Facebook encourages us to create a social network including everyone we know. But it only captures one slice of our social lives. Our social connections and conversations sprawl across many other forms of communication, including face-to-face chat, e-mail, phone calls, and text messages.
Much of these communications is increasingly channeled through one device: the smart phone. A new app from a startup called Aro exploits this fact by digesting everything that takes place on a phone—from e-mails to call logs—to learn about all of your connections and friendships, and to track relationships that span different forms of communication.
“We’re building your true social network from all of your services on the phone, and your [social] graph grows with every new message,” says Andy Hickl, chief technology officer with the company, which is funded by $20 million from Microsoft cofounder Paul Allen.
Aro is currently in a closed beta and is available only for Android phones (you can apply to join here), but an iPhone app is in the works. Once installed on a phone, the software asks for access to your e-mail account and calendar accounts and accesses the handset’s call logs, contacts and stored text messages. It then scans through everything to look for social connections and stores a digest of the resulting web of links on a cloud-computing platform. Aro adds several tools to the phone that can tap into that network to make doing tasks like using information from an email to create a contact or calendar event easier .
One of these tools is for search. Tapping in a person’s name brings up a list of every call or message of any kind they received from you or sent to you. Aro can also find messages from other contacts that refer to that person, even if only by their first name. That’s possible because of software developed by Hickl and fellow engineers that can recognize people, places, companies, and dates mentioned in messages. This ability also surfaces in Aro’s e-mail, text messaging, and calendar tools.
Most commercially successful examples of this kind of “semantic” technology are used to mine very standardized text such as news reports, says Hickl. Aro has to handle the messy, misspelled, and abbreviated world of personal communications. The Aro team used collections of messages from Twitter and Facebook to help teach the software how to deal with that.
Aro also draws on the social network it has uncovered to refine how it handles the terms it recognizes. For example, if someone refers to a “Mike” or “that executive” in an e-mail, Aro tries to work out to whom this refers based on what it knows of the user’s network and other cues in relevant messages.
Anything that the Aro software identifies is highlighted and can be tapped by the user to bring up a Web-like menu of actions. These include searching for all other mentions of, and messages from, a person, or adding a new event into a calendar.
Gain the insight you need on artificial intelligence at EmTech MIT.