Announcing a new demo and tool for interactive visualization of word
embeddings.  It's free and runs in the browser.  You can access it here:
https://www.cs.cmu.edu/~dst/WordEmbeddingDemo

The tutorial is here:
https://www.cs.cmu.edu/~dst/WordEmbeddingDemo/tutorial.html

Our EAAI paper about the demo is here:
https://www.cs.cmu.edu/~dst/WordEmbeddingDemo/EAAI-2022-Word-Embedding.pdf

TEACHING STRATEGY:
A good way to teach students about word embeddings is to first have them
play Google's Semantris game (https://research.google.com/semantris/)
in Blocks mode.  Then ask how Semantris knows that two words are related.
The answer is word embeddings.  See the tutorial linked above to find
out what these are.

BACKGROUND: There are two major paradigms for doing AI.  The classic
symbolic paradigm uses symbols, rules, search trees, knowledge graphs,
etc.  The numerical paradigm, exemplified by neural networks, uses high
dimensional vector spaces and nonlinear functions.  Word embeddings are
an excellent way to introduce elementary and middle school students to
the numerical paradigm for AI without getting into the technical details
of deep neural nets.

-- Dave Touretzky