I was recently sent this fascinating infographic that pitted IBM’s Watson against transcriptionists trying to transcribe song lyrics. It’s pretty fun to see the results and how they compare. However, I’m not sure it’s really fair to call it IBM Watson. It looks like they used the voice recognition software in Watson to do the voice to text. If they were really using Watson, then they would have done the voice to text and then compared it against all the song lyrics ever written and then produced the exact song lyrics. However, they said that the transcriptionists weren’t allowed to Google the lyrics. I assume the same was true for Watson.
It’s also worth mentioning that song lyrics can be much harder for a computer to hear than just a straight transcription. I also bet my friends at Nuance would argue that their voice recognition would have performed much better. Plus, I wonder how long the transcriptionists took on these songs. A double or triple checked transcription would be very accurate.
All of this said, this infographic and the results of their experiment (specifically stated as non-scientific) illustrates what most of us that have used the technology already know. Transcriptionists are more accurate than voice recognition. The question most people in healthcare make is whether the costs of a transcriptionist is worth that increase in accuracy. That’s the challenge.
Ok, without further ado, here’s the infographic: