Automatic speech/voice recognition software for the Oracle professional Oracle Tips by Burleson Consulting January 19, 2008

The 21^st Century steno pool – Jott

There are few tools that are not a re-hash of older technology, but one tool that shows great promise is the Internet-based “Jott” tool. (Thanks Steve Karam, for bringing this useful dictation software to my attention!).

Jott is a great way to do direct cell phone to e-mail transcription, and after a short set-up you can effortlessly dictate for up to 30 seconds and have the transcribed text sent via e-mail to anyone you wish.

Few executives today remember the “steno pool”, where roomfuls of young women honed their skills with Gregg Shorthand.  The old phrase “take a letter” is long-gone, and the art of stenography has gone the way of the dinosaur.

"Ms. Jones, please come have a seat on my lap and take a letter"

The original voice recognition were plagued with problems and they were unable to differentiate between speaking and background noise, leading to botched transcriptions like this:

Background noise plagued early speech recognition software

Lets take a look at where we came come since the nascent beginning of ASR (VR), more than 15 years ago.

A brief history of voice recognition

I remember talking with the early developers of speech recognition interviewing a brilliant quadriplegic programmer in Toronto who originally invented the first speech recognition software out of necessity.  Back them, the problem was always RAM shortages, as a good voice recognition (VR) tool must be able to cache huge amounts of rules in order to be performant.

Back in the early 1990's I was asked to write some article for Computerworld magazine and I vividly remember by research for a 1994 Computerworld article on automatic speech recognition (ASR) technology.

"ASR provides a task-driven interface to the computer, but programmers need an entirely new skill set to create user-friendly systems,'' says Dan Thompson, co-founder of KolVox."

Dan was an inspiration to me, and I was happy to know that Dan Thompson worked as an inspirational speaker, telling people how he overcame adversity with sheer will and programming skill.

But voice recognition has come a long way in the past decade.

Jott promised to be a very functional tool for the busy executive because it allows you to use your cell phone to dictate messages (but only up to 30 seconds) and automatically transcribe your verbal message and send it via text-messaging or e-mail. 

Testing Jott Voice recognition

I’ve been waiting for 30 years for voice recognition technology to become a reality, and there has yet to be a tool that will fully conquer contest-sensitive grammar. 

For more details, see my notes on context-sensitive grammar and my 1994 Computerworld article on automatic speech recognition (ASR) technology.

But Jott comes close to being quite useful.  On my cell phone, I had this interaction to send myself a e-mail with a note. 

Jott is great for transcribing cell phone dictation

While Jott is great for ordinary transcription, how does it work for complex grammar?  For a more advanced test, I interjected three phrases that can be confusing to voice recognition software:

• Complex terminology – I asked for a “tachistoscopic episcatistor”, a super-fast camera and projector shutter that is used in psychology perception experiments.
   
• Sounds-alike phrase – I used the term “Courier and Ives”, which some VR tools will transcribe as “Career knives”.
   
• Context-sensitive sounds – The age-old test has always been “Please write to Mr. Wright right now.”

Here is a sample session with Jott.  I used a very clear enunciation (I'm a licensed auctioneer) and my dictation was clear, crisp and free of background noise:

1. Me - Press voice-dial button on cell phone

2. Phone - “Who do you wish to call?
3. Me - "call Jott"

3. Phone - "Calling Jott"

4. Jott - “Who do you wish to Jott?”  self

5. Me “Jott self” 
6. Jott - "Jotting self"
7. Me - "I need a tachistoscopic episcatistor and a Currier and Ives print for the meeting.  And please write to Mr. Wright, right now."

6. Me - Hang up

Here is what I received in my e-mail inbox:

I need [Unclear speech] tester(?) and the courier in i-Sprint(?) for the meeting. And please write to Mr. Wright, right now.

Conclusions on Jott

Jott is great for most common communications, but there are some areas for improvement of the transcription software:

• Note that Jott did great with the context sensitive grammar (Please write to Mr. Wright right now". 
   

• Jott did did not know the programmable "cue" that Courier and Ives go together, transcribing "courier in i-Sprint" instead of "Currier and Ives Print".
   

• Jott failed my test on complex terminology.  This is an easy software fix by adding scientific words to their dictionary. 

As of January 2007, the Jott software will not personalize and accept feedback on mistakes or learn your voice cadence, but that may be coming soon.

Overall, I'm finally ready to concede that voice recognition has matured to the point where it's is usable.

If you like Oracle tuning, you might enjoy my book "Oracle Tuning: The Definitive Reference", with 950 pages of tuning tips and scripts.  You can buy it direct from the publisher for 30%-off and get instant access to the code depot of Oracle tuning scripts.