Speech Recognition Technology

Front-End vs Back-End

As speech recognition continues to be more accepted as part of the clinical documentation process, the pros and cons of front-end versus back-end speech recognition need to be weighed. The front-end process allows physicians or end users to have complete control over the documentation from dictation to final document. The front-end process occurs as the physician dictates, creating speech recognized text on-screen in real time. This also allows for quicker access to clinical documentation as it removes the delay involved in the traditional transcription process. However, the downside is that the end user must also take the time to train the front-end speech recognition engine and edit the document. Although there is a clear reduction in transcription cost, the time spent editing by the physician, end user, takes away from their main billable activities of patient care.

The back end process is transparent to the physician, end user, as it handles the speech training process on the server via the audio file. Through reading the audio file multiple times and receiving feedback from trained MT editors, the speech recognition engine improves in accuracy. This process also allows for a second pair of eyes to review the document for errors. The downside for the back-end technology is that the physician, end user, does not have any incentive to change their dictation habits. Since the physician is not involved in the training process, more time is then required to edit the speech recognized text by the MT editor and back-end speech recognition engine.