A View from Erica Naone
An Intelligent Software Assistant Debuts
Now, you can try the software out for yourself, as the app has be added to Apple’s app store.
Siri, which the company’s CEO Dag Kittlaus describes as “the mother of all mashups with a big brain in the front,” tries to perform all sorts of useful tasks based on simple voice or text commands. The iPhone version can, for example, be used to find upcoming local events, make reservations at a restaurant, or check the weather. The voice technology is licensed from Nuance and Siri’s core technology, which allows it to understand, classify, and respond to user requests, is licensed from the CALO project–a massive artificial-intelligence effort that has made major strides in machine learning in recent years.
Siri responds impressively to fairly vague requests. For example, I asked it, “Who’s playing tonight at the Paradise?” (The Paradise Rock Club is a music venue located in downtown Boston, but I didn’t give Siri that context). The software returned a list of music venues with the Paradise at the top, and, after a single click to confirm that this was correct, showed me who was on tonight. In contrast, entering that same vague question into a search engine yields completely nonsensical results.
I also had good results from questions such as, “Is it snowing now in Annapolis?” and “Where can I buy guitar strings?” I was particularly impressed with the latter, since Siri had to recognize that guitar strings are something I would buy at a store that sells musical instruments, and then identify likely locations from there. It offered me a list of stores within walking distance of my current location.
I did manage to trip the software up a bit a couple of times though. For example, I asked, “Where’s the best place to get sushi in Allston?” and, as often happens with voice-recognition software, the app offered me Austin, TX, instead of Allston, MA. I was able to correct the city name easily, however, and Siri then offered me top-rated sushi restaurants, using ratings taken from Yelp. It was also pretty easy to request reservations at these restaurants using the app.
Another error occurred when I asked for “action movies at Boston Common,” which was an admittedly casual way of referring to the AMC Loews Boston Common movie theater. Siri understood what I was asking for, but thought that “common” was the name of an actor I wanted to see. Again, it was fairly easy to correct the query and get the question answered.
These examples certainly show how Siri is “a great big mashup.” After figuring out what I want, the software relies on integration with a variety of Web services, including Yelp, OpenTable, and airline websites.
The company plans to release versions for Blackberry and Android devices soon.
Keep up with the latest in natural language processing at EmTech MIT.
Discover where tech, business, and culture converge.
September 11-14, 2018
MIT Media Lab