Tech Industry

Apple could smarten up Siri using crowdsourcing

A newly-published patent application describes how Apple might use information from actual human beings to supplement Siri's own knowledge, or lack thereof.

Screenshot by Rick Broida/CNET

Apple device users who are frustrated when Siri fails to answer a question may one day be able tap into the crowd to get a better response.

The Apple patent application "Crowd sourcing information to fulfill user requests," published Thursday by the US Patent and Trademark Office, envisions a few ways to help Siri when it comes up empty.

Siri typically relies on such sites as Bing, Yelp, and WolframAlpha to gather the data needed to answer a question. But sometimes those sources aren't equipped to provide an accurate response, especially if the question itself is phrased in a way that seems to confuse Siri.

Instead, Apple's invention would use other sources, such as expert information services, general information sources, and online Q&A forums to provide data. Questions and answers could also be stored in a crowd-sourced knowledge base that Siri would access. Such a knowledge base could be searched using natural language queries, according to the patent application, offering a better shot at answering questions that would otherwise stump Siri.

These other sources could come into play automatically if Siri is unable to answer a question, or you can select a specific source to check when asking the question. You could also rate the response to your question, which would then help Siri learn from its hits and misses.

But the harder the question, the longer you may have to wait for a response. The patent application suggests a range of time frames, from several minutes to several weeks, depending on the nature of the request and time needed to crowdsource useful information.


As Apple describes it in the usual patentese:

A user request is received from a mobile client device, where the user request includes at least a speech input and seeks an informational answer or performance of a task. A failure to provide a satisfactory response to the user request is detected. In response to detection of the failure, information relevant to the user request is crowd-sourced by querying one or more crowd sourcing information sources. One or more answers are received from the crowd sourcing information sources, and the response to the user request is generated based on at least one of the one or more answers received from the one or more crowd sourcing information sources.

A patent application doesn't necessarily mean the technology will reach the consumer market. But crowdsourced information seems like a logical next step in Siri's ongoing evolution.

(Via AppleInsider)