Vectors Similarity - Spacy

Similarity is determined by comparing word vectors or "word embeddings", multi-dimensional meaning representations of a word. Word vectors can be generated using an algorithm like word2vec and usually look like this:

python

array([2.02280000e-01,  -7.66180009e-02,   3.70319992e-01,
       3.28450017e-02,  -4.19569999e-01,   7.20689967e-02,
      -3.74760002e-01,   5.74599989e-02,  -1.24009997e-02,
       5.29489994e-01,  -5.23800015e-01,  -1.97710007e-01,
      -3.41470003e-01,   5.33169985e-01,  -2.53309999e-02,
       1.73800007e-01,   1.67720005e-01,   8.39839995e-01,
       5.51070012e-02,   1.05470002e-01,   3.78719985e-01,
       2.42750004e-01,   1.47449998e-02,   5.59509993e-01,
       1.25210002e-01,  -6.75960004e-01,   3.58420014e-01,
       # ... and so on ...
       3.66849989e-01,   2.52470002e-03,  -6.40089989e-01,
      -2.97650009e-01,   7.89430022e-01,   3.31680000e-01,
      -1.19659996e+00,  -4.71559986e-02,   5.31750023e-01], dtype=float32)

To make them compact and fast, spaCy's small pipeline packages (all packages that end in sm) don't ship with word vectors, and only include context-sensitive tensors. This means you can still use the similarity() methods to compare documents, spans and tokens – but the result won't be as good, and individual tokens won't have any vectors assigned. So in order to use real word vectors, you need to download a larger pipeline package:

diff

- python -m spacy download en_core_web_sm
+ python -m spacy download en_core_web_lg

</Infobox>

Pipeline packages that come with built-in word vectors make them available as the Token.vector attribute. Doc.vector and Span.vector will default to an average of their token vectors. You can also check if a token has a vector assigned, and get the L2 norm, which can be used to normalize vectors.

python

import spacy

nlp = spacy.load("en_core_web_md")
tokens = nlp("dog cat banana afskfsd")

for token in tokens:
    print(token.text, token.has_vector, token.vector_norm, token.is_oov)

Text: The original token text.

has vector: Does the token have a vector representation?

Vector norm: The L2 norm of the token's vector (the square root of the sum of the values squared)

OOV: Out-of-vocabulary

The words "dog", "cat" and "banana" are all pretty common in English, so they're part of the pipeline's vocabulary, and come with a vector. The word "afskfsd" on the other hand is a lot less common and out-of-vocabulary – so its vector representation consists of 300 dimensions of 0, which means it's practically nonexistent. If your application will benefit from a large vocabulary with more vectors, you should consider using one of the larger pipeline packages or loading in a full vector package, for example, en_core_web_lg, which includes 685k unique vectors.

spaCy is able to compare two objects, and make a prediction of how similar they are. Predicting similarity is useful for building recommendation systems or flagging duplicates. For example, you can suggest a user content that's similar to what they're currently looking at, or label a support ticket as a duplicate if it's very similar to an already existing one.

Each Doc, Span, Token and Lexeme comes with a .similarity method that lets you compare it with another object, and determine the similarity. Of course similarity is always subjective – whether two words, spans or documents are similar really depends on how you're looking at it. spaCy's similarity implementation usually assumes a pretty general-purpose definition of similarity.

📝 Things to try

Compare two different tokens and try to find the two most dissimilar tokens in the texts with the lowest similarity score (according to the vectors).

Compare the similarity of two Lexeme objects, entries in the vocabulary. You can get a lexeme via the .lex attribute of a token. You should see that the similarity results are identical to the token similarity.

python

import spacy

nlp = spacy.load("en_core_web_md")  # make sure to use larger package!
doc1 = nlp("I like salty fries and hamburgers.")
doc2 = nlp("Fast food tastes very good.")

# Similarity of two documents
print(doc1, "<->", doc2, doc1.similarity(doc2))
# Similarity of tokens and spans
french_fries = doc1[2:4]
burgers = doc1[5]
print(french_fries, "<->", burgers, french_fries.similarity(burgers))

What to expect from similarity results {id="similarity-expectations"}

Computing similarity scores can be helpful in many situations, but it's also important to maintain realistic expectations about what information it can provide. Words can be related to each other in many ways, so a single "similarity" score will always be a mix of different signals, and vectors trained on different data can produce very different results that may not be useful for your purpose. Here are some important considerations to keep in mind:

There's no objective definition of similarity. Whether "I like burgers" and "I like pasta" is similar depends on your application. Both talk about food preferences, which makes them very similar – but if you're analyzing mentions of food, those sentences are pretty dissimilar, because they talk about very different foods.
The similarity of Doc and Span objects defaults to the average of the token vectors. This means that the vector for "fast food" is the average of the vectors for "fast" and "food", which isn't necessarily representative of the phrase "fast food".
Vector averaging means that the vector of multiple tokens is insensitive to the order of the words. Two documents expressing the same meaning with dissimilar wording will return a lower similarity score than two documents that happen to contain the same words while expressing different meanings.

sense2vec is a library developed by us that builds on top of spaCy and lets you train and query more interesting and detailed word vectors. It combines noun phrases like "fast food" or "fair game" and includes the part-of-speech tags and entity labels. The library also includes annotation recipes for our annotation tool Prodigy that let you evaluate vectors and create terminology lists. For more details, check out our blog post. To explore the semantic similarities across all Reddit comments of 2015 and 2019, see the interactive demo.

</Infobox>