Hi all, I am new to Sphinx. I have read a lot of documents about it though. I
was able to run the demo (an4) using SphinxTrain and s3 flat decoder. Now I
want to make my own language model. My end goal is to build an application
that will perform word spotting. The dict is very small (20-25 words). I made
a language model with CMU SLM toolkit version 2. It works well for predicting
words that are in the grammar, but for out-of-grammar words it predicts
randomly, one of the words out of the vocbulary. I tried to make an open vocab
of type 2 but no luck. Can anyone help me out on this????
Any help will be greatly appreciated.
Thanks in advance.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
In particulary, you don't need a language model with a finite state grammar.
Also you need to use a pocketsphinx API to get confidence result and filter
the hypothesis.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
-- Please search the forum before asking. Questions like your one are asked
every week.
Please pardon the redundancy.
-- In particulary, you don't need a language model with a finite state
grammar.
Means???
Also, AFAIK Finite State Grammar is for command/control applications which is
not my case. What I would like to do is pick up a keyword from a long string
of words. Should I implement FSG??
Anyways, I will try PocketSphinx. Thanks for the pointer.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi all, I am new to Sphinx. I have read a lot of documents about it though. I
was able to run the demo (an4) using SphinxTrain and s3 flat decoder. Now I
want to make my own language model. My end goal is to build an application
that will perform word spotting. The dict is very small (20-25 words). I made
a language model with CMU SLM toolkit version 2. It works well for predicting
words that are in the grammar, but for out-of-grammar words it predicts
randomly, one of the words out of the vocbulary. I tried to make an open vocab
of type 2 but no luck. Can anyone help me out on this????
Any help will be greatly appreciated.
Thanks in advance.
Please search the forum before asking. Questions like your one are asked every
week.
https://sourceforge.net/projects/cmusphinx/forums/forum/5471/topic/3400265
In particulary, you don't need a language model with a finite state grammar.
Also you need to use a pocketsphinx API to get confidence result and filter
the hypothesis.
Thanks for the reply.
-- Please search the forum before asking. Questions like your one are asked
every week.
Please pardon the redundancy.
-- In particulary, you don't need a language model with a finite state
grammar.
Means???
Also, AFAIK Finite State Grammar is for command/control applications which is
not my case. What I would like to do is pick up a keyword from a long string
of words. Should I implement FSG??
Anyways, I will try PocketSphinx. Thanks for the pointer.
> The dict is very small (20-25 words).
The spotting is usually implemented with a finite state grammar like JSGF,
it's not only about command and control.