In the article the compression doesn’t make sense

If you are only sending one word, and the recipient already needs to know the word, then you only need 1 bit, essentially just signaling that you are saying that specific word

If you want a richer vocabulary, you could create an index of about 300k words (from the English dictionary), shared between the parties

Then to send any word you only need to send one number, and in binary it would have between 1 and at most 19 bits, for any word in the index (2^19 is around 500k)

That’s without even sorting the index by frequency of appearance/usage

27 bits for just one word seems wasteful