Predicting Human Metaphor Paraphrase Judgments with Deep Neural Networks
View/ Open
Number of pieces
45 - 55 (10)
Publisher
Publisher URL
Location
Journal
Proceedings of the Workshop on Figurative Language Processing 2018
North American Association of Computational Linguistics Workshop on Figurative Language 2018
Metadata
Show full item recordAbstract
We propose a new annotated corpus for metaphor interpretation by paraphrase, and a novel DNN model for performing this task. Our corpus consists of 200 sets of 5 sen- tences, with each set containing one reference metaphorical sentence, and four ranked candi- date paraphrases. Our model is trained for a binary classification of paraphrase candidates, and then used to predict graded paraphrase ac- ceptability. It reaches an encouraging 75% ac- curacy on the binary classification task, and high Pearson (.75) and Spearman (.68) correla- tions on the gradient judgment prediction task.