MT Speech Recognition & Editing 2 — Questions and Answers
Question 1: When an SR system transcribes 'ileum' as 'ilium,' what type of error has occurred?
- Homophone confusion (Correct answer)
- Phoneme substitution
- Language model failure
- Acoustic model dropout
Correct answer: Homophone confusion
Ileum (small intestine) and ilium (pelvic bone) are homophones, making this a classic homophone confusion error common in medical SR.
Question 2: Which editing technique is recommended when SR misrecognizes a dictated medication dosage?
- Delete the entire sentence and retype
- Correct only the erroneous token and verify context (Correct answer)
- Accept the error and add a flag comment
- Skip the correction and forward to QA
Correct answer: Correct only the erroneous token and verify context
Best practice is to correct only the specific error and verify surrounding context to ensure dosage, route, and frequency remain consistent.
Question 3: What does 'back-end SR' mean in a medical transcription workflow?
- SR that processes audio after the dictation session ends (Correct answer)
- SR integrated directly into the dictation handset
- A manual transcription fallback system
- SR used exclusively for radiology reports
Correct answer: SR that processes audio after the dictation session ends
Back-end SR processes recorded audio files after dictation is complete, as opposed to front-end SR which operates in real time during dictation.
Question 4: A physician's dictation includes 'the patient denies any fevers, chills, or night sweats' but SR outputs 'the patient denies any fevers, chills, or night seats.' The MT should:
- Change 'seats' to 'sweats' and continue (Correct answer)
- Query the physician about intended meaning
- Leave the text as dictated to preserve accuracy
- Delete the phrase entirely
Correct answer: Change 'seats' to 'sweats' and continue
Night sweats is a recognized clinical symptom; the SR misrecognized the final word and the MT should correct it to the clinically accurate term.
Question 5: Which factor most significantly degrades SR accuracy in a medical setting?
- High dictation volume
- Strong regional accent combined with background noise (Correct answer)
- Long report length
- Use of abbreviations
Correct answer: Strong regional accent combined with background noise
The combination of a non-standard acoustic profile (accent) with background noise creates compounding errors that lower SR confidence scores significantly.
Question 6: When editing SR output, an MT notices the system consistently misrecognizes a specific physician's pronunciation of 'subcutaneous.' The best long-term fix is:
- Flag every instance manually
- Retrain or update the physician's SR voice profile (Correct answer)
- Replace the word with its abbreviation 'SQ' throughout
- Reassign the physician's dictations to manual transcription
Correct answer: Retrain or update the physician's SR voice profile
Retraining the physician's individual SR voice profile corrects systematic recognition errors caused by unique speech patterns.
Question 7: In SR-assisted transcription, what is the purpose of a 'confidence score threshold'?
- To rank report priority in the work queue
- To flag low-confidence words for MT review (Correct answer)
- To calculate the physician's dictation speed
- To determine billing codes automatically
Correct answer: To flag low-confidence words for MT review
Words or phrases below the confidence threshold are highlighted or flagged so the MT knows to pay closer attention and verify those segments.
MT Speech Recognition & Editing 2
When an SR system transcribes 'ileum' as 'ilium,' what type of error has occurred?