Word Embeddings Learned with Different Context Types

The word embeddings used in our paper below are available for download here:

[25 dimensions]
[50 dimensions]
[100 dimensions]
[200 dimensions]
[300 dimensions]
[600 dimensions]

These embeddings were learned with a 10-billion-word corpus, comprising English Wikipedia 2015, UMBC (web) and Gigaword (news). The downloads include embeddings learned with context window sizes of 1,5 and 10, as well as dependency-based and substitute-based contexts.

Oren Melamud, David McClosky, Siddharth Patwardhan, Mohit Bansal. The Role of Context Types and Dimensionality in Learning Word Embeddings. In Proceedings of NAACL, 2016.