is there an API to get latest wakeword when there is a false positive wakeup, in order to retrain mycroft-precise model?
If i recall correctly precise-listen in conjuncture with the
-s -d flag (save - false positives - to directory; you have to dig into the code for this) from mycroft-precise has this functionality.
The model is not updated with every false activation. It’s done on a semi-regular rotation, and the code checks for a newer version on startup.
As i understand he wants to retrain the model by himself
You can turn on wakeword saving to keep local copies for things like this.
thank you for replying
but I want to say after false activation for example “bad boy”, then collect latest activation command and put to “not-wake-word” folder in order to retrain model
that’s exactly what I’m looking for