SPEECH RECOGNITION
Multi-threaded NewsWatcher

Contents


Speech Recognition

MT-NewsWatcher incorporates speech recognition capabilities, and is probably the first speech-controlled Usenet newsreader. Speech Recognition enables the program to respond to certain spoken commands, which is useful for those with disabilities and those who want a rest from the keyboard.

Setting up

Speech recognition requires that you have Apple's Speech Recognition software installed, with version 1.5 or later of the Speech Recognition extension. SR will only run on a Power Macintosh.

Please do not contact the author of MT-NewsWatcher with questions related to the installation and troubleshooting of Speech Recognition software.

To start up Speech Recognition (hereafter: SR) in MT-NewsWatcher, go to the Speech Recognition option in Preferences. Check the box to enable SR. (If the check box is disabled, it means that SR software is not installed, or that the version of the extension is prior to 1.5.) This preferences panel is not available in the 68k version (since you can't use SR anyway on such machines).

The Speech Recognition Preferences panel. In this case, the 'Use Speech Recognition' check box is disabled, because SR is not installed. The scrolling list shows you the available speech-controlled commands, with the action on the left, and the speech string on the right (these are the words that you say to perform the action).
The scrolling list in that preferences panel shows you what commands are speakable. You can edit the speech strings by clicking on a list item, tabbing to the edit field, and editing the text.

When changing the speech commands, you should bear in mind that speech recognition is not foolproof. It is best to choose commands such that they sound as different as possible. For example, have two commands, "Get next article" and "Get next group", is not a good idea, because these two strings are difficult to distinguish. It would be much better to have "Get next article" and "Open next group".

Available commands Current speech commands are:
  Get next article            Get the next article (same as 'Next article' menu item)
  Read this article           Open selected article in a subject window, else as above
  Show next thread            Same as 'Next thread' menu item
  Next group                  Same as 'Next group' menu item
  Open this group             Open the selected group in a group window
  Move up                     Move up in the group/subjects list
  Move down                   Move down in the group/subjects list
  I read it                   Mark as read
  Mark unread                 Mark as unread
  Expand thread               Expand a thread in a subject window
  Collapse thread             Collapse a thread in a subject window
  Reply to this               Reply
  Forward this                Forward
  Are there any new groups    Check for new groups
  Show me new articles        Check for new articles
  Close window                Close a window
  Save this                   [not implemented yet]
  Help me                     Show help
  Set preferences             Show preferences dialog
  Tell me about News Watcher  Show about
  Quit News Watcher           Quit NewsWatcher
  

In addition to these basic commands, the following windows add additional capabilities:

Group windows:
In group windows, you can speak the name of a group (without the dots, for example "comp sys mac apps") to open that group.

Beware that the SR software is not good at differentiating between similar-sounding group names.

Subject windows:
When a subject window is open, you can (try to) speak the first few words of a subject to open that article. MT-NewsWatcher will open the first article of the thread, if the article is a thread member.

The speech commands are built from the first 3 words of the subject, ignoring non-letter characters, and ignoring any "Re: " parts. Subjects which are not plain english (e.g. "OT/PPP 1.1.1 available") are best selected with "Move up", "Move down" and "Read this article". (The latter string, I discover as an Englishman, requires quite an American twang on "article" to work.)

Potential problems

The SR software does not work well with language model with more than a couple of dozen entries; when the model is built for the subject window, each visible article gets an entry. Thus, SR will work best in subject windows with 50 or less items. You can set the "Max articles to fetch" preference item to ensure that your subject windows do not get too big.

More Info

For more information about Speech Recognition on the Macintosh, check out these sites:

MT-NewsWatcher


Download Basics Speech Recognition Filtering Misc options