12/8/95 -- Note to developers regarding Apple╒s Speech Recognition Toolbox:
We are pleased to seed you with version 1.5a1 of the Apple's new Speech Recognition Manager. This is alpha software and documentation -- not ready for wide distribution or release in products yet. When finished, version 1.5 of the Speech Recognition extension will embody the first officially supported developer release of the Speech Recognition Manager.
The draft documentation included with this note, "Speech Recog Draft 11/21/95.pdf" in the "SR 1.5a1 Documentation" folder, is an Adobe Acrobat file. The documentation describes version 1.5 of the Speech Recognition extension and the API and toolbox it embodies. (Version 1.5 contains many features that were not available in version 1.4.1 or earlier versions of the extension.)
Input from early adopters of Apple's speech recognition technology has greatly improved the Speech Recognition Manager API. If you find bugs in the extension or documentation, please let us know:
AppleLink: PLAINTALK
Internet: PLAINTALK@applelink.apple.com
Among the new features in version 1.5 of the Speech Recognition Manager:
Ñ Routines for saving and loading language models to and from handles or files.
Ñ Better control over whether your application uses the feedback window and the listening method preferences (like the push-to-talk feature that enables users to leave the recognizer on all the time without worrying about misfires) set by users with the Speech control panel. (The Speech control panel is part of the English Speech Recognition package in PlainTalk 1.4.1, which can be found on Apple's sites on the net. The location is given below.)
Ñ SRIdle () call that can be used by applications that do a lot of processing without calling WaitNextEvent regularly. Most applications do not need to call SRIdle (), but applications like fast-action games that rarely call WaitNextEvent and do a lot of processing (and that use the Speech Recognition manager) should call SRIdle () often so the recognizer gets a chance to process sound and send results.
Ñ Ability to have the recognizer recognize from a 16bit/22.050kHz AIFF sound file -- useful during development, to create reproducible test cases.
See the documentation mentioned above for more details on these and other features.
The Speech Recognition extension runs on Power Macintosh computers with 16-bit sound in running System 7.5 or later. (Version 1.5a1 of the Speech Recognition extension even runs on the new PowerBook 5300s!) You will need a PlainTalk microphone or microphone of similar quality. The older omnidirectional microphones that were once provided with Macintosh computers will not work.
To use version 1.5a1 of the Speech Recognition extension, you might want to first install version 1.4.1 of English Text-to-Speech (to enable speech synthesis using the new SRSpeak... routines) and English Speech Recognition (to get the Speech control panel), which are two of the components of PlainTalk 1.4.1. You can find PlainTalk 1.4.1 on Apple's Internet, eWorld, CompuServe, and AppleLink tech-support sites, in areas related to U.S. Macintosh System Software updates. For example, on the Internet, look at ftp site:
ftp.info.apple.com
on the following directory path:
[Apple Support Area] [ Apple Software Updates] [ US] [ Macintosh] [System]
Then you just need to drag version 1.5a1 of the Speech Recognition extension to your extensions folder -- replacing version 1.4 of the extension -- and reboot.
The sample code provided with this note is a good place to look first for simple examples of how you can use the Speech Recognition Manager. The "Tiny SR Example" is a particularly simple, and complete, example.
