Thursday, November 5, 2009

Is Speech Recognition Ready for Prime Time - You Bet

In a posting on the American Medical News site titled: Is Speech Recognition Ready for Prime Time - You Bet Pamela Dolan refers to the history of speech recognition and how the technology was cited as one of the best things to hit healthcare - 10 years ago. In fact in 2005 I wrote an article for Health Management Technology Magazine (now available for purchase through Amazon): "Is Speech Recognition the Holy Grail":
Speech recognition technology has been lauded as the best thing to happen to healthcare technology since the advent of the computer, but is it really the Holy Grail? Speech recognition has the potential to overcome one of the most significant barriers to implementing a fully computerized medical record: direct capture of physician notes. Industry estimates from physicians and chief information officers at hospitals suggest that 50 percent of physicians will utilize speech recognition within five years. Coupled with this is the growing demand for medical transcriptionists, which is projected to grow faster than the average of all occupations through 2010
In pulling up the original article from my archive it made for interesting reading and while there were still problems with the technology in 2005 it had reached a tipping point and the summary at the end was pretty much on the money:
Speech recognition is good technology, but it is neither a panacea nor the Holy Grail. Speech recognition has been two years away for the last 10 years, but we may be approaching the Grail — finally.
Developments over the last several years have incrementally improved speech recognition systems to the point that some have intelligent speech interpretation—extracting the meaning, not just the literal translation of words—and producing high-quality documents with minimal human intervention. Further integration and embedding speech recognition with mainstream EMR solutions will allow for expedited capture of documentation as part of the clinical care process, offering clinicians a choice of methods to document creation. The last significant development in speech recognition technology was the recognition of continuous speech. The next big leap in this technology will be the merger of NLP and CSR to create natural language understanding. This development will take the technology to the next level and will offer a realistic opportunity to make speech recognition the de facto method of data capture for the medical community. The question is, When?
As the article from the American Medical News says:
"It (speech recognition) wasn't ready for prime time," Dr. Garber pointed out. "Now it is. No question"
But I disagree on the impediment to EMR usage that is linked ot the lack of discreet data. This is true with old style speech recognition - the process of converting the spoken word into text
The problem is when you talk into it, the data is not discrete ... it's still like a Word document
but not for speech understanding which is the the merger speech recognition and natural language understanding - available today. Already in use in many sites and delivering data in Healthstory CDA4CDT format.

So to answer the question - Is Speech Recognition Ready for Prime Time: You Bet!

So are you using it, what are your experiences or would you rather be entering data using forms and computer screens?


health care said...

Yet its not complete... let the imagination and efforts drive crazy to come out with the outlandish results for happening and to hit on the prime time. Definitely as this technique fulfills its demand, it will rocks and a great revolution in the health care industry will be done.

Richard said...

I believe what is misleading about speech recognition is the hardware requirements that are needed. I did server side speech recognition and this seems to be the most viable option for successful and near perfect text.

Nick van Terheyden, MD said...

The hardware requirements have been significant but hardware has improved dramatically and average configurations can be used. Server side versions do off load the heavy lifting to a data center and can improve the results both from a response as well as performance (due to the constant tuning and improvement carried out in the data center).
The technology is not perfect but it has reached a point that delivers results back that offer savings in time and resources

Skye said...

Speech rec has come far, but it will NEVER replace experienced personnel, such as transcriptionists. With the advent of speech rec, it was believed that it would replace coders as well. But that is slowly becoming evident that you cannot do without humans verifying the contents. Unfortunately, that finds our doctors sitting at monitors instead of having time to chat with their patients. The whole process still needs a lot of refining.