 66 bytes

-rw-r--r-- 1 root root 549117910 Dec 19 13:23 fstcompile.out-rw-r--r-- 1 root root 66 Dec 19 13:25 G.fst

No error output from fstrmepsilon

Any idea what is wrong ?

Thank you

fstrmepsilon removes everythinghttps://sourceforge.net/p/kaldi/discussion/1355348/thread/5f7c28c7/?limit=25#7d37

Sent from sourceforge.net because you indicated interest in
https://sourceforge.net/p/kaldi/discussion/1355348/

To unsubscribe from further messages, please visit
https://sourceforge.net/auth/subscriptions/

 
alternate