What have we accomplished
1. Performed Data Analysis on 《小說與詩》on various aspects: Word count, Sentiment, Lingustic Complexity, and Author-related information.
2. Performed Data Visualization such as: Histograms, Bar Charts, Box and whiskey diagrams, Correlation Matrices.
3. Implemented a closed-loop image generation workflow to visually represent the core message of each poem.
Findings
We successfully analyzed all 579 poems, with each poem averageing around 100 words. Sentiment analysis revealed that most poems convey extreme emotions—either positive or negative—with a few being neutral.
The generated images, along with their corresponding poems, are presented in a gallery format. We noticed that if the core message of a poem is hard to convey directly due to inappropriateness, the generated image will usually convey the core message symbolically.