TechWeb

Talk About Speech Recognition!

Nov 30, 2004 (09:11 AM EST)

Read the Original Article at http://www.informationweek.com/news/showArticle.jhtml?articleID=54800701


To date, most PC-based speech-to-text dictation solutions have been too unreliable for the typical user to embrace. ScanSoft's new Dragon NaturallySpeaking 8 should change those perceptions.

The latest version of Dragon NaturallySpeaking offers improvements in ease of use and speech-to-text accuracy. What's more, the product supports a wide variety of digital voice recorders, allowing a user to dictate on the run and transcribe to text at his or her convenience. The product also supports Pocket PC devices.

Setup is straightforward and CRN Test Center engineers were impressed with the speed at which the product learned a new user's voice. After only a few minutes of "training," the product worked with incredible accuracy on a PC and improved even further with additional use.

Dragon NaturallySpeaking was also tested with an Olympus DM10 digital voice recorder, and the final results were acceptable, though not stellar.

The product's limited audio file format support is its only downfall; currently, users can only import 16-bit Mono WAV files. This means that files must be converted from a recording device's native format to a WAV file format. Though not difficult, the process is slightly time-consuming and requires the addition of conversion software. Hopefully, future versions of Dragon NaturallySpeaking will support MP3 and WMA formats, which would greatly simplify the use of third-party recorders. Roving profiles and multiple PC support allow a user's voice files to transfer to any PC used.

ScanSoft offers specialized versions of Dragon NaturallySpeaking for solution providers selling into vertical markets such as the medical and legal fields. These are areas that could greatly benefit from accurate speech-to-text conversions.

Dragon NaturallySpeaking 8 comes close to approaching an ideal speech-to-text solution with its ease of use and overall accuracy and should have solution providers exclaiming, "Finally--speech recognition software for the masses."