Poetic Ontology Builder

Author: Willinton Triana Cardona
Version: 1.1.0

"Every text carries emotional weight. This tool extracts its meaning, weaves it into an ontology, and illuminates its emotional resonance"

Poetic Ontology Builder transforms any text into a living knowledge graph enriched with emotional embeddings.
It combines Google Gemini 2.5 Flash for knowledge extraction, Rosa (a fine‑tuned BERT model for GoEmotions), and ontology construction (OWL + RDF) into an interactive, beautiful visualization — including a full Enneagram personality profile derived from the text's emotional field.

---

Features

• Knowledge graph extraction:
  Extracts subject‑predicate‑object triplets with contextual qualifiers from any text using Gemini 2.5 Flash.
• Emotional embedding:
  Integrates ROSA — a fine‑tuned BERT model trained on the GoEmotions dataset — to detect the full probability distribution across 28 emotion labels per triplet.
• Precise emotional analytics:
  Computes mean, std dev, max, and total per emotion across all triplets. Outputs a dark‑themed bar chart with error bars and percentage labels.
• Enneagram chart:
  Maps the aggregated emotion distribution onto the 9 Enneagram types and renders a polar radar chart showing the dominant personality archetype of the text.
• Ontology export:
  Generates GraphDB‑ready OWL files for semantic storage and querying.
• Interactive graph visualization:
  Force‑directed, emotion‑color‑coded graphs with poetic tooltips.
• Timestamped outputs:
  Each run produces a clean folder:

  outputs/YYYYMMDD_HHMMSS/
  ├── sublime_graph.owl
  ├── sublime_graph.html
  ├── emotion_stats.json
  ├── emotion_stats.png
  ├── enneagram_stats.json
  └── enneagram_chart.png
  

---

Requirements

• Python 3.10 or higher
• A Gemini API key (Google AI Studio)
• A Hugging Face token with access to willt-dc/Rosa-V1

---

Installation

From PyPI:

pip install poetic-ontology-builder

For local development (editable):

git clone https://github.com/willt08/poetic_ontology_builder.git
cd poetic_ontology_builder
pip install -e .

---

Configuration

Create a .env file in the project root:

GEMINI_API_KEY=your_gemini_api_key
HF_TOKEN=your_huggingface_token

---

Usage

From a text file:

poetic-ontology path/to/textfile.txt

From an inline string:

poetic-ontology "And Socrates said: When the soul returns into itself..."

Open all outputs automatically after generation:

poetic-ontology rilke.txt --open

---

Output

1. Ontology:
   sublime_graph.owl — RDF/OWL representation, GraphDB‑ready.
2. Interactive Graph:
   sublime_graph.html — force‑directed graph with emotion‑colored nodes and tooltips.
3. Emotion Analytics:
   • emotion_stats.json — mean, std dev, max, min, total per emotion + dominant emotion counts per triplet.
   • emotion_stats.png — dark‑themed bar chart of mean probability scores ± std dev.
4. Enneagram Profile:
   • enneagram_stats.json — score and normalized score for each of the 9 types, plus dominant type.
   • enneagram_chart.png — polar radar chart of the text's Enneagram emotional profile.

---

Enneagram Mapping

Each GoEmotions label is mapped to its primary Enneagram type based on core motivational emotion clusters:

Type	Name	Key Emotions
1	Reformer	annoyance, disapproval, disgust
2	Helper	caring, love, gratitude, approval
3	Achiever	admiration, pride, optimism
4	Individualist	sadness, grief, disappointment, remorse, embarrassment
5	Investigator	curiosity, confusion, realization
6	Loyalist	fear, nervousness
7	Enthusiast	joy, excitement, amusement, desire, surprise
8	Challenger	anger
9	Peacemaker	neutral, relief

---

Emotional Palette

Each emotion is mapped to a distinct color based on its polarity and tone:

• Joy / Love / Pride → Warm yellows & pinks
• Grief / Fear / Anger → Deep blues & reds
• Neutral → Gray

---

Roadmap

• BFO/SUMO alignment for deeper ontological grounding
• PyPI release for global access
• Web interface for non‑technical users
• Multi‑modal inputs (image + text)

---

Acknowledgements

• ROSA: Emotional BERT model fine‑tuned on GoEmotions
• Google Gemini 2.5 Flash: Triplet & knowledge extraction
• OWLready2: Ontology management
• PyVis: Interactive graph visualization

---

License

MIT License © 2025 Willinton Triana Cardona