Speech Recognition

Speech recognition software is one of the fastest growing areas of computer science. This area has been continuously becoming more accurate and affordable. But it still has a way to go. Mainly because...

Finally!  Inept robots have replaced inept employees.

It's like word association with a sociopath.

Just The Facts

  1. Programmers have been attempting to create a perfect speech recognition program since the 1950's.
  2. The first speech-to-text program was created by Kurzweil Applied Intelligence in 1985.
  3. One of the goals driving advancement of speech recognition was ease of living for the disabled.
  4. Much like a wheelchair at a Wal-Mart, what was intended for the disabled has been adopted by the lazy.


The first speech recognition program was created by RCA labratories in the 1950's. It was capable of understanding ten syllables from one user. Many companies seem to use this model today. Progress was slow until the 1970s when DARPA decided to get involved. Enter the Speech Understanding Research program, better known as SUR.

They weren't comfortable addressing their superiors by their first names.

DARPA and Carnegie Mellon University created two speech recognition computers named Harpy and Hearsay. These systems were able to identify just over 1,000 words. That is to say, they had a vocabulary half the size of the average five year old and about a tenth the chance of hilariously falling over (I am a terrible person). Unfortunately for SUR, DARPA was not looking for recruits with the intelligence of a below average toddler and scrapped the project in 1976.

1980's to Present

In 1985, Kurzweil Applied Intelligence (now Kurzweil Technologies) created the first "useful" text-to-speech program which knew 1,000 words. In 1987, they released a program that knew 20,000 words. Both these programs had the flaw of needing the user... to... pause... after...each... word. It soon became abundantly clear that William Shatner was only going to buy one copy, so they focused on making the product more available to consumers everywhere.

Rocket... Man...

The... Future... Is... Here.

Speech Recognition technology steadily advanced since then. In 1999, DARPA got involved again. This time wanting to use technology to create a translator for the military. Called the Phraselator, it offered one way translations for soldiers. The Phraselator was validated for use on September 10th, 2001. I look forward to reading the insightful comments about controlled explosions and holographic planes. It is good to know that the voices returned after you stopped taking the brain-control pills the doctors were giving you. But that's an issue for another time and dimension.

You know what you must do

It says your neighbors are among them