Improving Knowledge Discovery in Document Collections through Combining Text Retrieval and Link Analysis Techniques

In this paper, we present Concept Chain Queries (CCQ), a special case of text mining in document collections focusing on detecting links between two topics across text documents. We interpret such a query as finding the most meaningful evidence trails across documents that connect these two topics. We propose to use link-analysis techniques over the extracted features provided by Information Extraction Engine for finding new knowledge. A graphical text representation and mining model is proposed which combines information retrieval, association mining and link analysis techniques. We present experiments on different datasets that demonstrate the effectiveness of our algorithm. Specifically, the algorithm generates ranked concept chains and evidence trails where the key terms representing significant relationships between topics are ranked high.