Applying machine learning toward an automatic classification of It

Abstract

In the majority of cases, the pronoun it illustrates nominal anaphora, tending to refer back to another noun phrase in the text. However, in a significant minor-ity of cases, the pronoun is used in exceptional ways that fail to demonstrate strict nominal anaphora. The identification of these uses of if is important in all fields where pronoun resolution has an impact After a survey of previous treat-ments of the pronoun it in the literature, some features of instances of it are proposed that can be used in a novel memory-based learning method to auto-matically classify those instances. On evaluating the method, it is found that the implemented system performs comparably well with respect to a rule-based system, and with an extended training set it is expected that the accuracy of the system will improve, offering greater coverage than rule-based methods. Correspondence: