Will speech recognition replace human in medical transcription?
SRT also known as voice recognition (VR), automated speech recognition (ASR), continuous speech recognition (CSR), etc., is basically a computer software that converts voice to text.
Different people from around the world speak English in different accents, with an accent similar to their mother tongue. Think of a speaker with a heavy guttural accent! So the voice recognition software needs to recognize these pronunciations. A lot of homework is needed to train the equipment before it can be used. Even a well-trained equipment needs human intervention like editing, proofreading and formatting. SRT cannot correct improper grammar, incorrect punctuation, incomplete dictation etc. Then there are many homonyms involved where human brain processes and uses the appropriate words, while voice recognition softwares need to prove their credibility in this regard. Background noise too reduces the recognition accuracy. SRT may alleviate the tedious process of typing but it cannot completely be on its own without human intervention. Currently it is used as a similar application like macros or word expanders.
Two types of SRTs are there, one is self editing (front end) and the other delayed (back end). With the first one, the speaker dictates into the PC and the voice is converted to text concurrently and the speaker himself corrects the errors made by the software, the doctors being reluctant to adapt to this system due to time constraints in their busy schedule. In delayed SRT, which is commonly in use now with large hospitals and clinics, the voice to text conversion is done at back office after the speaker has dictated rather than simultaneously. Then an editor must listen to the voice file, edit for errors and proofread the draft.
So in a nutshell, SRT can be a productivity enhancing tool to reduce turnaround times for the medical transcriptionists but not a creature eating an MT’s own bread at least for now.
