Blog chevron right Transcription

Manual Transcribing vs. Automatic

Daniel Chang
Daniel Chang
Posted in Zoom Jan 12 · 14 Jan, 2016
Manual Transcribing vs. Automatic

In the age of information, where one is surrounded by easily accessible and affordable software, it is no wonder that words “computerized” or “automated” are more and more associated with “fast”, “cheap” and sometimes even  “better”. The previous implies that the human input is no longer necessary for some of the tasks. However, unlike some sci-fi films about AI takeover suggest, audio transcriptions is a field where a significant amount of time will pass before automatic transcriptions will surpass the manual ones. Let’s have a look at a brief comparison between the two.

1. Accuracy. Although manual transcriptions might take more time, mostly one can be sure that the accuracy of the manual one will be higher. Of course, it would be a good idea to do your own research on transcription providers and their transcription quality.

2. Word omission. While it is true that manual transcriptions contain “inaudible” and “unintelligible” parts, but imagine if your transcription would contain only 1/3 of the speech of your audio file. There are several reasons why such omission is more likely to occur when using automatic transcriptions. 

  • Although it is possible to get a good transcription when using a software, the problem is that it requires a high quality audio file. This means that if there are some background noises, overlapping dialogues and other obstacles, you will end up with a text that has a ton of gaps in it. For this reason, automatic transcriptions are best for monologues in good audio quality.
  • Another problem is that automatic transcription softwares do not deal well with accents or dialects, which again means that some of the words will either be misspelled or omitted.
  • When it comes to such linguistic intricacies as coarticulation (overlapping sounds), assimilation, sound omission and other word linking techniques in faster speeches, automatic transcriptions might have the same problems as audios containing accents.

3. Proofreading. To err is human, but when it boils down to the matter of proofreading, it is very unlikely that a software will correct itself. With manual transcriptions, however, the errors either will be non-existent or minimal. Also, if one puts too much trust into automatic transcriptions, sometimes extra proofreading costs might be involved.

4. Price. Automatic transcriptions are most likely cheaper than the manual ones. This, of course, also very much depends on the verbatim type, audio length and turnaround time.

All in all, most find manual transcriptions more reliable and accurate, even though sometimes they are considered to be time consuming and more expensive. Looking on the bright side, however, automatic transcriptions could serve as frameworks for manual transcriptions.