Published Paper Details:

A metaphor is a figure of speech used to express a specific meaning or idea by drawing comparisons between two entities or ideas that are not necessarily related to one another. Without using the terms "like" or "as," metaphors use implicit comparisons. Instead, they substitute a word or phrase that doesn't directly relate to the topic at hand but instead makes an analogy or comparison. Understanding and processing metaphors is a difficult topic in natural language processing. Consequently, the current need for metaphor classification involves an innovative approach. In order to conduct tasks like entity linking and relation extraction, we employ a method known as Knowledge Graph Embedding to represent the data in a knowledge graph as numerical vectors. When examining metaphors, we notice a structural pattern where the source and target are linked by a relationship that depends on a specific attribute. The target, the attribute, and the source make up the metaphor triple that describes this pattern. We use dependency parsing to extract collocations of concepts and attributes. In our method, we jointly embed the metaphor Knowledge Graph and the Concept-Attribute collocations. After that, the goal of metaphor identification can be seen as representation-enhanced concept pair classification. We also work on prediction of the possible attributes when given two concepts i.e. target and source from knowledge graph. By improving metaphor detection accuracy, this technique will fill a research gap.