Automatic Transcription: How Often Does it Fail and How Much Will It Cost You?

oopsAutomated speech recognition software is a relatively new development that can produce transcripts of audio files for use in a variety of ways. The frequent errors associated with automatic transcription, however, dramatically limit its usefulness, especially in cases where accuracy and precision are of the utmost important (think medical transcription). Automated transcribing software cannot compete with professional transcription services in terms of accuracy or polished presentation; in fact, most automatic transcription programs require a great deal of costly proofreading in order to be made presentable for public use.

Software has no common sense.

While increasingly complex algorithms are used to generate the transcripts produced by transcription programs, these programs are not generally equipped to handle exceptions to their fixed rules or to learn new words or new linguistics rules. Awkward, stilted or simply incorrect transcripts are extremely common when using these speech recognition programs. A study published in the 2010 proceedings of the 28th International Conference on Human Factors in Computing showed that error rates of 20% or less were required in order for students to find automated transcripts of class discussions even somewhat helpful. By way of comparison, many modern automatic transcription programs still have error rates in excess of 40%, making them nearly useless for practical purposes.

Software lacks an interpretive human element.

No set of rules can encompass the scope of an entire language. Additionally, automatic transcription software is notoriously tone-deaf; it cannot pick up nuances or flow. Punctuation errors are extremely common with speech transcription software; the differences between a comma, semicolon and period typically are not recognized and must be corrected manually. While in some cases these errors are merely annoying, they can be costly and damaging to companies who must depend on these communications to stay in touch with clients and prospective customers.

The cost of failure is very high.

In most cases, the results produced by automated transcribing software require significant additional proofreading in order to render a usable end product. When combined with the cost of the automatic transcription programs, the resulting time and money expended normally exceed the cost of hiring a professional transcription service. Worse yet, the errors produced by these programs can create an unfavorable and unprofessional impression and negatively impact the reputations of companies and individuals.

Professional transcription services provide a more accurate and reliable way to create transcripts of audio files. Rather than the hit-or-miss results produced by automated systems, professional transcriptionists can pick up on the nuances and the flow of the audio and produce extremely precise transcripts that require no additional proofreading. This can save time and money and enhance a company’s or individual’s reputation.

Leave a Reply

You can use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>