Bridging the Gap: Condensing Knowledge Graphs for Metaphor Processing by Visualizing Relationships in Figurative and Literal Expressions

Metaphorical expressions, being unique phrases with complex traits like disjointedness, non-compositionality, diversity, and flexible structure, are crucial in natural language tasks like translation, opinion analysis, and summarization. To handle these expressions, an inventive approach is needed. This method, using Design Science Research, relies on identifying patterns in how concepts connect through dependency parsing. It stores this knowledge by organizing it into structured scripts, utilizing Cypher statements to integrate it into a graph database. Neo4j helps visualize various nodes and attributes, allowing the study of their patterns and behaviors in both literal and figurative sentences. The accuracy of this process is assessed using machine learning models like Naive Bayes, k-nearest neighbor, random forest, gradient booster, and support vector machine to classify literal versus metaphoric expressions.