Related Experiment Videos
A history-taking system that uses continuous speech recognition
K Johnson1, A Poon, S Shiffman
1Section on Medical Informatics, Stanford University School of Medicine, CA 94305-5479.
Proceedings. Symposium on Computer Applications in Medical Care
|January 1, 1992
Summary
Q-MED, an automated medical history system, uses speech recognition for patient symptom entry. Its natural language parser achieved 87% semantic accuracy, improving diagnostic efficiency.
Area of Science:
- Medical Informatics
- Artificial Intelligence in Healthcare
- Natural Language Processing
Background:
- Automated systems can streamline patient data collection.
- Speech recognition offers a natural interface for medical history taking.
- Challenges exist in accurately interpreting patient utterances in a dialog system.
Purpose of the Study:
- To evaluate the performance of the Q-MED automated history-taking system.
- To assess the semantic accuracy of the natural language parser within Q-MED.
- To determine the system's effectiveness in capturing patient-reported symptoms.
Main Methods:
- Q-MED utilizes speaker-independent continuous speech recognition for patient interaction.
- The system employs a dialog-based approach for symptom elicitation.
- Error-recovery mechanisms are integrated to handle misrecognitions and parsing errors.
Main Results:
- The natural language parser demonstrated an overall semantic accuracy of 87 percent.
- Q-MED effectively captures volunteered findings and clarifies unparsed information through targeted questions.
- The system facilitates automated patient history taking.
Conclusions:
- Q-MED's automated history-taking system shows significant potential in clinical settings.
- The high semantic accuracy of its natural language parser supports efficient symptom capture.
- Q-MED's dialog capabilities enhance the completeness of patient-reported data.