Applies to older version SpeechServers (Dragon v5/6 or IBM v8+) only. Older version included Standalone desktop client edition that is no longer available.
There are four basic tabs:
|
Contains a log of the server. | |
|
Settings of the current engine on local machine. | |
|
Settings of the current engine on all connected machines. | |
|
Default settings for new users. |
The information tab shows a historical log of what the server has been doing.
Use this tab to specify settings for the current engine on the local computer. Changes made on this tab make no changes to other engines or computers are not affected.
Enable Server -- This will enable the server. Unchecking this option will disable the server.
Check Every n Seconds -- This specifies how often the server will query Command! as to what jobs need to be processed. Set this to a reasonable value to insure work is checked in a timely manner. Pressing the Check Now button will force checking the queue immediately.
Disable between the hours of -- This specifies a time range to disable the server. You may want to run the server after hours or when resources are most available. When SpeechServers is enabled, it is recommend that other applications are not be running on that computer.
Log Events to File -- This will log all information in the log window to a file. The file is placed into SpeechServers program directory. Be careful to not leave this set for long periods. The file will continue to grow, filling up disk space.
Dragon User Storage Folder: -- A directory to store new users created by the SpeechServers must be set before any services can be turned on. Set a directory on one of your drives by clicking the ... button and browsing for a folder.
User Repository Cache Size -- This determines how many users are stored on the local computer from the repository at a time. Larger numbers consume more disk space, but can limit the time required to update users out of the repository onto the local computer.
Services Provided by this Engine -- Check the box of the services you want the engine to provide for this computer. Your options are:
|
TransWaveX |
transcription of .wav file |
|
SpeechTrainer |
automated corrective adaptation |
|
CompleteProfile |
profiles new users |
|
SaveSession |
provides extended session information for speech recognition |
Limiting what runs on a server can help to maximize resources. SpeechTrainer and CompleteProfile operation may take a large amount of time to complete. Having a computer(s) dedicated to these processes can help shorten the length of time for TransWaveX (transcription) services to complete.
This tab specifies setting for the current engine on all the computers connected to the current Command! database.
User Repository Storage -- Specify the location of the user repository for the current engine type (IBM or Dragon). The User Repository for each engine must be in different subdirectories.
Transcription Options -- Different options you can have performed when doing basic transcription services (TransWaveX).
Corrective Training Options -- Changing settings in this box will alter how the correction engine will perform general training. The SpeechServer will repeatedly train the file using the current Initial Bookends setting until the accuracy stops increasing. Once this occurs the Max Error Cycles will decrease by one. Every Training Cycle that does not improve the accuracy will decrease the Max Error Cycles by 1. If accuracy increases on a subsequent training cycle, the Max Error Cycles will be reset to the starting value. Once the Max Error Cycles reaches 0 the Initial Bookends is decreased by 1. This continues until either the Target Accuracy, or Max Training Cycles is reached. The settings are as follows:
'Record word errors after training' Checkbox -- This is used to generate a list of untrainable or untrained words from the SpeechServers.
Target Accuracy -- This specifies the target accuracy to attempt to train to. Once the accuracy has reached this level, training will stop.
Max Training Cycles -- The maximum number of training cycles that will be performed. Once this limit is reached, the current training job will end.
Max Error Cycles -- The maximum number of successive failures to improve the accuracy beyond the current maximum. This is used because some engines may improve over a succession of cycles.
Initial Bookends -- The number of matching words in a row to form a bookend. This can help to align text to audio by adding correct words to the beginning and ends of the correction text. Large values will cause the length of differences to increase dramatically and should be kept below two for optimal performance. The default value is one.
To explain how the Initial Bookends setting functions we will demonstrate using a fictitious transcription. If the sentence “The quick brown fox jumped over the lazy dog.” Was transcribed as “The brick brown sox slumped over the lazy hog.”(errors in red). The Speech Servers will make corrections differently based on what the Initial Bookends is currently set to. For example:
If Initial
Bookends is set to 2
The Speech Engine would make the following corrections.
1. “The brick
brown sox slumped over the” ->
“The quick brown fox jumped over the”
2. “the lazy hog”
-> “the lazy dog”
If Initial
Bookends is set to 1
The Speech Engine would make the following corrections.
1. “The brick
brown” -> “The quick brown”
2. “brown sox slumped over” -> “brown fox jumped over”
3. “lazy hog”
-> “lazy dog”
If Initial
Bookends is set to 0
The Speech Engine would make the following corrections.
1. “brick”
-> “quick”
2. “sox slumped”
-> “fox jumped"
3. “hog”
-> “dog”
Minimum Threshold -- This is the minimum accuracy that must be met before training can begin. This can help to avoid training a mismatched audio and verbatim text file.
The Speech Defaults tab is used to specify default settings that are used when the SpeechServers automatically creates new users. The selected vocabulary should be available to all speech engines used by all servers on the system.
