Apple Thursday published a new entry on its Machine Learning Journal blog which goes highly technical on Siri’s ability to recognize obscurely-named points of interest on a map, like your local restaurants, businesses and other places.
“The accuracy of automatic speech recognition systems has improved phenomenally over recent years due to the widespread adoption of deep learning techniques,” notes the post.
“Performance improvements have, however, mainly been made in the recognition of general speech; whereas accurately recognizing named entities, like small local businesses, has remained a performance bottleneck.”
Apple has met that challenge by incorporating knowledge of the user’s location into its speech recognition system and so-called geolocation-based language models (Geo-LMs).
As a result of the combo, Siri is able to better estimate the user’s intended sequence of words.
This has reduced Siri’s error rate by between 41.9-48.4% in Boston, Chicago, Los Angeles, Minneapolis, New York, Philadelphia, Seattle and San Francisco, excluding mega-chains.
In the United States, Apple has one Geo-LM for each of the 169 Combined Statistical Areas that cover about 80 percent of population. There’s also a global Geo-LM that covers all areas which are not defined by the Combined Statistical Areas around the world.
The combination of location and Geo-LMs lets the system provide customized results in terms of the names of points of interest, or fall back to the global Geo-LM if location is unavailable.
All you need to know: Siri’s regionally-specific language models for speech recognition make finding local destinations a breeze. The method is language-independent, meaning the expansion of Geo-LM support for other locales besides US English is pretty straightforward.