Supervised Metaphor Detection using Conditional Random Fields

Abstract

In this paper, we propose a novel approach for supervised classification of linguistic metaphors in an open domain text using Conditional Random Fields (CRF). We analyze CRF based classification model for metaphor detection using syntactic, conceptual, affective, and word embeddings based features which are extracted from MRC Psycholinguistic Database (MRCPD) and WordNet-Affect. We use word embeddings given by Huang et al. to capture information such as coherence and analogy between words. To tackle the bottleneck of limited coverage of psychological features in MRCPD, we employ synonymy relations from WordNet ®. A comparison of our approach with previous approaches shows the efficacy of CRF classifier in detecting metaphors. The experiments conducted on VU Amsterdam metaphor corpus provides an accuracy of more than 92% and Fmeasure of approximately 78%. Results shows that inclusion of conceptual features improves the recall by 5% whereas affective features do not have any major impact on metaphor detection in open text.

Proceedings of the Second Workshop on Figurative Language Processing

“…Many approaches have been proposed for automatic detection of metaphors, using features of lexical information Wilks et al, 2013), semantic classes , concreteness , word associations , constructions and frames (Hong, 2016) and systems such as traditional machine learning classifiers (Rai et al, 2016), deep neural networks (Do Dinh and Gurevych, 2016) and sequential models .…”

Detecting Figurative Word Occurrences Using Recurrent Neural Networks

“…In (Rai et al, 2016), a conditional random field (CRF) algorithm is proposed. The approach is based on features from the MRC psycholinguistic dictionary (Wilson and Division, 1997) and WordNetAffect database (a subset of WordNet with emotion annotations).…”

An LSTM-CRF Based Approach to Token-Level Metaphor Detection

“…As for our other baseline, we considered the results from Rai et al (2016), as reported by them. They used conditional random fields (CRF) for detection of metaphors and experimented on each of the genres contained in VUAMC, as well as on the complete dataset.…”

“…For our training and testing purpose, we had the text ids and sentence ids as provided for the shared task, from which we could get the respective sentences from the VUAMC and thus generate the feature vectors for each of their tokens (leaving aside the punctuation marks), as described in Method Precision Recall F 1 -score Do Dinh and Gurevych (2016) 0.5899 0.5355 0.5614 Rai et al (2016) 0.6333 0.5871 0.6093 Bi-LSTM-CRF (Embeddings only for tokens) 0.7036 0.5755 0.6327 Bi-LSTM-CRF (All of the considered features) 0.7283 0.6253 0.6740 Table 2: Results for Feature Selection on the complete VU Amsterdam Metaphor Corpus with Bi-LSTM-CRF.…”