Joint Parsing and Named Entity Recognition

Human Language Technologies: The 2009 Annual Conference of the North American Chapter of the ACL, pages 326–334, Boulder, Colorado, June 2009. c⃝2009 Association for Computational Linguistics Joint Parsing and Named Entity Recognition Jenny Rose Finkel and Christopher D. Manning Computer Science Department Stanford University Stanford, CA 94305 {jrfinkel|manning}@cs.stanford.edu Abstract For many language technology applications, such as question answering, the overall sys- tem runs several independent processors over the data (such as a named entity recognizer, a coreference system, and a parser). This eas- ily results in inconsistent annotations, which are harmful to the performance of the aggre- gate system. We begin to address this prob- lem with a joint model of parsing and named entity recognition, based on a discriminative feature-based constituency parser. Our model produces a consistent output, where the named entity spans do not conflict with the phrasal spans of the parse tree. The joint represen- tation also allows the information from each type of annotation to improve performance on the other, and, in experiments with the OntoNotes corpus, we found improvements of up to 1.36% absolute F1 for parsing, and up to 9.0% F1 for named entity recognition. 1 Introduction In order to build high quality systems for complex NLP tasks, such as question answering and textual entailment, it is essential to first have high quality systems for lower level tasks. A good (deep analy- sis) question answering system requires the data to first be annotated with several types of information: parse trees, named entities, word sense disambigua- tion, etc. However, having high performing, low- level systems is not enough; the assertions of the various levels of annotation must be consistent with one another. When a named entity span has crossing brackets with the spans in the parse tree it is usually impossible to effectively combine these pieces of in- formation, and system performance suffers. But, un- fortunately, it is still common practice to cobble to- gether independent systems for the various types of annotation, and there is no guarantee that their out- puts will be consistent. This paper begins to address this problem by building a joint model of both parsing and named entity recognition. Vapnik has observed (Vapnik, 1998; Ng and Jordan, 2002) that “one should solve the problem directly and never solve a more gen- eral problem as an intermediate step,” implying that building a joint model of two phenomena is more likely to harm performance on the individual tasks than to help it. Indeed, it has proven very diffi- cult to build a joint model of parsing and seman- tic role labeling, either with PCFG trees (Sutton and McCallum, 2005) or with dependency trees. The CoNLL 2008 shared task (Surdeanu et al., 2008) was intended to be about joint dependency parsing and semantic role labeling, but the top performing systems decoupled the tasks and outperformed the systems which attempted to learn them jointly. De- spite these earlier results, we found that combining parsing and named entity recognition modestly im- proved performance on both …

Content truncated to 500 words