从文本文件加载词向量 - GENSIM PYTHON

Question

Hello i have a txt file in this form, in the first column is the word and in the second its vector.

word 0.256 0.2659 0.326595
word1 0.528 0.6589 0.62326 ...

i am trying to load this as keyedvectors because I want to compute after the cosine similarity between the words and find the most similar words but I always get an error.

Answer 1

I'm guessing the actual format includes line breaks, like:

word 0.256 0.2659 0.326595
word1 0.528 0.6589 0.62326

That's more-or-less the format common for GLoVe-trained vectors, & very similar to the text format used by Google's original word2vec.c code - which adds a 1st line with a count of vectors & their dimensionality.

(If your vectors came from one of those tools, or a public place, & there are more hints as to their format from the filename or origin, that would have been helpful to note in your question.)

If I'm guessing your true format correctly, then Gensim's KeyedVectors class can load the GLoVe format via the .load_word2vec_format() method, with the no_header=True optional parameter:

vecs = KeyedVectors.load_word2vec_format(filename, binary=False, no_header=True)

See the docs for more options:

https://radimrehurek.com/gensim/models/keyedvectors.html#gensim.models.keyedvectors.KeyedVectors.load_word2vec_format