Duplicate sentences during validation

Common Voice
#1

Is anyone else noticing duplicate sentences during English validation? There are over a million sentences in the wiki dump alone so there should be enough unrecorded sentences to present to users without needing to repeat them.

Interestingly, they are always Sentence Collector sentences, never wiki sentences. One example is the sentence “The mouse lived in a small hole in the skirting board”, which I have been hearing over and over again by different speakers over the past few days.

I have a theory that this is mainly affecting users who have recorded a large number of sentences but I don’t have any data to back that up. It’s hard to provide an actionable bug report when I don’t fully understand the behavior that’s occurring.

Has anyone else noticed this and spotted any patterns?

P.S. There’s a Github issue here: https://github.com/mozilla/voice-web/issues/2128

CC @Codigo_Logo_Programacao_e_Inteligencia_Artificial @Michael_Maggs

#2

I’ve heard a couple.

#3

I’ve noticed that, during validation and recording as well.

#4

I hear it sometimes during validation. I see it a lot during recording. Usually the repeated instances are close together, 4-6 items apart.

dabinat:

I have a theory that this is mainly affecting users who have recorded a large number of sentences but I don’t have any data to back that up.

FWIW, I’m one of the high-volume recorders.

#5

It may be that someone recorded the sentence, but a validator clicked “No” for it, so it immediately went to another recorder. This conceivably could happen several times, if a sentence was “unlucky” enough to be badly recorded several times in a row.