Decoder

[treena908]

Hi,

in the paper, you said that, you haven't trained any markov model rather used hard coed probabilities. how did you specify the probability values and what is the transition diagram (is it same as in the 2015 Interspeech paper )? I looked into the deep_disfluency/decode/models directory and found the state diagram with probability for complex disfluency tags? What the tags rpSE and rpSM means? Is there any file I should look into to understand it better?