felipe.lavin.blog

On making stuff with code. And making code to do stuff.

Understanding embeddings

tldr: embeddings are numeric representations of text on an n-dimensional space that can be used to calculate semantic proximity.

For more details, check:

  • Embeddings are underrated – a concise theorical and practical approach to embeddings for discovering connections between texts.
  • The Illustrated Word2vec – has lots of diagrams and visualizations for word embeddings, language model training, sliding word windows (skipgrams) and text generation.