Social networks allow users to create virtual identities online, typically following an advertisingbased business model. Users may create multiple accounts on different social networks, so that each social network platform has access only to the information a user posts there, rather than a full set of profile information available across all networks. Advertising schemes on social networks could be more effective if networks could see the full set of information an individual posts across several platforms. This requires matching user identities across platforms, which is challenging because users may not identify themselves in exactly the same way on each platform. Existing methods primarily rely on static profile data to match user identities. In this work, we propose to extend existing identity matching methods based on a topic modeling approach using dynamic post data. We implemented our method, and evaluated its effectiveness using profiles drawn from Facebook and Twitter. Our initial evaluation showed an accuracy improvement of 8% in precision and 21% in recall when topic modeling is considered, compared with results based on static profile characteristics alone.