Paper: A simple feature-copying approach for long-distance dependencies

ACL ID W09-1124
Title A simple feature-copying approach for long-distance dependencies
Venue International Conference on Computational Natural Language Learning
Session Main Conference
Year 2009

This paper is concerned with statistical meth- ods for treating long-distance dependencies. We focus in particular on a case of substantial recent interest: that of long-distance depend- ency efects in entity extraction. We intro- duce a new aproach to capturing these efects through a simple feature copying preproces, and demonstrate substantial performance gains on several entity extraction tasks. 1 Long-distance dependencies The linguistic phenomena known as long-distance dependencies have a long history in computational linguistics. Originaly arising in phrase-structure gramar, the term aptly describes phenomena that are not strictly gramatical, and has thus gained curency in other endeavors, including that of con- cern to us here: entity extraction. The comon thread, h...