Accuracy & Turnaround
How good and how quick?
Accuracy is a function of:
- Language and Voice Model (words used and how they ‘sound’)
- Acoustic Model (audio data)
- Lexicon and Dictionary of Common Phases and Word Combinations
- Pre-supplied transcriptions prior / post launch
The VST solution easily integrates into pre-existing Automated Speech Recognition (ASR) systems – which are plentiful and already operate in a wide variety of languages and capacities. A number of robust ASR’s systems are available with preloaded local accents and ‘slang’ dictionaries. Most ASR systems may be trained and improved upon by enabling a “correction feedback loop”; the net result is that your system will improve in accuracy as you continue to use it.
Accuracy can also be rapidly increased with samples of transcripts to perform word-count, language use and word combinations calculations. By scanning 1000s of pages of old interviews and extracting word use you can very easily and quickly pre-populate your ASR engine and increase its accuracy. Scanned documents can also help to create dictionaries and lexicons of phrases / acronyms / definitions used in your environment.
Our goal is to deliver 100% accurate transcripts and we will work with you to refine the system to meet your business requirements. Achieving your results is a combination of understanding your environment and employing the most appropriate combination of VST’s technology and capabilities.
Turnaround (Based on an average 30 minute interview)
Pre-processing of file
: 3 min
: 10-15 min
PDF Assembly and Archival
: 2 min
Seeing is Believing:
Please click on this link and download a direct example of our services. This is from a Police Interview. If you open the document in Adobe Acrobat, you will then be able to ‘hear’ the actual conversation by clicking on any of the text.