Category Archives: Healthcare Technology

Speech Recognition: Can it walk the talk?

Speech recognition technology has been rapidly developing over the years. Companies like Dragon Naturally Speaking and Via voice have a significant market share.

There is a lot of concern within the Medical Transcription Community about the rapid adoption of Voice recognition technologies. Medical transcriptionists believe that Speech Recognition (SR) software will take over the industry and leave them out in the cold.

Well there is a lot of debate on all sides about this issue but the salient points stand as follows:

Yes, SR is rapidly developing in terms of accuracy and speed. However, the rates of accuracy can be 90% at best, in the case of the health-care industry. Health-care has a unique vocabulary that includes words that are not used in everyday life. Words that are complex to pronounce and sound very much like other words. For instance, how does the software figure out if you meant “French eyes” or “franchise; “urine” or “you’re in”; “dilate” or “die late”; “cauterize” or “caught her eyes”; “nitrate” or “night rate”, there are innumerable examples.

Thus speech recognition has some problems to surmount, problems that won’t exist for a human transcriptionist. There are some who say that speech recognition has reached a point of very high accuracy, only a few percentage points below human transcriptionists. But, is it worth sacrificing a couple of percentage points of accuracy for saving some money? After all, we are talking about human lives here.  Cost should not be a primary consideration, and we should aim at very highest levels of accuracy.

There is a school of thought that says that a doctor can use speech recognition to render documents, and then can read and check the document for errors later on. So the question that now rises is: is the doctor supposed to do this sort of work? The doctor’s primary purpose is to treat patients. A doctor would want to spend as much time with the patients as possible, and does not have the time for data entry and document checking.

The extension of this argument is: if the doctor does not want to proofread the documents generated by the SR software, then he can hire Medical Transcriptionists to proofread and check the documents rendered by the SR software. But this defeats the purpose of having SR software. If you are hiring MTs to proofread the software rendered documents, then you might as well have MTs actually doing the transcription work for you. Why pay for the software and then hire MTs to proofread it for you? It doesn’t make sense. Even from the point of view of the MT, the transcriptionist would much rather prefer to listen to a dictation from an audio file and then type the words at a fast rate (most MTs can type at 50-plus WPM), rather than look at a flawed document and try to spot errors.

Thus, it is undeniable that SR software is a marvel of modern technology, but in the health-care environment, the marvel falls far short of what an actual human can do. It makes sense for health-care providers to have human MTs and be secure in the knowledge that they are being assured the highest levels of accuracy.