shimo-lab/Universal-Geometry-with-ICA
Discovering Universal Geometry in Embeddings with ICA (Published in EMNLP 2023)
This project helps natural language processing (NLP) researchers and computational linguists understand the underlying structure of word embeddings, which are numerical representations of words. By applying a technique called Independent Component Analysis (ICA), it transforms raw word embeddings (from English or multiple languages) into visualizations like heatmaps and scatter plots. The output reveals consistent 'geometric' patterns that indicate how words are related in meaning across different languages.
No commits in the last 6 months.
Use this if you are a researcher analyzing word relationships in single or multiple languages and want to uncover fundamental, universal patterns in how words are represented.
Not ideal if you are looking for a tool to build or apply pre-trained language models for specific tasks like sentiment analysis or machine translation, rather than analyzing their internal structure.
Stars
20
Forks
—
Language
Python
License
—
Category
Last pushed
Jun 17, 2025
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/embeddings/shimo-lab/Universal-Geometry-with-ICA"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Higher-rated alternatives
cosmosgl/graph
GPU-accelerated force graph layout and rendering
Clay-foundation/model
The Clay Foundation Model - An open source AI model and interface for Earth
nomic-ai/nomic
Nomic Developer API SDK
omoindrot/tensorflow-triplet-loss
Implementation of triplet loss in TensorFlow
sashakolpakov/dire-jax
DImensionality REduction in JAX