ASR (Automatic speech recognition) has evolved and hence is extensively used for commercial purposes. Consequently, the chances of error rate are higher in most speech recognition areas. It is one of the main obstructions when it comes to the implementation of speech technology. This is especially in the case of applications with a larger volume of vocabulary speech recognition. The continuous incidence of errors has increased the need to look for some alternative solutions to identify and correct such errors automatically. It is very crucial to correct such transcription errors to enhance the accuracy of speech recognition and also to avoid the broadcast of the errors to the language modules like machine translation.
Common transcription errors and corrective measures
Missing or misspelt words, inappropriate usage of medical language, distinct language patterns and user accents often lead to errors in medical reports or documents when generation through SR technology. Here are some steps to prevent such errors.
- Add dictionaries of clinical specializations.
- Use reminders for clinicians to speak clearly and slowly when an SR technology identifies discrepancies during a speech to reduce the possibility of misapprehension.
- Use semi-automatic correction that helps users to check errors related to spellings quickly.