Front-end VS back-end: The battle of speech recognition and dictation
Whilst scrolling through the VoicePower twitter feed, a particular news article caught our attention. The article made reference to a recent study, in which the precision of front-end speech recognition was questioned. As a company which provides both front and back-end software, we thought we could contribute an alternative perspective to the discussion.
Front-end, also known as real-time speech recognition (SR), refers to software that works live, picking up your words and writing them on the screen as you speak them. Back-end dictation involves speaking into a handheld device, which records the dialog. This recording is then typed up and they are both sent to a secretary, where the accuracy and grammar of the text is checked and edited.
The study founded that front-end SR produced a 7.4 percentage of words that were incorrectly transcribed. Whereas, the back-end approach reduced these errors down to a significantly lower percentage of 0.4. Results such as these, will understandably leave readers wondering, why should we use front-end speech recognition?
Well, aside from the fact that the articles display image depicts a doctor using the software whilst wearing a surgical mask (duh!) and using a virtually pre-historic head-set, the answer is pretty straight forward… dedication!
Most SR users, medical professionals especially, are used to dictation (back-end) and forget that with front-end SR, it is the software doing the transcribing, not a human! As a result of this, the dictator’s techniques, such as the manner in which they dictate and their grammar, must be altered in order to achieve the full potential of the software. Thus, the dedication of the author produces the most effective results. The time spent during the initial training period, in which the author corrects any mistakes, practices their grammar and allows the software to learn their accent and the patterns of their voice, will contribute to an easier and more accurate result. The beauty of front-end SR is its ability to be made aware of, amend and remember errors for future reference. It would seem that physician resistance to allow time for this programming, is far greater than the actual inaccuracy of the software.
Something that we take pride in is the way in which our products and software contribute to a shorter turnaround period within the NHS – which would not be possible without front-end SR. The software allows doctors to dictate, proofread, sign and send information into EPRs (Electronic Patient Record systems), email and send clinical letters independently. This eliminates the processing time in which the secretary would proofread, edit and re-send the document back to the doctor in order to seek his approval and signature, before then sending out to the patient or required persons. This process can sometimes take up to four days, at which point the patient details may not be fresh in the doctor’s mind. Consequently, front-end SR saves time for more important tasks – like ensuring person-centered service.
Ultimately, both front and back-end dictation techniques are massively useful and pivotal in saving precious NHS and other professional’s time. Yet front-end SR particularly, is only as good as the dedication invested in it.
Here at VoicePower, we would love to know your thoughts… what do you think of front-end speech recognition?
Alternatively, if you would like to find out more, why not ask The Speech Recognition People? Contact us here https://voicepower.co.uk/contact-us/