Sysomos Product Update: Gender Analysis for Twitter and Facebook

By Sheldon Levine - Tuesday, May 31st, 2011 at 9:22 am  

Once again, we at Sysomos are pleased to bring you another big and unique update to both our MAP and Heartbeat software.

Unlike many of the other social networks, Twitter users are not asked to reveal their gender. On Facebook, users identify themselves as male or female, but that information is not publicly accessible. However, people who are trying to reach specific target audiences still want this information. So, over the weekend we updated both MAP and Heartbeat to include a new analysis feature that will help identify the gender for Twitter and Facebook users in an easy to read graph.

How are we able to do this? We spent a long, long time creating a database of names. Our list contains common names in English and many languages from around the world. In addition to just reading names, our latest analysis feature looks for clues in a user’s profile and content. There always are cases where a machine can not judge if a user is male or female, e.g., ambiguous names or brands. In a case like the name Jamie, which could belong to either a male or female, we look for clues in their profile like “mother of 3″ to determine that this Jamie is a female. We have tuned our algorithm to be conservative such that it makes almost no mistakes. Using disclosed names along with clues from user profiles and content, it is able to accurately classify two users out of three, with the remaining one-third being marked as “don’t know”.

This latest update is one of the most unique in the entire industry, especially with such a high level of accuracy. We think that everyone using our software to accurately approach and/or research their target audiences will find this feature extremely useful.

As always, if you are a current Sysomos subscriber and have any questions about our latest updates, don’t hesitate to contact your account manager.

Keep your eyes here on the blog for more updates coming in the future.

Tags: , , , ,