Sentence2Vec

This repository shows how to use Word2Vec to build a Sentence2Vec model.

How it works

Word2Vec can help to find other words with similar semantic meaning. However, Word2Vec can only take 1 word each time, while a sentence consists of multiple words. To solve this, I write the Sentence2Vec, which is actually a wrapper to Word2Vec. To obtain the vector of a sentence, I simply get the averaged vector sum of each word in the sentence. The similarity score of two sentences can be calculated by the cosine similarity of their result vectors.

image source: https://medium.com/@premrajnarkhede/sentence2vec-evaluation-of-popular-theories-part-i-simple-average-of-word-vectors-3399f1183afe

Here, I have used job titles as sentences for example data.

Getting Started

# install pipenv if you don't have it
pip install pipenv

# install all dependency packages
pipenv install

# to train the model file
pipenv run python train.py

# to test using the Sentence2Vec model
pipenv run python test.py

# turn job title to vector
# print(model.get_vector('Uber Driver Partner'))
# [ 0.21333599 -0.53108028  1.37453529  2.03607658 -1.72767097  0.46092097
#   0.81509189 -2.24843609 -1.38467287  0.3631835   0.21041276 -1.43147987
#   2.07694975 -2.04546102 -0.48704222  1.30528843  0.20731724  1.93968257
#   0.28302325 -1.71109595 -1.37670168  0.25310072 -1.14791857  0.52478451
#  -0.39728031  2.46906954 -0.1344102  -1.47068134 -0.51386118 -3.08029145
#   0.91159476  0.57291993  1.16054738 -0.53711121 -0.05862717 -0.50246223
#  -0.59518169  0.67068764  0.3773455   2.28154169  1.85961113  0.45566744
#  -1.93983853 -0.81725128  0.50716382  1.5355774  -0.3168966   0.7430951
#  -0.81434408 -0.31304452 -0.03718723  0.61110803 -0.57048208  0.27710366
#  -0.07769835 -0.30162389  0.33499967  0.20927837  0.96208079 -1.02033632
#   1.35363784 -0.8118807   1.77009585 -1.44495682 -2.28812462  0.72161403
#   0.84442989  0.87933841 -0.16113991 -2.33612454  1.17596538  2.86228576
#  -0.25639118  1.30844083 -1.04891106  0.33489262  0.40968706 -0.91616195
#  -0.49571678  1.66697404 -0.30117604  0.53850619 -1.60444642 -0.56396668
#  -0.6034843   1.0579209  -0.27829613 -1.11163335  0.20202117 -0.63776878
#   0.36250977  0.74587251 -0.51649491 -2.0207209   0.33673185 -0.04135378
#  -0.5516142  -1.25613269  2.31798068 -0.68068302]

# not similar job
# print(model.similarity('Uber Driver Partner',
#                        'Carpenter/ Modular  building installer'))
# 0.252182726939546

# a bit similar job
# print(model.similarity('Temporary Barista 30 hours per week',
#                        'Waitress / Waiter Part-Timer'))
# 0.3569998251067342

# similar job
# print(model.similarity('Sandwich maker / All rounder',
#                        'Cafe all rounder and Sandwich Hand'))
# 0.8558831986704775

API

constructor(model_file)

Provide a model file when create the model

load(model_file)

Load from a model file

get_vector(sentence)

Get the vector representation of the sentence

similarity(x, y)

Get the similarity score (range from 0 to 1) for 2 sentences x and y