As we can apply more processor power, voice recognition has made many strides over the last few years. Now we expect to use voice to input data for call centers. We expect to use voice to navigate through menus. We expect to use voice input for commanding and controlling applications. We can pretty accurately use voice to input textual data as in correspondence for example.
However currently it is not an effective tool to enter transactional data. With traditional forms capture solutions, it is quicker at several thousand keystrokes/hour, to type information. However there are some transactional applications that are being successfully employed – the most visible of these is transcribing medical records, which Nuance has invested heavily in. The input is semi-structured and the application conceptually very similar to traditional paper based forms processing. The physician is trained to follow a specific structure in identifying and diagnosing the most likely problem and the proposed treatment. So rules can be built that can understand enough to substantially reduce transcription needs.
But other transcription services have been more challenging – legal transcripts remain un-automated. Bank traders’ voice recordings must be manually searched.
We believe that voice recognition is in a similar state to OCR when we started in 1989 encouraging the use of OCR pattern recognition to automate data capture from paper forms. This was supplemented by rules and key data entry in low cost areas such as India and has now expanded to encompass fax, emailed images, handwriting, barcodes, OMR and PDF’s.
Voice recognition is where OCR was 10 or so years ago. There are many opportunities for pattern based transactional capture from unstructured voice records. HSA is the leader in identifying and understanding these opportunities. For more information please email Harvey Spencer Associates.
Copyright © 2017 Harvey Spencer Associates