I am having trouble to convert my grammar, in order to use it with Julius.

I used the following commands :

HLStats -b bigfn -o wordlist wrd.mlf

HBuild -z -n bigfn wordlist grammar.slf

So my grammar.slf is about 1.2 megabytes, I have ~6200 words in my dictionary (it is the TIMIT corpus).

The first stage of the SLF2DFA.SH script, remove_null.pl, generates a 900mb file. I find this to be quite large, especially considering the name of the script.

Then, there is the second step : moore2mealy.pl.
I tried running it on a supercomputer with 64gb of ram and it filled them all (and was subsenquently killed by the linux OOM killer), so I guess this is very wrong. 

Any help ? Thanks. 

--- (Edited on 6/27/2013 2:23 am [GMT-0500] by ) ---