I bought the book with much anticipation since I am a regular reader of Mr. Geitgey’s posts. The book did not disappoint. I particularly enjoyed the introductory section on neural networks, specially, , the lucid description of forward and back propagation. I have read many references on the web and have taken the famous machine learning course by Andrew NG but none of those references explained how a neural network works as well as machine learning is fun did.
The code examples are easy-to-read and are well organized.
The VMware virtual machine is a nice touch.
I would have liked to see more discussion about adversarial neural networks and generative neural networks. In addition, more details of commonly used optimization algorithms such as gradient descent would have been welcome.
Finally, a section on how to install the several libraries mentioned would also be handy.
Speech Recognition
Five things to like about J-Say 13
The J-Say product has undergone a significant face lift. It is now owned by Hartgen Consultency, a company founded and headed by the developer Brian Hartgen. J-Say is middleware technology that interfaces Dragon NaturallySpeaking Professional with the Jaws for Windows screen reader. It allows users to control Jaws for Windows using their voice, and allows them to get full control of their computer. J-Say version 13 has just been released. There are five features I really like about this upgrade.
- The “forget it” command which deletes whatever was dictated in the last utterance and speaks what has been left. This allows me, as a user to retain the context of my dictation.
- The new correction feature using the “select” command. If the corrected phrase is in the choices, it improves the speed of editing significantly. No need to redictate or to open the spell box and re-enter the text.
- This is probably the fastest version of J-Say so far. I can feel this specially in Microsoft Outlook when I have to get through a large number of messages quickly. I prefer using the keyboard for this but now, speech recognition has become an acceptable option.
- J-Say tags allow the user, to mark files across folders and drives and then to delete, move or copy them. This beats holding down the shift key and moving around with arrow keys or, holding down the control key and marking files using the arrow keys.
- You can stay current with Jaws for Windows versions. J-Say no longer restricts you to a specific build of jaws.
Finally, for those of you contemplating buying this technology, the price has been reduced significantly. Please see the J-Say website for more information.
The role of bridging technology in speech-recognition
If you are a screen reader user and want to use speech-recognition, then you will be asked to use bridging technology. As of this writing, there are two options you have.
- J-Say wwith Dragon Naturally Speaking
- System Access with Windows speech-recobnition.
This post is not about which technology is better. One question that is asked often is can you use speech-recognition without bridging technology? The answer is yes you can but there are things that bridging technology helps you to do. You need to decide if these things matter to you. Most of my experience has been with J-Say so the below points relate to that technology.
- Reading the training text
- This is true of DNS version 8 and I suspect this has not changed for Dragon version 12.5. Training the text was very difficult. The difficulty was in knowing what text to speak next. I had to use the jaws color reporting feature accessed via jaws key + 5 to check the color of the text.
- Screen reader control
- Bridging technology gives you screen reader control so you can use your voice to do screen reading tasks. I have found that tasks such as commanding the computer to speak the contents of a dialog box, line or paragraph work nicely but scrolling and editing text is better done by the keyboard.
- Checking the status of the microphone
- When using speech-recognition, it is crucial to know if the microphone is off or on. Bridging technology like J-Say may give you an easier way of determining this.
There are advantages to not using any bridging technology. The main advantage is that you are not tied to any screen reader or other product so can update as soon as updates are released. You do get third party programs such as knowbrainer which give you commands that make formatting easier but these products are not particularly screen reader compatible. For example, knowbrainer works by pressing keys to execute various commands. That gets noisy quickly when using a screen reader. The other advantage is cost. It is significantly cheaper to just buy Dragon Naturally Speaking and Windows speech recognition is built into windows.
controlling jaws for Windows through Dragon NaturallySpeaking using J say
this post is going to walk you through the steps of controlling jaws for Windows from Dragon naturally speaking. You cannot speak the keystrokes to control Jaws for Windows directly since the Jaws keyboard driver runs at a lower level than Dragon’s keyboard driver. There are two ways you can control jaws from Dragon.. One way is to use the com interface that Jaws provides. The second way is to call Jaws scripts. I am going to be showing the second way since I’m assuming that you have J-Say technology installed.
The first thing you need to do is to determine the name of the script you want to call. You can see the name of the script from the Jaws scripting manager or by looking at the Jaws keyboard manager. Once you have that information, you can begin writing your command. We will use the example of initiating a jaws tandem session. This is a globally defined jaws script and is not in the set of current J-Say commands. A jaws tandem session is initiated by calling the StartOrEndTandemSession script.
I am going to assume that the StartOrEndTandemSession script is installed and working.
1. Go to the desktop.
2. From the tools menu of dragon, invoke the “add new command” dialog. If you have dragon running, say “add command”.
3. Dictate or type a name and a description.
4. Make the command a global command as well as an advanced scripting command.
Note:
On my computer, the combo box to select the type of the command is not spoken automatically. You may have to use the Jaws read current line command to read the value that has been selected in the combo box.
5. Keep tabbing until you reach an edit area where you can type your script. 6. A set of begin and end statements will already be inserted for you. 7. Between them, enter the following line.
DllCall “MSGW1004″,”JFWRunScript”,”StartOrEndTandemSession”
Note:
I do not know what MSGW1004 stands for but I suspect, after reading the help it is a dll name.
8. Now, activate the jaws cursor and click the save button. This is found at the bottom of the scripting edit area.
9. Assuming there are no errors, you will be back at the desktop or perhaps in the command browser.
10. In any case, your command is ready now so all you now need do is to speak the name of the command and it should work.
You can have as many dllCall statements as you want in a script. The key thing to remember is that the Jaws and Dragon scripts should be synchronized that is they should be loaded at the same time otherwise, if the Dragon commands are loaded and the Jaws script files are absent, you will get an “unknown script call” to the script if you speak the relevant Dragon command.
Let us look at another example. I have created a macro in Outlook that deletes messages that have the same thread. I need to invoke this macro when I say “delete thread”. The Dragon command is below. I have added comments to each statement.
Sub Main
DllCall “MSGW1004″,”JFWRunScript”,” SpeechOff” ‘turn jaws speech off SendDragonKeys “{alt+f8}” ‘invoke the macro ecxecution dialog SendDragonKeys “{alt+r}” ‘only 1 macro in outlook so run it
DllCall “MSGW1004″,”JFWRunScript”,” SpeechOn” ‘enable jaws speech End Sub
Masking user interface complexity using speech-recognition
Shrinking budgets are nothing new in IT departments. Unfortunately, training budgets are one of the first to go. This leads to a situation where new software is deployed but does not yield the expected return since users are unable to utilize it fully. However, there is one way around the problem. Use speech-recognition. It is a more natural way to compute and can help automate processes by the use of macros. Moreover, it has a uniform command set that does not change with changes to the software it is driving. For instance, to copy text to the clipboard, you will say "copy that" irrespective of whether you are in notepad, Microsoft Word 2003 or Excel 2010. This has huge benefits since users can continue to use familiar commands in new user interfaces so the "how do I do this" type of queries will be significantly reduced.
Another aide in masking user interface complexity with speech-recognition is third party software in the form of command add-ons. For example, Dragon Naturally Speaking has several add-ons like Knowbrainer http://www.knowbrainer.com, VoicePower <http://www.voicepower.co.uk/ and J-Say http://www.tandt-consultancy.com/products/ss00014_j-say_professional.php. These add-ons compliment Dragon’s functionality and their developers update them as their supported software changes. Some of these add-ons like knowbrainer, even have their own scripting languages which simplify custom command creation.
Table describing the above add-ons
Add-on | What it does | Distinguishing features |
Knowbrainer | Contains over 10000 commands to enhance the use of Dragon Naturally Speaking across various applications. | Verbal Basic- a custom command creation language that is speech driven and a popular set of speech-recognition forums on http://www.knowbrainer.com. |
VoicePower | Another command add-on that provides a plethora of commands for computer control. | Provides context sensitive training to users as they work. |
J-Say | Connects a screen reader Jaws for Windows with Dragon Naturally Speaking | Several productivity commands and is the only add-on in its class. |
From the Keyboard to the Microphone: A More Natural Way of Computing
This is my article in <a href="http://www.blindskills.com/dialogue.html"Dialogue magazine It deals with migrating from using the keyboard to using speech-recognition.
The article also makes a special reference to screen readers and other bridging technology like J-Say
The article has been published in the November December Issue.
Unable to download Dragon Naturallyspeaking 10.1
Just a quick note to readers that it does not seem to be possible to download Dragon Naturally Speaking 10.1. When you navigate to
this link
you are asked for your contact information and dragon serial number.
Once you have entered these items, you are taken to another page where you need to wait some more. You will be told that Nuance is sending an e-mail with download instructions. The e-mail comes but the download link is not active.