Semantic and social spaces: Identifying keyword similarity with relations

Social media has two essential building blocks: content and people. These two components form a heterogeneous network with various types of relations: people produce and share content, various content items are related to each other, and people have social relations such as collaboration and discussion. Identifying people’s expertise and topics is the first step in evaluating the quality of information in a network. Text semantic analysis and social network analysis address this issue from different perspectives. Text analysis aims at constructing concept similarity networks based on documents and keywords within the documents. Linguistics or keyword co-occurrence is used to detect word associations. Social network analysis utilizes network structures to find prestigious individuals who connect with other experts in a relational network. This study proposes a three-layer framework to integrate the semantic and social networks in order to reveal peoples expertise based on the words they use and their relations. Using social tagging activities on CiteUlike as an example, we illustrate how social relations help identify similar concepts in semantic networks.