Special case use of OpenEars

[antonjazz]

Hello,

I’ve thoroughly enjoyed experimenting with OpenEars. But my need is rather different than the usual cases (conversational English) and the special cases OpenEars seems very well equipped for (specific vocabularies and grammars).

My need is simple: to listen to the user attempt to pronounce a written word and assess whether they’ve succeeded in speaking the word correctly. I can set up OpenEars with a one-word vocabulary, but OpenEars seems to be rather lenient recognizing the word (it gives false positives) and the recognition score doesn’t seem to be a workable grade to assess pronunciation.

I also experimented by putting several near-miss words in the dictionary, hoping that they would catch unsuccessful attempts at the target word; unfortunately they also caught successful pronunciations of the target word, creating false negatives.

If anyone has ideas for using OpenEars, or Sphinx in general, for single-word assessment, I’d love to know.

thank you!
Anton

[Halle Winkler]

Hi Anton,

Sorry, this is probably not a great match for OpenEars as a use case, but I recommend taking a look at the FAQ since it has some info about rejecting wrong hypotheses: https://www.politepix.com/openears/support. Unfortunately there isn’t really bandwidth at the moment to moderate a larger discussion about Sphinx, so I will close this up, but I recommend searching previous forum topics for “OOV” and “out of vocabulary” and “keyword” and/or “wake word” recognition to learn more about the options here – there should be lots of existing info about similar applications once you know the phrases to search for.

[Author]

Posts