Learning Reliability of Parses for Domain Adaptation of Dependency Parsing

The accuracy of parsing has exceeded 90% recently, but this is not high enough to use parsing results practically in natural language processing (NLP) applications such as paraphrase acquisition and relation extraction. We present a method for detecting reliable parses out of the outputs of a single dependency parser. This technique is also applied to domain adaptation of dependency parsing. Our goal was to improve the performance of a state-of-the-art dependency parser on the data set of the domain adaptation track of the CoNLL 2007 shared task, a formidable challenge.