Word Vector Analysis

The following example illustrates how to find most similar words, simple word vector maths and how to visualise similar words in a compressed vector space.

import matplotlib.pyplot as plt
plt.close('all') # very important for read the docs to avoid it crashing due to memory

import enlp.understanding.vectors as vts
from enlp.visualisation.word_vectors import similar_words

Download vectors - this can be swapped with loading your own vectors

import gensim.downloader as api

model = api.load("word2vec-google-news-300")  # download the model and return as object ready for use
wvs = model.wv #load the vectors from the model

Out:

/Users/clbi/Documents/eNLP_all_folders/eNLP_fork/eNLP/examples/ex_wordvectoranalysis.py:20: DeprecationWarning: Call to deprecated `wv` (Attribute will be removed in 4.0.0, use self instead).
  wvs = model.wv #load the vectors from the model

Most similar word to happy

print (vts.similar_words(wvs, 'happy', n=5))

Out:

[('glad', 0.7408890128135681), ('pleased', 0.6632171273231506), ('ecstatic', 0.6626912355422974), ('overjoyed', 0.6599286794662476), ('thrilled', 0.6514049768447876)]

Most similar word to zebra

print (vts.similar_words(wvs, 'zebra', n=5))

Out:

[('giraffe', 0.6372909545898438), ('hippo', 0.6137316823005676), ('zebras', 0.5988895893096924), ('hippopotamus', 0.5641686916351318), ('leopard', 0.5635697841644287)]

Vector Maths - Past tense of walk?

# ran - run + walk
pwords = ['ran', 'walk']
nwords = ['run']
print (vts.vector_maths(wvs, pwords=pwords, nwords=nwords))

Out:

[('walked', 0.7423241138458252)]

Vector Maths - Female equivalent of king?

# Woman - Man + King = ???
pwords = ['woman', 'king']
nwords = ['man']
print (vts.vector_maths(wvs, pwords=pwords, nwords=nwords))

Out:

[('queen', 0.7118192911148071)]

Vector Maths - Country of which Edinburgh is the capital?

# Norway - Oslo + Edinburgh
pwords = ['Norway', 'Edinburgh']
nwords = ['Oslo']
print (vts.vector_maths(wvs, pwords=pwords, nwords=nwords))

Out:

[('Scotland', 0.7331377267837524)]

Visualising vectors

fig, (ax1, ax2) = plt.subplots(1, 2, figsize=(14, 7))
similar_words(wvs, 'pizza', ax=ax1)
ax1.set_title('Most similar words to pizza')
similar_words(wvs, 'pasta', ax=ax2)
ax2.set_title('Most similar words to pasta')
plt.tight_layout()