---------- ---------- PC Pro Realworld Computing: Paul Lynch - PDAs

Talking to that scribble thing

There are quite a few implications of installing voice technologies into a PDA, and all of them are very exciting. However, they fit into the general category of ideas whose time hasn't quite come. Text to speech is a great one; the technology to implement with almost acceptable quality was available on the BBC microcomputer and even the original Macintosh, and yet there isn't an implementation available for any current PDA; it is such a crucial component of voice technology that this really should be very surprising. Voice activated controls (not speech recognition, which is a different story altogether) have been widely available at the toy level for PCs for some time. Usable speech to text is harder, but even that has had reasonably successful implementations (the Dragon Dictate system available from Dragon Systems, for example) for several years. The only missing piece has been natural language processing, which shows every indication of being even further away from realisation as a competitive ches

A PDA that fully supported voice input and output could be made without a screen or keyboard, with the only alternate source of input a serial port or Ethernet connection. You wouldn't need to draw attention to yourself by putting it on the conference room (or pub) table, although if you really wanted to do that I am sure a headset would do the job nicely. However, you would be giving up the now familiar and cosy windows style of display, and a completely new set of conventions defining a voice user interface would have to evolve. One key to this might be natural language processing, although speaker dependent recognition has already proven that voice input can be a reasonable way of controlling a computer. If you have used a machine with voice control like that, then you will recognise the feeling of frustration that comes when telling some one on the phone, or in the same room as you and your computer's microphone, how to use the system - and it starts doing what you are describing!

When details of the Newton 2000 were first released, the Note recording and playback capability received some attention in the press, and it almost seemed as if Apple had promised text to speech and speech to text capabilities as part of the design. As the release date approached, promises dropped back and it appeared that Apple might eventually release a text to speech package of their own, and some third party, presumably Dragon Systems (ironically based in Newton, Massachusetts), were working very secretly on a port of their speech recognition system. However, rumours now indicate that Dragon have dropped this development, apparently claiming that the restricted memory available in the Newton OS design prevents them, probably with justification, from proceeding.

To record a voice note on the Newton, you have to create a new note with a stationery type of Recording (other stationery types available are Checklist, Outline or Note; the latter of which is for plain text and drawings). This then gives you a set of three buttons at the top of a normal text note: record, play and stop. There is a bar to indicate the progress of a playback, and a timer next to it which shows how long in seconds a recording is. You can write in the rest of the note, just like a plain text note; this could be use to manually transcribe a short recording, for example.

{Newt-record.bmp - "Recording stationery for Notes on Newton"}

Recording stationery for Notes on Newton
Recording stationery for Notes on Newton

One good point of the Newton recording is that it is economical with space; a four second recording, for example, took up 8K as a note. Playback quality is acceptable, although the volume is quite low compared to the input speech.

The user interface, however, is very sketchy. What is there is good, but it doesn't go far enough: compared to the full featured user interfaces of the key chain voice recorders, or of David Joyce's MemoVoc on Psion, there is a lot missing. There is no way to quickly move from one voice note to another, and pen only control is less convenient than key or button operated controls on the other devices. There is certainly room for a third party application on the Newton to do properly what the Newton Systems Division have started, and the efficient use of storage would make the Newton an ideal candidate for this application.

Words and design by:
Paul Lynch
Last updated: July 30, 1997

[PDA Columns | Home Page | Company]