Services like Google Translate are all the rage nowadays. The technology is able to listen to words and determine, to a degree, what the individual is saying and respond back with the requested information. However, is this service actually better than a live individual who is transcribing what they hear? To put that to the test, a recent study used two professional transcriptionists and pitted them against IBM Watson, known as the smartest machine on earth, in order to transcript song lyrics. How the two faired might surprise you.
In the test, IBM Watson was disconnected from the Internet. With an Internet connection, the computer could just access the lyrics and instantly know each and every word, which would prove detrimental to the test. To test the quality of the transcript services, a variety of musical styles were played. This included everything from Taylor Swift and Van Halen to Elton John and Brad Paisley.
As anyone who is familiar with singing and music can agree with, many singers do not over pronounce words and sometimes the words can string together. It may take a few times to listen to the song to fill in all of the lyric gaps. However, in the test, both the professionals and IBM Watson had one chance to listen to the lyrics and neither had heard the songs before.
In every case, the professional transcriptionists recorded the lyrics correctly, without an error or word missing. However, IBM Watson did not do as well. For instance, in Elton John’s “Tina Dancer,” the computer missed three words and had eight total errors in just a single verse. This trend proved to be no fluke as the professional transcriptionist nailed Brad Paisley’s “Ticks” while Watson missed five words and committed 12 errors.
So what exactly does this show, other than IBM Watson is not able to always pick up slurred together lyrics (especially with music playing with the singing)? There is a human element to listening to music. The human brain is able to fill in what the ears are unable to pick up. While Watson is extremely intelligent and houses more actual information than the human brain can ever store, it does not have the human element for listening to other humans, which is ultimately what led to all of the missing words and errors.
For anyone who is looking for a professional transcription service, it is necessary to look towards the pros for this kind of work. Google Translate and other similar forms of technology are great for just general, at home use, but when it comes to quality transcription, nothing is able to beat the ears and the mind of a human performing the task.
IBM Watson Against Two Transcriptionists