Court Reporters: Don’t Sweat Speech Recognition… Yet

There have been murmurings online about automatic speech recognition, AI, and machine learning technologies replacing court reporters.

The sad reality is this: Speech recognition will undoubtedly replace court reporters… eventually.

Automatic speech recognition (ASR) technology has made tremendous advances in recent years with the growing popularity of voice recognition features on smartphones and internet-connected appliances.

Is Speech Recognition 90% Accurate?

I read a blog post on the current state of ASR at 3Play Media, a tech-embracing closed caption, transcription, and subtitling company. The post mentions that ASR technology is at a point where it can be up to 90% accurate. But that 90% accuracy is IF

  • There is only one speaker
  • The speaker is reading from a script or is equivalently concise with virtually no grammatical or speech errors
  • The speaker is using a high-quality microphone and speaking at an appropriate distance from the microphone
  • There is little to no background noise in the audio, and
  • All the above conditions remain constant throughout the audio file.

If any one of those conditions is not met, accuracy suffers. If two or three of those conditions are unmet, accuracy drops to 50% — an error every other word.

And none of those conditions are met in a deposition or courtroom setting.

Add in speakers with heavily accented English and multiple speakers interrupting each other, and the obstacles to a speech-recognized transcript seem insurmountable.

Speech Recognition In Action

To get an idea of what a transcript might look like using today’s technology, try this:

clip art microphone
  • Get your friends together, and head out to lunch at the neighborhood Chik-Fil-A
  • Once everyone is seated, pull out your smartphone and open a notetaking app that supports voice-to-text dictation
  • Set the phone in the center of the table, and start the dictation
  • Keep everyone talking about something—anything. Try to involve everyone at the table
  • After about 3–5 minutes of lunchtime conversation, take a look at the “transcript” you’ve produced.

How accurately did it capture the conversation? Can you imagine having to format, scope, and proofread that mess?

Speech Recognition Is Coming

Nonetheless, interest in ASR research seems to have reached a critical mass. The race toward perfecting this technology is on. And as speech recognition evolves, so will the role of the court reporter.

Will the day come that automatic speech recognition replaces court reporters altogether?

The answer is yes, without a doubt, that day will come. But it is nowhere in the foreseeable future.

Share this:

Abbreviations
AP: Associated Press Stylebook
BGGP: Bad Grammar/Good Punctuation
CMOS: Chicago Manual of Style
GPO: U.S. Government Publishing Office Style Manual
GRM: Gregg Reference Manual
LMEG: Lillian Morson's English Guide for Court Reporters
MW: Merriam-Webster.com dictionary

Leave a Comment