Problems with Arpa model made with cmuclmtk

[Lyle Pratt]

I have created an arpa model with cmuclmtk. When I try to use it with
pocketsphinx I get this error every time:

ERROR: "ngram_search.c", line 221: Language model/set does not contain </s>, recognition will fail

I've been through the process dozens of times and always end up with that
error.

Can anyone tell me why this is happening?

Thanks!
Lyle

[Jake]

I guess the transcripts you are using to build LM don't have both and
at the start and end of each utterance.

[Lyle Pratt]

All of my transcripts do indeed have both and . That is not the
problem.

[Jake]

I'm sorry. If does exist in the .ARPA file, I have no clue at this point.

[Nickolay V. Shmyrev]

Maybe you need to follow the tutorial closely

http://cmusphinx.sourceforge.net/wiki/tutoriallm#arpa_model_training

I suggest you to share your language model so we can check ourselves if
is present there. For example case could be the issue. will not work.

[Lyle Pratt]

Thanks for the help and suggestions guys.

I'm just going to use mitlm. It seems to be simpler to use and I have not had
any problems with it so far.

[tina_gong]

hi, lylepratt
i met the same problem ,what can I do?

ERROR: "ngram_search.c", line 221: Language model/set does not contain ,
recognition will fail

[tina_gong]

though there is no answer for me，but at last,i find the answer.
i was so Careless。
when i generated the .lm.DMP,I missed the blank in the .txt.
just like that
go should -> go