Discriminative learning of selectional preference from unlabeled text

Abstract

We present a discriminative method for learning selectional preferences from unlabeled text. Positive examples are taken from observed predicate-argument pairs, while negatives are constructed from unobserved combinations. We train a Support Vector Machine classifier to distinguish the positive from the negative instances. We show how to partition the examples for efficient training with 57 thousand features and 6.5 million training instances. The model outperforms other recent approaches, achieving excellent correlation with human plausibility judgments. Compared to Mutual Information, it identifies 66% more verb-object pairs in unseen text, and resolves 37% more pronouns correctly in a pronoun resolution experiment.

Inducing Chinese Selectional Preference Based on HowNet

“…Erk [13] generalized seen arguments to unseen arguments with a similarity based method. Bergsma et al [14] formulated SP acquisition as a classification problem and utilized Support Vector Machine (SVM) as the classifier.…”

Perceptually Grounded Selectional Preferences

“…The systems compute similarities between distributional representations of arguments in a vector space. Bergsma et al (2008) trained an SVM classifier to discriminate between felicitous and infelicitous verb-argument pairs. Their training data consisted of observed verb-argument pairs (positive examples) with unobserved, randomly-generated ones (negative examples).…”

A Neural Network Approach to Selectional Preference Acquisition

“…We equally compare to their model for evaluation purposes. Bergsma et al (2008) present a discriminative approach to selectional preference acquisition. Positive examples are taken from observed predicate-argument pairs, while negative examples are constructed from unobserved combinations.…”