from Julius book v3.2:

pass1_best: First pass best hypothesis word sequence (interim result).

pass1_best_wordseq: Same as above but as a N-gram sequence.

pass1_best_phonemeseq: Best hypothesis phoneme sequence ("|" separates words)

pass1_best_score: The hypothesis score (Log-likelihood)

sentence1: is the final recognition result.

--- (Edited on 7/22/2020 11:00 am [GMT-0400] by kmaclean) ---

Dear Team,

Thank you for the response..

I have 2 queries.

A.  

% julius -C jconf_filename -input rawfile

what is the raw file?

B.

pass1_best: <s> DIAL EIGHT
pass1_best_wordseq: 0 3 5
pass1_best_phonemeseq: sil | d ay ah l | ey t
pass1_best_score: -3953.129395
sentence1: <s> DIAL TWO </s>
wseq1: 0 3 5 1
phseq1: sil | d ay ah l | t uw | sil
cmscore1: 1.000 1.000 0.471 1.000
score1: -5293.808105

As i observe sentence1 gives most of the time good accuracy output.

It indicates as The 2nd pass uses the interim results from the first pass and searches these results using a 3-gram stack decoding technique.

How we can only display the 2nd pass results only.

Is there any document to understand sample.jconf script how we can add more commands or remove inorder to test the best feature.

Regards,

Pradeep Karunakar

--- (Edited on 7/23/2020 4:18 am [GMT-0500] by Pradeepkk) ---

>what is the raw file?

raw file has no header information.

an audio is just an array of numbers.  A wav file, for example, has header information that describes the audio.  A raw file does not have any header information and is just a series of numbers corresponding to the audio.

>How we can only display the 2nd pass results only.

just filter out results using whatever scripting language you are using when you call Julius - e.g. search for the line of output that says: "sentence1" and capture whatever is after that as your result

>document to understand sample.jconf script

the julius script contains command line parameters to julius.  The julius book link I sent you has all that info, so does the julius website.

--- (Edited on 7/24/2020 10:27 am [GMT-0400] by kmaclean) ---