Recognizing Names by the Meaning of Defining Syntactic Structures

We present a named entity recognition system that relies on interpreting the meaning of the syntactic structures that define named entities. The syntactic heads of the defining structures are learned from the training data and are later used to interpret the meaning of the syntactic structures in the test data. Semantic features are extracted based on the interpretations and are used to build a decision tree classifier. The mapping phase iteratively matches previously labeled names with aliases and conjunctions in the same document. Evaluation on two test sets shows this system can generalize to some extent and perform moderately well.