SpaCy:如何加载Google新闻word2vec向量？

Question

我尝试了几种加载google news word2vec向量的方法(https://code.google.com/archive/p/word2vec/):

en_nlp = spacy.load('en',vector=False)
en_nlp.vocab.load_vectors_from_bin_loc('GoogleNews-vectors-negative300.bin')

以上给出:

MemoryError: Error assigning 18446744072820359357 bytes

我也试过.gz打包向量; 或者使用gensim将它们加载并保存为新格式:

from gensim.models.word2vec import Word2Vec
model = Word2Vec.load_word2vec_format('GoogleNews-vectors-negative300.bin', binary=True)
model.save_word2vec_format('googlenews2.txt')

然后,该文件包含每行上的单词及其单词向量.我试着加载它们:

en_nlp.vocab.load_vectors('googlenews2.txt')

但它返回"0".

这样做的正确方法是什么？

更新:

我可以将自己创建的文件加载到spacy中.我在每一行使用带有"string 0.0 0.0 ...."的test.txt文件.然后使用.bzip2将此txt压缩到test.txt.bz2.然后我创建一个spacy兼容的二进制文件:

spacy.vocab.write_binary_vectors('test.txt.bz2', 'test.bin')

我可以加载到spacy:

nlp.vocab.load_vectors_from_bin_loc('test.bin')

这有效!但是,当我为googlenews2.txt执行相同的过程时,我收到以下错误:

lib/python3.6/site-packages/spacy/cfile.pyx in spacy.cfile.CFile.read_into (spacy/cfile.cpp:1279)()

OSError: 

Answer 1

对于spacy 1.x,将Google新闻向量加载到gensim并转换为新格式(.txt中的每一行包含一个向量:string,vec):

from gensim.models.word2vec import Word2Vec
from gensim.models import KeyedVectors
model = KeyedVectors.load_word2vec_format('GoogleNews-vectors-negative300.bin', binary=True)
model.wv.save_word2vec_format('googlenews.txt')

删除.txt的第一行:

tail -n +2 googlenews.txt > googlenews.new && mv -f googlenews.new googlenews.txt

将txt压缩为.bz2:

bzip2 googlenews.txt

创建SpaCy兼容的二进制文件:

spacy.vocab.write_binary_vectors('googlenews.txt.bz2','googlenews.bin')

将googlenews.bin移至python环境的/lib/python/site-packages/spacy/data/en_google-1.0.0/vocab/googlenews.bin.

然后加载wordvectors:

import spacy
nlp = spacy.load('en',vectors='en_google')

或者在以后加载它们:

nlp.vocab.load_vectors_from_bin_loc('googlenews.bin')

Answer 2

我知道已经回答了这个问题，但是我将提供一个更简单的解决方案。此解决方案会将google新闻载体加载到空白spacy nlp对象中。

import gensim
import spacy

# Path to google news vectors
google_news_path = "path\to\google\news\\GoogleNews-vectors-negative300.bin.gz"

# Load google news vecs in gensim
model = gensim.models.KeyedVectors.load_word2vec_format(gn_path, binary=True)

# Init blank english spacy nlp object
nlp = spacy.blank('en')

# Loop through range of all indexes, get words associated with each index.
# The words in the keys list will correspond to the order of the google embed matrix
keys = []
for idx in range(3000000):
    keys.append(model.index2word[idx])

# Set the vectors for our nlp object to the google news vectors
nlp.vocab.vectors = spacy.vocab.Vectors(data=model.syn0, keys=keys)

>>> nlp.vocab.vectors.shape
(3000000, 300)