Skip to the content ForoHackers
Log inJoin

We both transcribed the same word. Our annotators disagree about where the breath ended.

Natural language processingResearch softwareSpeech
#17225 · · Quote · Report

Original clip is me saying "four" after a breath. Both of us wrote FOUR. My span: 0.31 to 0.72 s. Colleague: 0.37 to 0.72. The first six hundredths slide from breath into f. If I force one number, the resulting agreement will be extremely affordable.

#17495 · · Quote · Report

À vitesse normale j'entends bien le mot entier. Au ralenti j'aurais envie de mettre trois frontières de plus. La consigne demande le début du souffle ou le premier bruit que j'attribue au f?

#17496 · · Quote · Report

that's the fork. she wants first recognisable consonant, I included the continuous noisy onset. playback at normal speed didn't reveal a little door slamming at either timestamp. we did agree where voicing came in later.

#17497 · · Quote · Report

Keep both spans attached to this recording, plus the difference in instructions. If you average to 0.34 you've manufactured the only boundary neither listener actually chose.

#17498 · · Quote · Report

did that. the transcript for someone reading still just says "four". the training set gets the disputed 60 ms and both raters' choices. nobody has to perform a theatrical intake of breath every time the text is read aloud.

#17504 · · Quote · Report

cuatro. ya está. he esperado un minuto de rigor antes de empezar, espero que no tengáis que volver a etiquetarlo todo.

You are not logged in, so this goes up anonymously: no account, no name and no address is stored with it, and it may carry no link and no image. Log in or join to post under your name. Our web server's logs keep the address and time of every request for up to 14 days; to keep yours out of them, use Tor. To reach the staff in confidence, use the drop. Sign as a returning source (a key kept only in this browser)

Sending takes a few seconds here: your browser first works out a proof of work against spam.

Reply