W h i t e Pa p e r
Guidelines for Developing Windows Applications Compatible with Dragon NaturallySpeaking
W h i t e Pa p e r | Guidelines for Developing Windows Applications Compatible with Dragon NaturallySpeaking
INTRODUCTION These guidelines explain how to make a Windows application compatible with Dragon NaturallySpeaking. The guidelines are intended for all Windows application developers. The primary goal of these guidelines is to allow applications to work with Dragon NaturallySpeaking (DNS) out of the box, that is, without the necessity of using the Dragon NaturallySpeaking SDK to integrate speech into the application. However, following them will also make applications more compatible with other speech applications and accessibility tools. You may be reading these guidelines because your customers or end users have asked you to make it possible to use your application with Dragon NaturallySpeaking. They may not have provided you with specifics, so this paper explains the main reasons that users adopt DNS and what features are important when they use DNS with an application. You may need to ask your customers why they want to use DNS and which of its features are most important to them. The uses of DNS fall into two major areas: dictation (also known as speech-to-text) and navigation (also known as command-and-control). This paper addresses both of them. Note: if you are developing a web-based application, see the white paper Guidelines For Speech-Accessible HTML For Dragon NaturallySpeaking, available in tech note 3067 in the ScanSoft knowledge base at http://knowledgebase.scansoft.com. That white paper is more relevant than this one. Why Do Your Users Need Speech Recognition (Productivity or Accessibility)? Speech-recognition users generally adopt the technology for one of two basic reasons: Productivity: users want to accomplish tasks more quickly than they could by hand. Usually they are interested in the dictation capability of Dragon NaturallySpeaking, which is much faster than typing. Dictation increases task speed in textintensive practices such as creating physician notes and legal documents. (Another way to look at
productivity is in terms of transcription cost; DNS saves costs by allowing professionals to generate their own text when they would otherwise have a transcriptionist generate it.) Sometimes DNS can be used to improve productivity through its custom command (macro) capability to automate sequences of actions in the application. This capability is analogous to the macro capability in applications such as Microsoft Office or macro addon software such as Macro Express, with the added feature that macros can be triggered by voice. Accessibility: they need to use computer applications with limited or no use of their hands, so both dictation and navigation are important. Often they are more concerned with reducing mouse usage than keyboard usage because the mouse is likelier to cause repetitive stress injuries. In other cases, they may be using their hands for tasks such as handling papers or medical specimens. Ideally the hands-free user should be as productive as a fully abled user, but this may not always be possible, especially if the application involves more navigation than text entry. DNS features that address navigation include: o automatic generation of voice commands for standard Windows controls (menus, buttons, etc.) o automatic generation of voice commands for links and other web-page features in Internet Explorer o the ability to generate any keystroke by voice (useful in any application that can be navigated by keyboard) o the ability to move the mouse pointer by voice (a cumbersome last resort in an application that requires the mouse for navigation). DICTATION AND CORRECTION Full Dictation Support (Select-and-Say) Dragon NaturallySpeaking allows the user to dictate into any application. It also allows the user to correct and edit text by voice, a capability known to end users as
Productivity Applications Division
Select-and-Say. A key command is the select <text> command, in which <text> can be any word or group of words in the control that currently has the input focus. A related capability is the ability to play back dictated text to assist in proofreading. For more information on the voice commands that users expect to be able to use during correction, see the Correcting Recognition Errors section of the Dragon NaturallySpeaking Help. For the Select-and-Say and playback capabilities to work properly in your application, Dragon NaturallySpeaking must be able to set and get the text and the selection in any control in order to synchronize them with its internal dictation buffer. It can do this automatically if you follow these guidelines: Use standard window classes for edit controls, including Edit, RichEdit, RichEdit20A, RichEdit20W, RichEdit50W. If you need to subclass one of these controls, include the string "edit" or text in the name of the new class, because Dragon NaturallySpeaking treats controls with these strings in their names as standard Edit controls. It sends standard Windows messages to your control, such as EN_SELCHANGE, EN_REPLACESEL, EN_SETSEL, EN_GETSEL, WM_GETTEXT, WM_SETTEXT, so your control must support these messages. If you subclass one of the rich-edit controls,follow the instructions in tech note 4247 at http://knowledgebase.scansoft.com to enable Select-and-Say in this control because by default Dragon NaturallySpeaking treats them as Edit controls, not rich-edit controls. If your text entry areas are not standard Windows edit controls or subclassed from such controls, or if you cannot make it respond to the standard Windows messages such as the ones listed above, use the custom-dictation control of the Dragon NaturallySpeaking SDK to add dictation support to your application.

Another way to take advantage of Select-and-Say and playback capabilities is for your application to use Microsoft Word as its text editor. Dragon NaturallySpeaking 8 enables dictation support in Microsoft Word even if Word is embedded in or launched from another application. In versions of DNS previous to version 8, follow the directions in tech note 3402 at http://knowledgebase.scansoft.com to enable support for Microsoft Word in your application. NaturalText (Global Dictation) Dragon NaturallySpeaking can still be used even when dictation support is not enabled, which includes these cases: The application uses nonstandard edit controls, and dictation support has not been enabled for those controls using the Dragon NaturallySpeaking SDK. The application is running on a server through a client/server system such as Citrix or Microsoft Windows Terminal Server, and the user is running Dragon NaturallySpeaking on the client system. This capability is called NaturalText or global dictation. NaturalText sends keystrokes to the application in order to enter text and change the selection, and has no way to detect what text is present in the edit control (unlike dictation support, which does all of those things programmatically). The end user sees a message in the DragonBar that says dictating into a nonstandard window." The Select-and-Say and playback capabilities function in a limited fashion. For details of the limitations of nonstandard windows, see tech note 4247 or the topic "Dictating to nonstandard windows in the Dragon NaturallySpeaking help. Additional Guidelines for Dictation The following guidelines are equally applicable whether your application uses dictation support or NaturalText. Place the input focus in the appropriate text window, with an insertion point that makes it obvious to the user where typed or dictated text will appear. Make sure the visible focus is in the same control that would be returned by a call to a
Windows GetFocus() call, because the latter is where DNS places dictated text. This is because users can become very frustrated when their dictated does not appear where they expect it to, or does not appear at all. Return the focus to that window, at the same insertion point, if the application is deactivated and then activated again. The reason for this requirement is that the user needs to use commands that display the DNS Spell dialog, such as Spell That, to correct recognition errors. Your application window becomes deactivated when Dragon NaturallySpeaking displays the Spell dialog and reactivated when the user closes the Spell dialog. If the focus has changed, DNS cannot place the corrected word in the text and the user cannot continue dictating without first clicking into the text window to restore the focus; this situation causes frustration. If your application needs to take action as a result of changes to the text, do not assume that changes can come only from the keyboard or by pasting from the clipboard. Dragon NaturallySpeaking dictation support changes the text by sending messages such as WM_SETTEXT and EM_REPLACESEL. Try not to enter a state where the text cannot be changed for an extended period of time. For example, do not change a read-write window into a read-only window. This is because the user will try to dictate into the window and become frustrated. NAVIGATION BY VOICE Say What You See (Tracking) Background: Dragon NaturallySpeaking implements a feature called Say What You See. The ScanSoft engineering term for this capability is tracking. Tracking allows the user to say the name of any menu (such as File), menu command (such as Open), button, or static control that is used to set the focus to an edit control. The user can precede the command with the word click if necessary to distinguish it from dictation. (A user-settable option requires the user to say click before every such command.) For more information on what users expect to be able to do with

tracking, see the topic Controlling programs by voice quick reference in the Dragon NaturallySpeaking Help. Tracking commands are built at the start of every utterance (that is, when the user starts speaking after a pause), at which point Dragon NaturallySpeaking enumerates the child windows of the active window. For each one that is a menu, button, or text label with an accelerator key, it builds a voice command for that control. The voice command for a menu or button generates a mouse click on that control. The voice command for a static text label generates the corresponding accelerator key. Dragon NaturallySpeaking does not attempt to build commands for individual items in a list box or combo box (except in Internet Explorer; see the HTML white paper referenced above). Users can select items by saying commands that generate keystrokes, for example move down 2 or press B. You can also create commands for these items using the voice-command control of the Dragon NaturallySpeaking SDK. Therefore, in order for tracking to work properly in your application, you must do the following: Give each control in the application a label that is descriptive, unique, and short enough to be used as a voice command. If the label must be long, shorten the corresponding voice command by enclosing part of the text in parentheses; the user does not have to say the part of the label that is between the parentheses. Do not label controls with images rather than text. Make it visually clear which label corresponds to each control. Use standard window classes for controls, such as "Button." If you create custom controls, make sure they have the appropriate class and also respond appropriately to the WM_GETDLGCODE and WM_GETTEXT messages. Always use accelerator keys in text control labels that are associated with (that is, used for navigation to) captionless controls such as edit controls and list boxes.
In addition to (or instead of) following the above requirements, you can use Microsoft Active Accessibility (MSAA). MSAA makes information about your applications UI elements available to Dragon NaturallySpeaking as well as to other client programs that use MSAA, which include accessibility aids and automated testing tools. MSAA is a component of the Windows XP and Windows Server 2003 operating systems, and can be installed on Windows 2000, Windows 98, Windows ME, Windows NT. For more information, search for the Active Accessibility start page at www.microsoft.com. Keyboard Accessibility In general, applications that will be used by voice should not require the use of the mouse. There are three reasons: o Generating mouse actions by voice is slower than generating keystrokes by voice. o When the user wants to build a voice macro to automate a sequence of actions, programming keystrokes into the macro is far more reliable than capturing mouse movements and also results in quicker macro execution. o If the application is running on a server through a client/server system such as Citrix or Microsoft Windows Terminal Server, and the user is running Dragon NaturallySpeaking on the client system, tracking does not work and the user must rely on voice commands that generate keystrokes for navigation. Make sure that all controls have keyboard shortcuts or can be reached with the Tab key. If the Tab key is used for navigation, make sure there is at least one control in a predictable location (such as the first control on the screen) that can be reached by a keyboard shortcut. Also make sure the tab order corresponds to the visual layout of the screen. If your application uses toolbars (which Dragon NaturallySpeaking does not track), make sure each toolbar function can also be accessed through the applications menus.

Support the standard Windows keyboard shortcuts that are meaningful in your application. The Microsoft web site has several documents listing Windows keyboard shortcuts, such as knowledge base article 301583 (List of the keyboard shortcuts that are available in Windows XP.) Make sure all the keyboard shortcuts in your application are documented or that you provide visual cues to tell the user what they are. If your application places a shortcut on the Windows Start menu and/or desktop, make the name of the shortcut pronounceable and short enough to use in a voice command. This is because DNS uses MSAA to generate a Start <application> command for every shortcut on the Start menu and desktop. SYSTEM INTERFACE Here are some additional, general requirements. Include a VERSIONINFO block in your executable files, making it possible for other programs to query the version of your application. This allows Dragon NaturallySpeaking to behave differently in different versions of your application, should the need arise. Do not allow your application to go into a spin loop that consumes 100% of the CPU, even though it may be invisible to your applications performance except when using instrumentation. The reason for this restriction is that DNS needs to use 100% of the CPU while it is processing speech; if your application consumes the CPU then DNS becomes unusable. The task manager tray icon is useful for detecting this problem. Pop-up menus and mousecapture modes are the most common offenders. For keystroke input, if your application needs to check the current state of a key (such as the CTRL key VK_CONTROL), use GetKeyState and avoid GetASyncKeyState. This is because GetAsyncKeyState reads the current state of the physical keyboard, which is not meaningful if DNS is in the process of generating a keystroke sequence. GetKeyState reads the keyboard state after it has been manipulated by DNS.
CONCLUSION Dragon NaturallySpeaking attempts to adapt itself to your application, and you can help it by following the guidelines above. Still, it is best to perform realistic test scenarios with your application, using voice only, without touching the mouse or the keyboard. It is a good idea to follow general guidelines for accessibility (and also Section 508 compliance in the United States). For reasons listed earlier, even fully abled users can benefit when an application is made accessible. Microsoft provides accessibility guidelines for application developers on their web site. For assistance in developing your application, you can sign up for ScanSoft developer support. Developer support is not only for developers using the Dragon NaturallySpeaking SDK; it is for any developer whose application needs to work well with Dragon NaturallySpeaking and who needs help from ScanSoft engineers. For more information, see http://www.scansoft.com/naturallyspeaking/sdk/support/.

www.ScanSoft.com

Productivity Applications Division ScanSoft, Inc. 1 Wayside Road Burlington, MA 01803
2005 ScanSoft, Inc. All rights reserved. ScanSoft, Productivity Without Boundaries, Dragon NaturallySpeaking, and the ScanSoft logo are trademarks and/or registered trademarks of ScanSoft, Inc. or its subsidiaries in the United States and/or other countries. All other company names or product names may be the trademarks of their respective owners. WP06/05

Page 1 of 4

ScanSoft Ships Dragon NaturallySpeaking 7 Medical, Lowers Healthcare Costs through Highly Accurate Speech Recognition
Care Providers Can Use Speech to Enter Patient Information into EMR, HIS and Microsoft Office Applications, Reducing or Eliminating Manual Transcription Costs
PEABODY, Mass., April 15, 2003 - ScanSoft, Inc. (Nasdaq: SSFT), a leading supplier of imaging, speech and language solutions, today announced the immediate availability of Dragon NaturallySpeaking 7 Medical, a new release of the world's best-selling speech recognition solution designed specifically for healthcare organizations. Already in use at thousands of healthcare facilities worldwide, Dragon NaturallySpeaking converts speech into text at up to 160 words per minute, automating the input of patient information and saving thousands of dollars per doctor, per year, when compared to manual processes. Dragon NaturallySpeaking 7 Medical delivers accuracy levels as high as 99 percent, a 15 percent improvement over the previous release. It also includes new functionality important to healthcare organizations of all sizes, including eliminating the dependency on headset microphones through support for array microphones and Microsoft Pocket PC devices; the ability to speech enable electronic medical records (EMR) and healthcare information systems (HIS); and enabling enterprise-wide deployments by supporting network based administration of voice profiles, custom vocabularies and software installation. "Dragon NaturallySpeaking 7 Medical delivers the features and ease-of-use required to make speech recognition a mainstream solution for healthcare professionals everywhere," said Michael Feldman, MD of the University of Pennsylvania Health System. "ScanSoft has focused on delivering the most important features needed to cut transcription costs: highly accurate and easy-to-use speech recognition. Our organization saves time and money using Dragon NaturallySpeaking, and we get immediate access to data instead of waiting days to review and process patient information." Dragon NaturallySpeaking 7 Medical has built-in understanding for more than 250,000 words, and includes additional vocabularies covering terms found in general medicine, pathology, radiology, cardiology and surgery disciplines. Healthcare organizations can also easily add their own words to the solution to create fully customized vocabularies either on individual workstations or throughout a network. Care providers can take advantage of voice-activated shortcuts, allowing a single word or phrase to automatically populate fields in an EMR, or to simply generate full paragraphs of standardized text. For example, a voice-shortcut of "insert standard flu symptoms" could be used to populate an EMR system or Microsoft Word template with complete paragraphs of data. Over and above the accuracy and performance gains in the new release, ScanSoft has implemented new capabilities within the product that overcome historical limitations associated

8/6/2003

Page 2 of 4
with speech recognition in healthcare settings. New support for array and wireless microphones allow care providers to use the product in a more natural way, without being tethered to a PC by a headset microphone. Support for mobile dictation, a mainstay of the profession, has been expanded to process recorded speech from Microsoft Pocket PC devices, in addition to digital recording devices from Panasonic, Sony and others. ScanSoft has also delivered new features that add value to investments being made to automate processing patient information. The new release simplifies using the product to add speech input and voice control capabilities to commercial and in-house automation solutions, including EMR, PAC (Picture Archiving and Communication) and HIS solutions. The product also instantly speech-enables a wide range of commercial PC applications and is widely used to enter patient information into Microsoft Word templates and database applications - completely by voice. The new network administration capabilities within the product are of particular value to midand large-size healthcare facilities. Installation and maintenance of multi-seat and site deployments have been simplified through support for MSI (Microsoft Installer), while centralized management for voice profiles and custom vocabularies is made possible through the ScanSoft DNS Administrator. "The integration of speech recognition within EMR and HIS systems provides organizations with a powerful way to streamline the collection and workflow of patient information," said E. Ross Weinstein, President of Voicebrook Inc., a New York speech technology consultancy and software developer, specializing in the healthcare industry. "We congratulate ScanSoft on the new release of Dragon NaturallySpeaking Medical, as it delivers the accuracy, integration and network capabilities needed to bring the benefits of speech recognition to organizations of any size."

The World's Best Speech Recognition for Healthcare Professionals Made Better
To help healthcare organizations improve the quality and productivity of patient care, and reduce or eliminate manual transcription costs, ScanSoft has added a number of key capabilities to Dragon NaturallySpeaking 7 Medical, including:
Most Accurate Ever - Significant improvements to the Dragon NaturallySpeaking recognition system has yielded an unprecedented accuracy improvement of up to 15 percent over the previous release, resulting in speech recognition accuracy levels as high as 99 percent. Fastest Ever - More than 50 percent improvement in the initialization time of Dragon NaturallySpeaking 7 enables "10 seconds to dictation" on baseline systems. Users can train the software on their voice patterns in just five minutes. Extensible Integration with EMR, HIS and Microsoft Office - Nearly all standard Microsoft Windows applications, including EMR and HIS applications, are speech enabled simply by installing Dragon NaturallySpeaking, allowing the use of all interface menus and input fields simply by speaking. Dragon NaturallySpeaking 7 Medical comes with enhanced compatibility modules for the full Microsoft Office Suite and Corel Word Perfect. Mobile Dictation Support - Users can now "talk-and-dock" using standard Pocket PC devices, in addition to handheld digital recorders, to record mobile dictation for automatic transcription when they synch to their PCs.

Page 3 of 4

Support for Array Microphones and Wireless Headsets - This release represents the first Dragon NaturallySpeaking to support array and wireless microphones, giving users the freedom to dictate without being tethered to their PC by a wired headset. Integrated with Medical Vocabularies - Dragon NaturallySpeaking 7 Medical comes with preconfigured specialty vocabularies for general medicine, pathology, radiology, cardiology and surgery. Tools that are included with the software can be used to easily build a custom vocabulary for any user-specific lexicon. Speech Recognition for Developers and System Integrators -DNS SDK (Software Developers Kit) enables integrators to add custom speech capabilities to commercial and in-house EMR and HIS applications, and can be deployed on a central server for "off line" or "back end" speech recognition. New Vocabulary Optimizer - Users can instruct Dragon NaturallySpeaking to instantly analyze sentence structure and word use frequency in previously created documents. Dragon NaturallySpeaking then automatically adjusts the recognition engine to match the user's unique writing patterns, thereby further increasing accuracy. Improved RealSpeak Proofing - The time and effort needed to proofread a document created by voice has been significantly reduced with ScanSoft RealSpeak Proofing, which reads email and document text aloud in a human-sounding synthesized voice. Dragon NaturallySpeaking can also play back the original voice recording associated with any dictated document, which facilitates third party transcription editing. Expanded Custom Voice Shortcuts - The ability to create custom voice shortcuts, where a word or phrase will trigger the insertion of standardized text, has been expanded. Users can create a custom voice shortcut for standard letter closings -including a graphic of their signature - and have it automatically added to their email or Microsoft Word correspondence by simply saying the shortcut command (i.e. "Add signature") The inserted text now automatically matches the font of the destination document. Section 508 Accessibility - Dragon NaturallySpeaking 7 Medical can help people with disabilities achieve maximum productivity on PCs, as well as help prevent carpal tunnel syndrome and other musculoskeletal disorders. Dragon NaturallySpeaking 7 is the only PC dictation application that meets U.S. Section 508 mandates, which were established by the federal government to insure access to technology by people with disabilities. Network Administration - Dragon NaturallySpeaking Medical is available with a pre-configured MSI file for push installations throughout a network environment. The new ScanSoft DNS Administrator utility enables IT professionals to manage and update user profiles and vocabularies from a central network location.

"We are seeing provider organizations accelerate their investments in solutions that automate manual processes," said Robert Weideman, chief marketing officer at ScanSoft. "Whether driven by the desire to deliver improved patient care or to meet government mandates, such as the Health Insurance Portability and Accountability Act, Dragon Naturally Speaking 7 Medical is a powerful solution that enables the entire organization to become more productive."

Availability and Pricing

ScanSoft Dragon NaturallySpeaking 7 Medical is available immediately through ScanSoft's professional sales organization and global network of reseller partners. The suggested retail price is $995 for individual users, with multi-seat and site license discounts available. For

Page 4 of 4

additional information on features, pricing and volume licensing programs, please visit www.ScanSoft.com or call (1) 800-443-7077.

About ScanSoft, Inc.

ScanSoft, Inc. (Nasdaq: SSFT) is a leading supplier of imaging, speech and language solutions that are used to automate a wide range of manual processes - saving time, increasing worker productivity and improving customer service. For more information regarding ScanSoft products and technologies, please visit www.ScanSoft.com.
Trademark reference: ScanSoft, the ScanSoft logo, RealSpeak and Dragon NaturallySpeaking are registered trademarks or trademarks of ScanSoft, Inc. in the United States and other countries. All other company or product names may be the trademarks of their respective owners. The statements in this press release that relate to future plans, events or performance are forward-looking statements that involve risks and uncertainties, including risks associated with market trends, competitive factors, and other risks identified in ScanSoft's SEC filings. Actual results, event and performance may differ materially. Readers are cautioned not to put undue reliance on these forward-looking statements that speak only as of the date hereof.