Doc.similarity either outputting 0 or 1 (only when it's the same string)

[JayThibs]

How to reproduce the behaviour

When I run the code below, I get a score of 0.0. I only get a score of 1 if I use "apple" for doc2. Everything else it 0.

!pip install -U pip setuptools wheel --quiet
!pip install -U 'spacy[cuda113]' --quiet
!python -m spacy download en_core_web_trf --quiet
import spacy
nlp = spacy.load("en_core_web_trf")
doc1 = nlp(u"apple")
doc2 = nlp(u"green apple")
doc1.similarity(doc2)

Any ideas what's wrong?

Your Environment

• Operating System: Colab
• Python Version Used: 3.7
• spaCy Version Used: Latest

[polm]

This is expected behavior, because the Transformers pipeline doens't have word vectors and doc.similarity uses word vectors.

Moving to Discussions since this isn't a bug report.