Automatic Interpretation of Visual Metaphor For Creative Ideation
This is a visual metaphor and the main goal is to generate an unlimited number of ideas to create other visual metaphors with help of machine learning and AI.
With the help of Machine Learning tools and algorithms, a Computational Creativity model is developed to generate creative Visual Metaphor ideas. In this paper, with the study Combination Creativity, we tried to interpret visual metaphor to train a model which can generate innovative ideas for visual metaphor. The model generates an idea for a given context based on Case-based Reasoning. By processing similar ideas with the help of Divergent Thinking, the model tries to find new connections between different concepts and generates some concept candidates. Conceptual networks and similar words and concepts based on their’s embedding vectors are part of the divergent thinking module. Then a generative model generates creative ideas for each given candidate. A rating survey and a Turing test are provided to evaluate the generated ideas. The Turing test and a rating survey demonstrate that bot-generated ideas can compete with human-generated ideas.
More details on model architecture, dataset, and idea generation is in the report file named visual metaphor report.pdf . This project is part of a thesis submitted in partial fulfillment of the requirement for the degree of master of science (Doctor of Philosophy) in Artificial Intelligence