For instance, Elton John’s Tiny Dancer was played for both IBM Watson and an accomplished transcriptionist. In only three lines of text, the computerized program transcribed lyrics containing eight errors and three missing words. Meanwhile, Schenae, professional transcriptionist, delivered the same lyrics transcribed with no errors and no missing words. Because many computerized transcription program produce texts based attributes like tone, cadence, and duration of given sounds, errors are more likely to occur. A trained transcriptionist, on the other hand, has the ability to contextualize any given lyric in order ensure that both the words and lines make sense logically. For example, one error that the computer program yielded was, “Bue gene bear bear yeah,” instead of, “Blue-jeaned baby”---only the first line of text in the sample. Another key example is the subsequent phrase, “LA lady,” which was transcribed by the computer as “Lazy lately yeah.” In this way, it is clear that human beings consistently outperform the computerized program, and that more fine-tuning is necessary to establish a computerized transcriptionist service which can deliver the same quality of results as human beings.
Another professional transcriptionist, Colby, performed equally well compared with IBM Watson. To be sure, the computer program yielded fifteen missing words and nine errors in only four lines of lyrical text, while the same segment was transcribed by Colby with zero missing words and zero errors. For perhaps the most compelling evidence, we need only look at the third line of the IBM Watson transcription, which should have read, “Least I don’t need to beg or borrow.” Instead, the computer program produced, “The startled me truck bed though barlow.” As evident by this example, the computerized transcription is not only inaccurate by nonsensical, would do nothing but misinform the reader as to the true lyrics of the song.
As this study clearly shows, when comparing the abilities of a computerized transcription service, IBM Watson, with two highly skilled transcriptionists, the latter greatly outperformed the former. In this way, it is clear that technology still lags far behind when compared with human intelligence, and we should take this into account moving forward.
