Interactive Emotion Graph for Human-AI Collaborative Text Manipulation

Problem Statement

In this thesis, we introduce the concept of an interactive, coupled emotion visualization to help writers understand and refine the emotional trajectory of their narratives. By making the results of sentiment analysis visible and interactive in an emotion graph that maps the intensity of primary emotions across a text, such a tool can help authors evaluate and improve the coherence and impact of their storytelling. Writers can use it to maintain a desired emotional pacing, to experiment with different emotional angles, or to identify and change abrupt or flat emotional shifts. We use large language models (LLMs) to enable dynamic bidirectional interaction, where any manipulation in the emotion graph will also be instantly reflected back in the text, allowing for real-time adjustments during the creative process.

Tasks

• Perform a literature review. Identify key challenges and gaps in current approaches to text editors with visualizations
• Implement an interactive website for the users
• Conduct a user study
• Do a statistical evaluation
• Write a thesis and present your findings in the Disputationsseminar
• (Optional:) co-write a research paper