Once you have setup Express Scribe for Speech to Text, when you load a file it will process the file in the background and then include the full text in the Notes box. Even if your computer is fast, this can take some time (even longer than the recording itself) so this function works ideally when you have a queue of typing work. However, because the speech recognition runs in the background, you can start transcription immediately.
You should not be too excited by speech to text. It will be many years before machines will be able to transcribe accurately. At the moment, even a well trained engine will offer at best around 90% accuracy. It is our concept that you will use this feature as a tool to provide you with draft text which you can play through at a faster speed and correct or re-format.
Training is essential to get reasonable text to speech. You must "train" the engine to recognize each voice that you are going to use. The training function is managed by your speech recognition software, not Express Scribe. Please refer to your speech engine manual for more information.
Express Scribe automatically matches each dictation file with a trained speaker using the Express Dictate (or DialDictate) ID of the sender. This ID is obtained when they register their software (all unregistered users, WAV files and docked recordings are considered to be ID 0). After you train a Speaker to learn a User's voice, Express Scribe can set that speaker to run everytime that User sends you a file. You do this in the "User-specific profiles" section of the Options->Text to Speech tab.
Setup for Speech to Text
- Open the Options dialog and select the Speech to Text tab.
- Click "Enable speech recognition".
- Select the speech recognition engine from the Engine drop-down list. If you do not see any engines in the pull-down list, then you have not fully installed your speech recognition software or it is not SAPI compliant.
- Select a speech profile from the Default profile drop-down list. This is the profile that will be used for any dictations received from users who are not listed in "User-specific profiles".
- Get each user to train the engine so it can recognize them. They do that using your speech recognition software. Please refer to the manual provided with your speech recognition software. With most engines they can train the speech recognition engine using an other PC and then export the training file to your system.
- For each user who has completed the training, add their User ID and speaker Profile to the "User-specific profiles" list. If they are registered users of Express Dictate or DialDictate, use the ID that they obtained when they registered their software. User ID 0 for all unregistered users, WAV files and docked recordings.
Note: When opening the Speech to Text Options tab, or changing the selected Engine, the "Retrieving profiles" dialog may be displayed for some time.
When your next dictation is loaded, Express Scribe will run the speech engine in the background. When complete, it will "paste" the text into the Notes box.
Please be patient. The process can take some time (even longer than the recording itself). Thus the function is ideal when you have a queue of items. But remember that you can start transcribing immediately.
Please note that the Speech to Text process will only happen when a dictation is first loaded. If you run Express Scribe with an already loaded dictation the recognition process will not run. |