How frozsgy/semantle-tr works
Project type and purpose
• This is a full-stack web application (“Semantle Türkçe”) that serves a word-guessing game backed by semantic similarity.
• The README and file tree show a Python backend, a static/templated frontend, a SQLite data store initialized from Word2Vec vectors and hint-generation scripts, and a Docker-based deployment.
File structure
Top-level:
• Dockerfile, docker-compose.yml, .dockerignore – container and orchestration setup.
• semantle.py – the main Python web service (likely a Flask or similar WSGI app).
• requirements.txt – Python dependencies.
• conf.d/semantle-tr.conf – webserver (e.g. Nginx) configuration.
• dump-vecs.py, dump-hints.py, store-hints.py – data-preparation scripts that train/load vectors, generate hints, and populate the SQLite DB.
• word2vec/ – model-training code (train.py, train.sh for Wikipedia corpus).
• static/ & templates/ – frontend UI delivered via HTML, CSS, JavaScript and Jinja2 templates.
README for additional insights
• Describes a pipeline: train Word2Vec → dump vectors → generate hints → import hints → run Docker-Compose to serve the app.
• Dependencies are purely Python and a local SQLite DB; no external SaaS beyond the open Wikipedia dump.
Elements
• Model → Dump Scripts → DB
• Browser → Nginx → Backend → DB → Backend → Browser
• Blue for data pipelines
• Green for runtime services
• Orange for storage
• Purple for frontend/UI