Stable Release including:
VecShare: Library for Sharing Word Vectors
Release v1.2.2
This library provides functionality for rapidly sharing and retrieving word embeddings over the internet. Additional information on the VecShare framework can be found at: https://bit.ly/VecShare (Accepted at EMNLP 2017).
Download at pip install vecshare
Supported Functions
The VecShare Python library currently supports:
check: See available embeddingsformat: Autoformat a header to upload an embedding to the data store or Compress an embeddingupdate: Update an existing embedding or its metadataquery: Look up word vectors from a specific embeddingextract: Download word vectors for only the vocabulary of a specific corpusdownload: Download an entire shared embedding
Supported Selection Methods
maxtkn: Select embedding trained on most tokenssimscore: Select embedding scoring highest on 9 set similarity taskavgrank: Select embedding with highest avg rank signature score (See https://bit.ly/VecShare)