As the number of social networking services (SNS) and their users grow, so does the complexity of individual networks as well as the amount of information to be consumed by the users. Users of SNS exchange short and instantaneous messages interactively, which can be seen as conversations. We explore this conversational aspect of SNS and show how refined topic-based semantic social networks can be formed in order to reduce the complexity and information overload. Among other possibilities, we use the notion of topic diversity and topic purity of SNS conversations between two users and show different types of social relationships can be identified in that they break down a huge “syntactic” social network into topic-based ones based on different interaction types. Resulting semantic social networks can be useful in designing various targeted services on online social networks.