While not specifically named in U.S. Patent No. 8,296,383 for "Electronic devices with voice command and contextual data processing capabilities," the invention is closely associated with how Siri operates and may be, at least in part, powering the voice-recognizing assistant.
First filed for in 2008, well before Apple purchased Siri in 2010, the patent describes a method and system in which a portable electronic device like the iPhone records a voice command and corresponding contextual information. The iPhone then sends the data to an off-site server for processing, which either responds to the voice command or sends back information to the handset that allows the device to execute the command.
Flowchart illustrating '383 patent's operational steps. | Source: USPTO
At the time, some of Apple's iDevice products used on-board voice commands to play music, tell time, or in the case of the iPhone, call contacts. While useful, the system was limited to certain operations due to the processing power and storage requirements necessary to respond to advanced user commands.
With the advent of fast data networks, Apple's patent allows for more sophisticated commands to be off-loaded from the device and carried out remotely in near real-time.
Key to the invention is contextual information, which is referred to as metadata, regarding the operating state of the device. For example, a user can perform an operation to look for more songs like the one that is currently playing on the handset by saying, "Find more like this." Another example would be location data, which can be used to "find nearby American restaurants," or similar commands.
It is not known if the '383 patent's technology is being used as a basis for Siri, however the virtual assistant is known to work in much the same way. Apple recently filed a patent application for the features that power Siri, including the ability to recognize conversational input.