Adapting vs Training

[Nickolay V. Shmyrev]

Yes

[Zain]

Is there any link or tutorial how to employ "any-word" language model in CUM PocketSphinx.

[Zain]

Is there any change that I have to make in sphinx_train.cfg file to change from statistical language model to JSGF grammar?

I have the following error when trying to use JSGF grammar using the command:
pocketsphinx_batch.exe -hmm xyz -cepdir feat ... -lm etc/hello.jsgf ...

INFO: ngram_model_trie.c(354): Trying to read LM in trie binary format
INFO: ngram_model_trie.c(365): Header doesn't match
INFO: ngram_model_trie.c(177): Trying to read LM in arpa format
INFO: ngram_model_trie.c(70): No \data\ mark in LM file
INFO: ngram_model_trie.c(445): Trying to read LM in dmp format
ERROR: "ngram_model_trie.c", line 458: Wrong magic header size number 234a5347: etc/hello.jsgf is not a dump file

[Nickolay V. Shmyrev]

The option for jsgf is -jsgf, not -lm.

[Zain]

Is there any option to use only 1-gram or 2-gram of the 3-gram language model during decoding.

[Nickolay V. Shmyrev]

No, you have to modify the model outside for example with srilm:

 ngram -lm text.lm -order 1 -write-lm unigram.lm

[Zain]

Does CMU Sphinx support cross-word triphones as HTK?

[Nickolay V. Shmyrev]

Actually cmusphinx does not support within-word triphones, only cross-word ones.

[Bhiksha Raj]

We have four types of triphones --

a) word-internal (marked "i" in the mdef files)
b) begin-word (e.g. the triphone B (AH, AE) in the word-pair "A BAD
..". Marked "b" in the mdef file)
c) end-word (e.g. the triphone B (AE, AH) in the word-pair "GRAB A
..". Marked "e" in the mdef file)
d) single-phone-word (e.g. from the words "A" and "I", where the
entire word is only one phoneme, so all triphones are both-begin word
and end-word triphones. e.g. the triphone "AY (M, AE)" in "SAM I
AM". These are marked as "s" in the mdef file).

b,c and d are all cross-word triphones.

-Bhiksha

On Fri, May 19, 2017 at 2:57 PM, Zain zainzainzain3@users.sf.net wrote:

Does CMU Sphinx support cross-word triphones as HTK?

---

Adapting vs Training

---

Sent from sourceforge.net because you indicated interest in
https://sourceforge.net/p/cmusphinx/discussion/help/

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

--
Bhiksha Raj
Carnegie Mellon University
Pittsburgh, PA, USA
Tel: 412 268 9826

[Zain]

Could you please let me know what is the difference between cmuclmtk and srilm for creating language models.

[Zain]

HTK recognizer does differentiate between small letters and capital letter characters. For instance, the phoneme /B/ is different from phoneme /b/. Does CMU sphinx consider /b/ different from /B/ or it considers same phoneme?

[Nickolay V. Shmyrev]

For instance, the phoneme /B/ is different from phoneme /b/.

Same here

[Zain]

By "same", do you mean that /B/ and /b/ are same in Sphinx, or do you mean that Sphinx is same as HTK to handle them as two different phonemes?

[Nickolay V. Shmyrev]

Yes

[Zain]

I have the following error:
...
Decoding 202 segments starting at 1812 (part 10 of 10)
0%
Aligning results to find error rate
word_align.pl failed with error code 65280 at /usr/local/lib/sphinxtrain/scripts/decode/slave.pl line 173.

Any help.