Learn something new every day
More Info... by email
Automatic transcription is the process of producing a written transcript of spoken or recorded speech through the use of computers and without direct human intervention. Accurate automatic transcription requires high-quality transcription software and a device that can accurately receive auditory input. In general, it is also necessary that the speech or recording to be transcribed is reasonably free of distortion and interference from background noise. Attempts have also been made to transcribe music with the use of computers running specialized transcription software. Music transcription involves writing out the notes of a given piece of music, particularly when there is no preexisting notation for that music, as is the case with improvised solos.
Good transcription software is necessary for successful automatic transcription. The software is responsible for processing the auditory input, separating a continuous language stream into separate words, recognizing those words, and correctly representing them in text. A failure at any stage of this process generally results in a transcript that differs somewhat from the source material. Good automatic transcription software should be able to recognize the subtle differences between similar words and compensate for various styles and speeds of speech. Difficult-to-understand accents tend to be problematic for even the best transcription software.
Some people use automatic transcription software because they prefer dictation to directly typing or writing their text. Some forms of automatic transcription software are particularly good for this purpose because they can "learn" the voices of the people whose words they are transcribing. In such cases, automatic transcription is not used to create transcripts of speech from a variety of different sources, so there is no need for the software to be open to a wide variety of speech patterns. Allowing this process of optimization, which can occur during extended use of the software by one person or during a preliminary calibration, can greatly increase the accuracy and potential speed of dictation.
Computers are, unfortunately, not as well suited to consistently and accurately recognizing human speech as humans are. They cannot, for instance, make use of contextual clues if they fail to understand a particular word. As such, it is often necessary for a human to proofread transcripts created through automatic transcription. Minor errors in formatting and various errors in transcription are, in many cases, common unless the transcribed speech is very clear. Still, using computerized transcription can quickly make a solid foundation for a transcript that requires only limited human intervention prior to submission or use.
One of our editors will review your suggestion and make changes if warranted. Note that depending on the number of suggestions we receive, this can take anywhere from a few hours to a few days. Thank you for helping to improve wiseGEEK!