embeddix 1.2.0

A general purpose toolkit for word embeddings

embeddix

A small toolkit for processing word embeddings with numpy. You can use embeddix to convert .txt embeddings (such as glove files) to numpy and vice-versa.

Install

pip install embeddix

or, after a git clone:

python3 setup.py install

Use

Extract vocabulary from a txt embeddings file

embeddix extract --embeddings /absolute/path/to/embeddings.txt

Convert from txt to numpy

embeddix convert --to numpy --embeddings /absolute/path/to/embeddings.txt

Convert from numpy to txt

embeddix convert --to txt --embeddings /absolute/path/to/embeddings.npy

Align vocabularies across models

embeddix reduce --embeddings /absolute/path/to/embeddings/directory/

Evaluate DSM on intrinsic tasks

Evaluate on lexical similarity (men, simlex, simverb) or concept categorization (essli, ap, battig)

embeddix evaluate \
--embeddings /absolute/path/to/embeddings.npy \
--vocab /absolute/path/to/embeddings.vocab \
--dataset instrinsic_task_dataset_name