MIT scientists have worked out a way to take online service reviews and turn the data into a voice-recognizing recommendation system–no matter how complex or specific your request. It’s a sneak peak of how a semi-AI smartphone of 2015 will help your daily life.
MIT’s news release notes “the proliferation of websites such as Yelp and CitySearch has made it easy to find local businesses that meet common search criteria” when the criteria are pretty simple–cheap food of a particular type in a particular location, for instance. But what about “not so common criteria,” MIT wonders–“how big are the portions? […] Does the bartender make a good martini?” This data is sometimes available in the body text of wordy reviews, but as we all know, it can be a time-intensive slog through search results to find it, often because we need to verify that a particular review is truthful.
MIT’s Computer Science and Artificial Intelligence Lab’s Spoken Language Systems Group has built a system that automatically “combs through users’ reviews, extracting useful information and organizing it to make it searchable.” Though the MIT algorithm looks for phrases like “excellent martinis,” the code doesn’t assign a value to words like “excellent.” Instead, it matches such epithets (whether they’re positive or negative) with the numeric ratings that some review sites allow–“good steak” and “9/10 stars” for example. It then takes this collection of ratings and descriptions and builds a sophisticated map that adapts to new reviews, and also winds in data from the service’s websites to extend its understanding.
The upshot is a program that can return results associated with “nice ambience” alongside ones that have a “friendly vibe” because it’s seen the words “good,” “nice,” “friendly,” or “mood” associated with highly starred reviews for a particular restaurant or bar.
This is a glimpse of how smartphones of the future will work. We’re seeing a growing reliance on services like Yelp, and associated recommendation engines–Google itself just announced a global release of its Hotpot location-centric recommendation system. Professional reviewers still have a key job to play, but peer review is getting more important, particularly when you’re trying to make on-the-spot decisions and don’t have time to trawl through the review pages of the New York Times. Mobile apps on smartphones are how more and more of us are doing this–and if Google gets involved with the same kind of research as MIT is, this is how our smartphones will help us find services in the future. It’s really sci-fi come true.
To read more news on this, and similar stuff, keep up with my updates by following me, Kit Eaton, on Twitter.