jonathnmh12/Greek-Semantic-Analysis

Greek Semantic UI

Interactive UI for exploring Greek semantic embeddings from the Bible.

Overview

This application provides a Pygame-based interface for exploring semantic relationships in Biblical Greek text. Select any Greek word to see its closest semantic neighbors based on embedding similarity.

Project Structure

├── test_ui.py         # Main Pygame application
├── models/            # Model files (symlink or copy from training repo)
└── pyproject.toml

Setup

uv sync

Usage

python test_ui.py

Streamlit With Precomputed LDA Insights

Run the one-time artifact build after training or updating LDA outputs:

python build_lda_artifacts.py

This writes runtime-ready files in data/greek-nt-sbert_v2/lda/:

• —verse_topics.csv
• —topic_summary.csv
• —topic_term_weights.csv
• —artifacts_manifest.json

Then launch the Streamlit app:

streamlit run pages/streamlit_app.py

For corpus-wide word retrieval, place your training text at greek_corpus.txt in the project root. The app extracts unique Greek word tokens from that file and uses them as similarity candidates.

How it works:

1. 1.Load a pre-trained Greek embedding model
2. 2.Generate embeddings for your corpus
3. 3.Click on Greek words to see semantically similar terms
4. 4.Cosine similarity determines ranking

Data

Currently loads verse data from John 1:1 for clickable UI text. Similarity candidates come from greek_corpus.txt (if present), plus visible Greek tokens in verse_data.

Models

Place trained models in the models/ directory. Currently configured to load from models/Greek_v2_Expanded.

Next Steps

• —Load from entire Greek corpus for broader similarity matching
• —Add filtering by part of speech
• —Implement search functionality
• —Support batch similarity comparisons