ml-stuttering-events-dataset
ml-stuttering-events-dataset copied to clipboard
apple
Annotation
Could you please help me with the label files.
https://raw.githubusercontent.com/apple/ml-stuttering-events-dataset/main/SEP-28k_labels.csv What does the numbers 0, 1, 2,3 mean here? because it is confusing Could you please clear it bit
For example Consider this row: Show | EpId ClipId | Start | Stop | Block | SoundRep | WordRep | Prolongation | Interjection| NaturalPause| HeStutters | 1 | 5 | 28390400 | 28438400| 0 | 0 | 3| 0 | 1 | 0 |
What does these number mean here
Thank you
The number refers to how many annotators (out of 3) labeled a clip with the given label type. e.g., In the line you included all three annotators said there was a sound repetition and one person thought there was a prolongation.
In general we assume that a label is correct if at least 2 (of 3) annotators chose that label.
Please explain little bit of https://github.com/apple/ml-stuttering-events-dataset/blob/main/fluencybank_labels.csv ... annotation is clear but (start ,stop )can't guess .. Please explain if possible
Hello, sorry I was on vacation last week. The units here are audio samples at 16,000 hz. Each clip is 3 seconds long. So in the first row the audio clip starts at sample 88,960, ends at sample 136,960, and one (of three) annotators labeled the clip with a 'block'.