Entity Recognition for Duplicate Filtering.

We propose a system for automatic detection of duplicate entries in a repository of semi-structured text documents. The proposed system employs text-entity recognition to extract information regarding time, location, names of persons and organizations, as well as events described within the document content. With structured representations of the content, called "metamodels", we group the entries into clusters based on the similarity of the contents. Then we apply machine-learning algorithms to the clusters to carry out duplicate detection. We present results regarding precision, recall, and F-value of the proposed system.