In 1954, the American author, Sylvia Wright, invented the term “mondegreen,” which means substituting words for others that sound nearly the same as the correct ones. Scientifically the creation of a mondegreen comes from words that have near-homophony. This happens frequently in song lyrics.
Here is a quick fun challenge.
Try to recognize the songs from these mondegreens of their lyrics:
We are living in a Cheerio world and I am a Cheerio girl.
Police have a dog.
Hit me with your pet shark.
There’s a baboon on the rise.
Slow walking Walter, the fire engine guy.
Material Girl by Madonna, as in “We are living in a material world and I am a material girl.”
This is “Feliz Navidad” by Jose Feliciano.
Pat Benatar actually sang, “Hit me with your best shot.”
Credence Clearwater Revival warns in their song, “There’s a bad moon on the rise.”
“Smoke on the water, fire in the sky,” is what Deep Purple sang.
The Human Versus Machine Transcription Smack Down
IBM used a computer named “Deep Blue,” which took on the challenge to see if a computer could beat a human player in chess. The computer won, but only because it had the advantage of being able to calculate the millions of possible moves to find the one with the highest probability of succeeding. That is lots of brute force math, which machines can do faster than humans, but what about something more subtle…
A new more difficult challenge was raised to see how advanced voice recognition software compares to the high-level skills of a human transcriptionist.
The contestants were:
1) In this corner, professional transcriptionists and in that corner…
2) IBM’s Watson system, which claims to be the “Smartest Machine on Earth.”
What happened was that the IBM Watson system got its silicon chips kicked!
The results are shown in this infographic.
Schenae and Colby who are both professional transcriptionists had zero errors and zero words missing. IBM Watson had 33 errors and 29 words missing.
The human professional transcriptionists were able to transcribe the lyrics perfectly from a variety of songs, with zero mistakes. IBM Watson was not able to do any of them correctly. The results show that clearly the intricacies of the English language, not only includes the words, what they sound like, but more importantly the meaning of the words and how they fit in the context. At this moment, humans are far superior in this skill than machines are.
Human professional transcriptionists have no need to fear that computers will replace them any time soon. What transcriptionists do is exceptional, in terms of the computing power needed to replicate their skills.