markov chain text generator trained on a few hundred thousand lines of poetry from project gutenberg—I decided this output worked best as short haiku-like poems

(people will try to make sense of anything as long as it looks vaguely like a haiku)

finding random rhyming couplets in project gutenberg poetry books

I love Charles Theonia's poetry and had the pleasure of hearing them read some of these poems in their early stages when we were on book tour together a few years ago—this is such a beautiful and appropriate way to publish them

debugging my project gutenberg scraper unintentionally produced an emphatic rallying cry

things about python I will never support or understand

(I've probably tooted about this before but whatever)

there's this new genre of spam that keeps getting through the filter and it seems to be taunting me. NA NA NA NA NA

I took that combined phonetic/semantic similarity vector space I was playing around with and tSNEd it, this is the result of just picking a random line segment in the 2d tsne space and finding the word closest to evenly-spaced points on the line, it's weirdly unnerving

still had this open in textedit, an artifact from when I was making an animation for a presentation about predictive typing

I made a clone of ios quicktype that you can train on arbitrary texts... here's an example trained on pride and prejudice

realized you could make a sort of markov chain text generator using concatenated word vectors instead of the tokens themselves, which has the benefit of being able to cope pretty well with out-of-vocab strings. anyway, here's word-vector-markov Jane Austen elaborating on what the Internet is

🎵 people... who boost posts to reach more people... are the luckiest people in the world 🎵

wondering how many other python projects have gh issues that look like this right now (latest version of python made "async" a reserved word) (this is from

also Hayes proposed using the arithmetic on the n-gram vectors in the model for doing a kind of rudimentary "style transfer," which is definitely one of the the earliest examples of this kind of proposal I've seen (though he doesn't actually get it to work except in the simple "add two models together" example)

I combined the 50d pre-trained GloVe vectors for semantic similarity with my phonetic similarity vectors (just by concatenating them to make 100d vectors) so I could find words that are similar both in meaning and in sound... here's an example in use, replacing each word in _The Road Not Taken_ with its closest sound/meaning equivalent according to this method (original on left)

so, for example, you can easily make a text input field that automatically suggests pokemon names as you type

prototyping a generic version of my gutenberg poetry autocomplete site (gutenberg-poetry.decontextuali) for a workshop... this version uses the SortedSet from and you can put whatever text you want into it. (right now it's showing the words that begin with the given prefix, sorted in reverse order of log probability)

the blossoms on my cornflower plant have very intense color for like two days and then they almost instantaneously lose their color and wither... managed to capture one (on the left) just as the color was fading

while working on a friendly wrapper for a nearest-neighbors library today, I had the occasion to play around with the xkcd color corpus a bit and made these constrained color word poems

I made a very simple corpus-driven chatbot that replies to you with the line of text from the corpus that comes after/is in response to the line of text most similar to what you just typed. here's a sample interaction where the database is built from the Cornell Movie Dialogs corpus... (my typing in green, bot response in blue; I typed the first turn and it alternates after that)

Show more

Follow friends and discover new ones. Publish anything you want: links, pictures, text, video. This server is run by the main developers of the Mastodon project. Everyone is welcome as long as you follow our code of conduct!