Chapter 11: Speech Recognition

Quad Grams

(system looks at 4 words prior to being identified) When the speech recognition system tried to determine the word, the system looks at the system looks at four words in the context to help determine the correct word or phrase.

Hidden Markov Model (HMM)

A statistical model that output a sequence of symbols or quantities. Another technology use.

The advantage of front-end speech recognition is

The turnaround time is shortened

Speaker Dependent

means that it has been trained to a particular user and already has an established vocabulary for that user

Continuous Speech Recognition

means that the user can use the system naturally and does not have to speak as slowly and distinctly as previous systems

Keyword spotting

A command would be triggered when a user would say a simple phrase or keyword subfield of speech recognition that identifies keywords in spoken language. (a subfield of speech recognition that identifies keywords in spoken language and prior to the 1980s, speech recognition was based on which a command would be triggered when a user would say a simple phrase or keyword. It could expand as the capabilities of technology were rapidly advancing. Still used in call routing systems by a client who when prompted by the system will speak a name or key word and the system will route the caller to the appropriate area or department for phone assistance.)

Back End Speech Recognition (BESR)

A medical specialist editor reviews the document for accuracy which makes the physician even more efficient. (the specific use of speech recognition technology in an environment where the recognition process occurs after the completion of dictation by sending voice files through a serve. The physician dictates in the usual manner, the audio is sent as a draft text along with the speech file to the medical language specialist who serves as an editor. Can be used to offer clinicians a way to immediately dictate, sign, and produce a report independently.)

Gesture Recognition Technology

A method of encoding handwritten, print, or cursive characters and interpreting the characters as words or the intent of the writer. These networks are part of a technology known.

Speech recognition technology (SRT)

Also known as voice recognition, allows the replacement of written or data entry text with the human voice entry. Uses the human voice as the medium for input. First introduced back in late 1980s.

ADHI

Association for Healthcare Documentation Integrity (formerly American Association for Medical Transcription)

Front End Speech Recognition (FESR)

Dictation with its once-and-done approach allows the provider to create, edit and sign the report, and to dictate directly into EHR fields or select from menu options using voice commands. The physician or the dictator is the editor of the document that is dictated. A medical specialist editor reviews the document for accuracy, which makes the physician even more efficient. It also can be used to offer clinicians a way to immediately dictate, sign, and produce a report independently.

Explain the process of dictation by the dictator in front-end speech recognition as well as the back-end process of editing by the transcriptionist.

Front -end speech recognition (FESR) can be used to offer clinicians a way to immediately dictate, sign and produce a report independently. Back-end speech recognition (BESR), a medical specialist editor reviews the document for accuracy, which makes the physician even more efficient".

What is expected to happen with the role of the medical language specialists as more speech recognition technology is used in the future?

Increase in productivity and decrease in staffing

Interactive Voice Response (IVR) system

One of the earliest speech recognition systems, taking its root from the keyword spotting systems from the early 20th century.This was the era where voice commands were popular by telephone such as when phoning the doctor's office, for example, one would say "appointment" if they wished to make an appointment.

The physician wants to dictate a question. What lets the SRT know that it is a question?

Physician changes his inflection

Issues with SRT

Quality of product training with the software editing of dictated reports

Identify the history of speech recognition technology from its earliest beginnings until now and how this is integrated into the computerized health record.

Since it was first introduced in the late 1980s, this technology has improved. Early on based on keyword spotting Also used hidden Markov model -Was slow and based on enunciation -ANNs and gesture recognition technology -natural language processing / speech recognition / AI

Natural Language Processing (NLP) Technology

Speech recognition but converts human language (structured or unstructured) into data that can be translated then manipulated by computer systems. This technology basically teaches the meaning of a word or phrase to a software application after several uses. Don by a statistical algorithm, as the applications can then compare and code these similar expressions accurately and quickly.

Turn Around Time (TAT)

The elapsed time from completion of dictation to the delivery of the transcribed document either in printed medium or electronically to a repository. It also used as a measure of quality.

As seen with dictation-transcription process, turnaround time is a measure of what?

The time frame of completion of the dictated reports

trigrams

When the speech recognition system tried to determine the word, the system looks at the system looks at three words in the context to help determine the correct word or phrase.

bigrams

When the speech recognition system tried to determine the word, the system looks at the system looks at two words in the context to help determine the correct word or phrase. (system looks at 2 words prior to being identified)

Macro

a pre-approved document outline designed so that the practitioner only dictates to fill in the blanks in the facility may chose to use.

Neural networks

are designed to work similarly to the brain's neural networ. It recognize the speaker's voice and remember the way the user says each word, regardless of varying accents and inflection.

Artificial Neural Networks (ANNs)

computational techniques based on artificial intelligence and machine learning in which the structure and operation are inspired by the properties and operation of the human brain. another mathematical model that was introduced in the late 1980s. It continue to learn as they gather information.

Discreet speech

in older voice recognition applications, each word is required to be dictated slowly and distinctly by the user to allow the software to determine where one word begins and the next stops.

The term used to describe computers working similarly to a person's brain is

neural network

The HIM professionals must review clinical documentation not only for quality and accuracy but also for

readability errors from copy and paste overdocumentation Granularity of ICD-10 coding

The SRT has been trained to use a specific physician's voice. This system is:

speaker dependent

Template

the format used to control documentation to be included in a progress note, history and physical, or other document. It can be use in both the paper and electronic environment.

If a provider wants to ensure the correct spelling of a word while dictating using SRT, he or she should

use the military alphabet and spell the word out