Is Voice Recognition Software (VRS) getting good enough that it could soon replace court reporters? Perhaps another way to look at it, is the VRS error rate starting to approach the human error rate? Or are court reporters and their little machines going to be a fixture of courtrooms for the next 100 years?
I think that in most courts now, the audio recording is itself considered the official record. No need for a human or computer program to transcribe it.
Really, is that true? So all of those “Can you please read that back?” remarks on “Law and Order” are all bogus? TV has let me down yet again…
I find that hard to believe, frankly. While an audio record could be the ultimate official record, I’d think transcripts which were searchable would be absolutely necessary for filing and future reference purposes. Certainly, all the court records I’ve seen have transcripts, or redacted transcripts available for perusal.
IANAAttorney, however. So I don’t know what the finely honed legal experts are using these days to research cases.
To answer the OP, if Dragon Naturally Speaking represents the cutting edge of VR right now, then no, its error rate is not yet as low as human transcriptionists/recorders. At least not when I dictate, it isn’t.
I think that in the courtrooms I’m thinking of, a transcript is only produced if one of the parties wants (and is willing to pay for) it.
Last I checked it had plateaued somewhere between 80 and 90% for general recognition (not domain specific)
Humans use context to distinguish between similar sounding phrases, it will probably be very slow progress to approach human levels