Tagged Acquires Topicmarks To Improve Friend Suggestions With Natural Language Processing

Tagged’s mission is to help strangers meet each other online, so it has to offer friend suggestions of people you’ll like and who’ll like you back. That’s why it acquired Topicmarks, a natural language processing and machine learning company. Topicmarks will allow Tagged to analyze the profiles of its 100 million registered users and match them with others with similar interests and vocabulary. Topicmarks’ technology, CEO, CTO, and 3 senior engineers will join Tagged in exchange for cash and stock. Its existing service will remain active for the foreseeable future.

Tagged’s CEO Greg Tseng tells me the acquisition price was “somewhere in the middle” between covering Topicmarks’ $150K in seed funding and significantly impacting his online gaming and meeting network’s bottom line. Apparently it wasn’t just a graceful, low-payout exit for Topicmarks as many suspect Facebook’s acquisition of Gowalla was.

Tagged has been profitable for four consecutive years thanks to its 10 to 20 million monthly active users. It grew from 50 to 150 employees this year, including 17 from acquisitions of Topicmarks, game discovery platform WeGame, and social messaging client Digsby. One Topicmarks employee won’t be joining the Tagged.

Topicmarks’ existing service scans text and returns short content summaries. It process news article links, RSS readers, desktop files, cloud storage services like Dropbox and Box.net, and more. Now its technology will be applied to reading the profiles, messagees, comments, and statuses of Tagged users. It will look for what topics users talk about, their punctuation and emoticon usage, and whether they write in a more urban or rural fashion. It will then produce “bi-directional recommendations” of users who will both be interested in each other.

Internally, Tagged has been calling this a “Pandora for people”. Though really it’s more complicated. As Tseng explains, “When Pandora recommends music, the music doesn’t need to like you back.” More accurate friend suggestions could increase engagement, time on site, and interconnections between users that make Tagged sticky. They could also help Tagged fend off competitors such as 3-D avatar-based Shaker which won TechCrunch Disrupt 2011, and Badoo which now has 130 million registered users and $100 million a year in revenue.