When using MindMeld and other software programs that leverage speech recognition technology, there is a lot of work that occurs under the hood to transform a person’s speech into signals a computer can understand. Here is an extremely simplified breakdown of how speech recognition works, courtesy of "How We Talk To Computers:”
- First, speech is translated into signals.
- The program then isolates these signals into phonemes, or a word’s sound units.
- Using statistical probability, a likely guess is made. This guess is calculated by comparing the word against the existing corpus (or vocabulary) of the entire system.
Click here to explore more of the steps involved in making software understand the nuances of human speech.
(via Medical Transcription, h/t Visually)