Identifying non-referential it: a machine learning approach incorporatinglinguistically motivated patterns

摘要

In this paper, we present a machine learningsystem for identifying non-referentialit. Types of non-referential it are examinedto determine relevant linguisticpatterns. The patterns are incorporatedas features in a machine learning systemwhich performs a binary classification ofit as referential or non-referential in aPOS-tagged corpus. The selection of relevant,generalized patterns leads to a significantimprovement in performance.