We all have that person who is just impossible to buy for. Either they already have everything or maybe you just don’t know them well enough to know what kind of fun gift they’d enjoy. And you can’t ask – that’s tacky.

Well, if they tweet you can find something that’s very “them.” So very them, in fact, they’ll wonder how you knew to buy it. But don’t give away your secret, just smile evenly and maintain eye contact for a full five minutes.

What wizardry is this?

The folks at Knight Lab, Shawn O’Banion, a Northwestern Ph.D. Computer science student, and Larry Birnbaum, a Northwestern professor of computer science, who brought you Tweetcast just in time for the U.S. Presidential Election now bring you BookRx – just in time for the holidays!

BookRx tells you which books someone will likely love based on their tweets. And it offers a varied selection sorted in categories that change based on the user’s interests: Business, Fantasy & Science Fiction, Fiction, Humor, Mystery, Romance, Social Sciences & Politics, and Sports & Fitness.

BookRx shows the categories that each user is most likely interested in and does so in order (eg. if someone is most interested in Business, that category will be displayed first. If someone has very little or no interest in Romance, the system likely won’t show results from that category at all.

The technology learns what words, hashtags, and users are correlated with the various categories/genres by analyzing the Twitter behavior of users that we know enjoy books about Romance, Mystery, Sports, etc. We can then apply this model to new users by looking at their own tweets to make recommendations. The terms that show up for any given category are those that are distinctly associated with readers of that particular genre.

Basically the technology tracks words, @-mentions, hashtags and links in the Twitter feeds of people who we know had an interest in specific books and compares same elements in the feeds of people who enter hashtags in BookRx. The greater the similarity between the two, the more likely that the book recommendations will be valuable.

(Surprised image from Shutterstock)