I am interested in using Sphinx3 to perform a phonetic word search, that is I would like to specify a phonetic representation of a word, then have Sphinx decode the input audio and return a list of occurrences of that word. Presumably this would be faster than generating a full transcript because hypotheses that do not match the target word (within a threshold) would immediately be discarded.
Is it possible to do this in Sphinx without significantly rewriting some source code?
Alternately, I have considered using allphone to generate a phonetic lattice, then matching up my search string with the output. The reasons for doing this instead of using the regular decode word lattice are that the phonetic lattice does not incorporate the language model, and that a phonetic representation of the search string allows for out-of-vocabulary words. Anyways, I have been unable to generate usable results from the phonetic lattice file (.phlat) because there are no scores generated. Is there a way to obtain these values?
Thanks in advance for the assistance.
Jimmy
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I am interested in using Sphinx3 to perform a phonetic word search, that is I would like to specify a phonetic representation of a word, then have Sphinx decode the input audio and return a list of occurrences of that word. Presumably this would be faster than generating a full transcript because hypotheses that do not match the target word (within a threshold) would immediately be discarded.
Is it possible to do this in Sphinx without significantly rewriting some source code?
Alternately, I have considered using allphone to generate a phonetic lattice, then matching up my search string with the output. The reasons for doing this instead of using the regular decode word lattice are that the phonetic lattice does not incorporate the language model, and that a phonetic representation of the search string allows for out-of-vocabulary words. Anyways, I have been unable to generate usable results from the phonetic lattice file (.phlat) because there are no scores generated. Is there a way to obtain these values?
Thanks in advance for the assistance.
Jimmy