Plotting Semantics
with the Universal Sentence Encoder

measure:

What is this??

Input some language to visualize & compare language as shapes.
You may also enter a word or phrase to "measure" the similarity. This is really interesting to measure different aspects of things, or to discover information about how language is used (for example: we could actually quanitfy racial/political bias)

How does it work??

This page uses the Universal Sentence Encoder to create numerical representations of semantic concepts called embeddings.

We can represent these embeddings in multidimensional space, to understand their relationships to each other and do math on them.

Similarities are calculated using the cosine similarity formula.

github

Plotting Semantics with the Universal Sentence Encoder

Plotting Semantics
with the Universal Sentence Encoder