How to use SRILM to build a language model?

[rezaee]

I have downloaded SRILM and did make world in it's main directory successfully(without errors),Also i tried make test without errors.

I fount there is a file named ngram-count in this path../srilm-1.7.2/bin/i686-m64 then I copied it to inside my etc file inside my database, and tried to run this command as cmusphinx tutorials say here https://cmusphinx.github.io/wiki/tutoriallm (I copied all my words and sentences inside train-text.txt file) :

./ngram-count -kndiscount -interpolate -text train-text.txt -lm your.lm

But I get this result:

one of modified KneserNey discounts is negative~~~

error in discount estimator for order 2
~~~

[rezaee]

I tried to change my command to :
ngram-count -text train-text.txt -lm your.lm

And got this result:

warning: discount coeff 1 is out of range: 0
warning: discount coeff 4 is out of range: 2.60909
warning: discount coeff 5 is out of range: 1.21455
warning: discount coeff 6 is out of range: 1.17879
warning: discount coeff 7 is out of range: 1.15325
warning: count of count 8 is zero -- lowering maxcount
warning: discount coeff 3 is out of range: -0.0630631
warning: count of count 4 is zero
warning: count of count 5 is zero
warning: count of count 6 is zero

I don't know is this normal and works fine or not? if not, whats the problem?

[Nickolay V. Shmyrev]

It works as expected.