search

Shahabks / myprosody

A Python library for measuring the acoustic features of speech (simultaneous speech, high entropy) compared to ones of native speech.
https://shahabks.github.io/myprosody/
MIT License
232 stars 63 forks source link

Models throw error when used with .predict #15

Closed ma7555 closed 4 years ago

ma7555 commented 4 years ago

KNN and SVC models are not working, other models .predict returns a symbol 'c', not sure what is it exactly.

I also want to ask how was these models trained?

Shahabks commented 4 years ago

to get KNN and SVC work on a local machine, you need to train the models on your machine (apparently the pre-trained models won't work on other machines) ...... Symbol "C" refers to one of the spoken language proficiency level

ma7555 commented 4 years ago

can you explain how was these models trained? which dataset was used?

Shahabks commented 4 years ago

the original algorithm was developed for a client who provided us with 9000+ audio files of non-native English speakers (from the archive of its students who being prepared for exams over a few years). We used those audio files + their associated human-rater scores to train the models

ma7555 commented 4 years ago

understoond, thanks

shalabhsingh commented 3 years ago

What are the different spoken language proficiency levels, from good level to bad level?