You have to click an icon to activate voice... I'm sure you can write a quick Python script to compare new audio input, stored in a watched directory, to a base

You have to click an icon to activate voice... I'm sure you can write a quick Python script to compare new audio input, stored in a watched directory, to a base audio dictionary that fires an API based on normalized matches to "terms" in that dictionary. If you say "search" the script compares/approximates to existing waveforms. We should have a Speech-Computer Interface API, rather than boring ourselves with the standard "Search for...", that makes a canonical set of speech patterns which computers should respond to habitually. Things like "I'm thinking..." (fires off search, feeling lucky, wikipedia, etc. mashup results pages), "I'm hungry..." (fires off search, geolocation information, personal biometric data, etc...), "Who is..." (wiki? pipl? facebook?), "What is..." (wiki?), "When is..." (shows a calendar)...