Word2Vec Embeddings Implementation

2020

Architecture diagram of the Word2Vec skip-gram model mapping input words through an embedding layer to context word probabilities

2D t-SNE projection of word vectors trained with standard skip-gram showing semantic clusters

2D t-SNE projection of word vectors trained with negative sampling showing improved semantic relationships

Project information

Project URL: GitHub Repository

About this project

Implementation of the Word2Vec algorithm in PyTorch using the skip-gram architecture with negative sampling. Visualizations compare standard skip-gram word embeddings with negative-sampling embeddings via 2D t-SNE projections to show learned semantic relationships.

Back to Projects