I've been doing a bit of reading, and I understand that Sphinx has been used very often with the Pheonix parser. I have a couple of questions about this. First, are there any plans to release the source for the Pheonix parser? Second, was there any kind of interaction between the parser and the word recognizer? In particular, was the parser used to correct mistakes, help select the most probable word or to supervise on-line learning in the recognizer? Third, will some of the interface of Sphinx be opened up so that _any_ parser may help with error detection or on-line learning in the recognizer?
Thanks,
Amit Dubey
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Phoenix has (and still is) used with Sphinx on certain projects here
at CMU and elsewhere, but we have some issues still about releasing
it. We may have to create another parser if we don't manage to
get everything together for Phoenix to release it.
Phoenix is a parser that uses semantic fragment grammars. With it,
one can, for instance, take the top N possible word strings from
sphinx2 and find parses that seem to work, and reject some. But
we don't have this tool to give out widely at the moment. I'll make
a point of asking the people who ultimately will make the decisions
on this component again.
kevin
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi,
I've been doing a bit of reading, and I understand that Sphinx has been used very often with the Pheonix parser. I have a couple of questions about this. First, are there any plans to release the source for the Pheonix parser? Second, was there any kind of interaction between the parser and the word recognizer? In particular, was the parser used to correct mistakes, help select the most probable word or to supervise on-line learning in the recognizer? Third, will some of the interface of Sphinx be opened up so that _any_ parser may help with error detection or on-line learning in the recognizer?
Thanks,
Amit Dubey
Phoenix has (and still is) used with Sphinx on certain projects here
at CMU and elsewhere, but we have some issues still about releasing
it. We may have to create another parser if we don't manage to
get everything together for Phoenix to release it.
Phoenix is a parser that uses semantic fragment grammars. With it,
one can, for instance, take the top N possible word strings from
sphinx2 and find parses that seem to work, and reject some. But
we don't have this tool to give out widely at the moment. I'll make
a point of asking the people who ultimately will make the decisions
on this component again.
kevin