The Use of Granularity in Rhetorical Relation Prediction

We present the results of several machine learning tasks designed to predict rhetorical relations that hold between clauses in discourse. We demonstrate that organizing rhetorical relations into different granularity categories (based on relative degree of detail) increases average prediction accuracy from 58% to 70%. Accuracy further increases to 80% with the inclusion of clause types. These results, which are competitive with existing systems, hold across several modes of written discourse and suggest that features of information structure are an important consideration in the machine learnability of discourse.