Semantic Text Mining with Linked Data

Linked Data is an open data space that emerges from the publication and interlinking of structured data on the Web using the Semantic Web technologies. How to utilize this wealth of data is currently a focused research theme of the Semantic Web community. In this paper, we aim to utilize Linked Data to generate semantic annotations for frequent patterns extracted from textual documents. First, we extract semantic relations from textual documents and merge them into a set of semantic graphs. Then, we apply a frequent subgraph discovery algorithm on the set of graphs to generate frequent patterns. Finally, we annotate the discovered patterns using Linked Data. Our approach can be applied in such domains as terrorist network analysis and biological network analysis. The efficacy of our approach is demonstrated through an empirical experiment that discovers and validates relationships between political figures from large number of news on the Web.