An Entity-Centric Query Expansion Approach to Cumulative Citation Recommendation in Knowledge Base Acceleration

Most existing Knowledge Bases (KBs) are maintained manually and difficult to be kept up-to-date. To address this problem, Cumulative Citation Recommendation (CCR) is introduced to filter relevant documents for target entities from a temporally ordered stream of documents and detect centrally relevant documents as candidate citations. This paper proposes an entity-centric query expansion approach to address CCR task. Given a target KB entity, we first extract its related entities from three sources, including its profile page, existing citations in the KB and annotation documents. Then the three kinds of related entities are leveraged as expansion terms in different combination strategies. Extensive experiments have been conducted on KBA-CCR-2012 dataset, and experimental results have validated the effectiveness and robustness of our approach.