From the post: "This was made with the help of a computer program [word2vec] that tries to express the meaning of any word by an adjective and a noun pair. Phrases like 'abstract astronomy' for 'space' and 'aquatic archipelagos' for 'islands' were generated by the program."
The author, Doug Summers Stay, explained the method to me a few years ago:
---
The idea is very simple.
You have a target word and a library of possible adjectives and nouns that start with the letter a, for example.
You add an extra element to each vector, so they are now 301 dimensional vectors. For adjectives, you set this to 10, for nouns to -10. In the target word, you set this element to zero. (You might have to play with the values 10 and -10 a little: they should be weighted high enough compared to the other elements that getting them right is a necessity rather than just a nice thing to have.)
A sparse decomposition function such as LASSO takes in a (1 x 301 target vector) and a library of n possible vectors stacked up as a (n x 301 matrix) and outputs a (n x 1 vector) which shows how each of the n possible vectors in the library should be weighted in a sum. You tune the sparsity parameter so that it is looking for exactly two non-zero weights. So it needs to find two vectors that add up as closely as possible to the target vector, and one of them must be an adjective and one must be a noun to make it work out right.
You perform the sparse decomposition and the non-zero weighted elements are the adj. and noun you want. You can run it again with those words removed from the dictionary if you want other choices.
This is just something I came up with and tested. It's never been published anywhere that I'm aware of. You can easily imagine variations to find any number of terms with particular properties or relationships to each other.
Kind of reminds me of the book Thing Explainer my Randal Monroe where he uses only the 1,000 most common English words. (Of course the drawings do a lot of the heavy lifting as far as explaining).
"This end should be pointed toward the ground if you want to go to space. If it starts pointing toward space, you are having a bad problem and you will not go to space today."
I was inspired last year to do an equivalent with B: https://godexperiment.org/beginnings-an-alliterative-rewrite... ("Blessed Being began by building blue bowl, bottom base..."). (That's not my own site where it's posted, it's my friend Jeremiah's site.)
I was also inspired by this S version: https://calvinballing.github.io/saga/ ("So started saga: Supreme sentience shaped skies, secular sphere...").
Reminds me of Stanisław Barańczak’s paraphrase of Hamlet’s soliloquy where every word on the first line starts with A, on the second with B, etc. Unfortunately, I’m afraid you need to understand Polish fluently to appreciate this:
There's a surprising amount of flexibility here for new translations. For example, I prefer "Autochthon" as a translation for "God", rather than "ancient archangels" (although "Almighty" is the perfect word for later on). A while ago I decided I preferred the following as verses 1 and 2:
1. An advent: Autochthon architects Area, above and abase.
2. Actions articles apparatus, all amorphous and absent; And all asleep across aquatic anarchy. And Autochthon advanced across area.
"Abase" makes sense to me by analogy with its cognates "abaixo" and "abajo" in Portuguese and Spanish, or French "à bas", but it's not clear to me that it's ever literally had the meaning of "down(ward)", "below", or "beneath" in English.
Maybe "abased", which apparently has such a literal geometric meaning at least in heraldry?
Hmm, yes, I think you're right. There's the more prosaic "… above and all around", which gets at something slightly different from "heaven and earth". Douglas Hofstadter was right all along about how hard translation is!
Well, to the people to whom the Hebrew Bible belongs, there's a reason Genesis (Bereshit) starts with the second letter of the aleph-bet ("Bet") and not the first ("Aleph")
---
The idea is very simple.
You have a target word and a library of possible adjectives and nouns that start with the letter a, for example.
You add an extra element to each vector, so they are now 301 dimensional vectors. For adjectives, you set this to 10, for nouns to -10. In the target word, you set this element to zero. (You might have to play with the values 10 and -10 a little: they should be weighted high enough compared to the other elements that getting them right is a necessity rather than just a nice thing to have.)
A sparse decomposition function such as LASSO takes in a (1 x 301 target vector) and a library of n possible vectors stacked up as a (n x 301 matrix) and outputs a (n x 1 vector) which shows how each of the n possible vectors in the library should be weighted in a sum. You tune the sparsity parameter so that it is looking for exactly two non-zero weights. So it needs to find two vectors that add up as closely as possible to the target vector, and one of them must be an adjective and one must be a noun to make it work out right.
You perform the sparse decomposition and the non-zero weighted elements are the adj. and noun you want. You can run it again with those words removed from the dictionary if you want other choices.
This is just something I came up with and tested. It's never been published anywhere that I'm aware of. You can easily imagine variations to find any number of terms with particular properties or relationships to each other.