🌿 Lines of Longing
“The machine does not write the poem.
But it may one day understand why we do.”
Lines of Longing is an AI project focused on emotion detection in poetry. It goes beyond binary sentiment and explores the subtle hues of feeling — yearning, surrender, nostalgia, awe — within poetic language.
This project is about giving AI the tools to recognize depth, not just data.
💡 What It Does
- Uses NLP techniques and transformer models (BERT, DistilBERT)
- Fine-tuned on emotion-labeled poetic datasets (or adapted literary datasets)
- Classifies poetry into nuanced emotional categories
- Includes visualizations to show emotion curves across verses
🌸 Goals
- Train a model to recognize emotional layers in short poetic texts
- Explore explainable AI to show why a certain emotion was predicted
- Contribute to emotionally intelligent systems that don’t flatten feeling
🧠 Model Stack
- DistilBERT (or BERT) from HuggingFace Transformers
- Scikit-learn for preprocessing
- SHAP or LIME for explainability
📊 Possible Dataset Sources
- GoEmotions (adapted)
- Custom-curated poetry datasets from Project Gutenberg
- Poetry Foundation (manual annotation)
🔮 Future Plans
- Add emotion arc visualization for full-length poems
- Build a lightweight API to analyze uploaded poems
- Compare classical vs. contemporary poetry emotion signatures
🕯 Why It Matters
Poetry has always been a container for unspeakable emotion.
This project asks: Can machines learn to respect that silence between the lines?