Innovation or Imitation? Assessing Novelty in AI-Generated Metaphors through Automated Conceptual Metaphor Analysis with MetaNet, Word Embeddings, and Conceptual Mapping Techniques

The following thesis explores whether metaphor generation can be creative using an automatic metaphor interpretation system and a creativity evaluation mechanism. After presenting the theoretical foundation of metaphors, including Conceptual Metaphor Theory, Frame Semantics, and central works in Metaphor Analysis within Natural Language Processing, the thesis focuses on developing an algorithm to detect conceptual metaphors in a sentence. This is followed by statistical and close reading analyses of the system’s functionality. The second part delves into the concept of creativity in text, particularly in text generation by large language models, and examines how creativity in metaphors can be quantified computationally. Finally, the thesis evaluates AI-generated metaphors using the developed methods and offers a conclusive assessment.