Allison Parrish @aparrish@mastodon.social

since the dawn of the written word people have made amulets and talismans from holy texts and in this sense a pinned toot is like a horseshoe you hang over a barn door

this is the Cornell Movie Dialogs corpus btw http://www.cs.cornell.edu/~cristian//Cornell_Movie-Dialogs_Corpus.html

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)

why are all of the entries about math on wikipedia so bad 😐😐😐😐😐😐😐😐😐😐😐😐 it's like listen I know I messed up by not dedicating my life to math but I really need to understand this one concept can you just cool it for a second and help me out

and by "good old fashioned" I mean "like literally state of the art 36 months ago" as is the case for almost everything machine learning-related at the moment

trying to use google's "universal sentence encoder" but it's trying to use 32 gigs of RAM (out of my laptop's 8 gigs) to find embeddings for "this is a test" so maybe it's back to good old fashioned averaging-the-word-vectors for me https://www.tensorflow.org/hub/modules/google/universal-sentence-encoder/2

reading/reading about Emily Short's _Parrigues Tarot_ and I'm completely astounded, it's so good https://emshort.blog/2018/06/26/parrigues-tarot-draft/

"[S]tenciling letters itself is self-evidently neither writing nor typography, [but it] often reaches in these directions"—super fascinating blog post about 20th century stencil alphabets https://www.fontshop.com/content/stencil-alphabets (via https://mailchi.mp/theleagueofmoveabletype/weekly-typographic-2900097)

(note I have never seen Westworld so I can't discuss whether or not this is "one of the best things" about its second season. I just liked these sentences)

"But one of the best things about Westworld's second season is it recognizes that the question of sex with robots is boring. Because it is. It sound provocative, but there's not much you can extrapolate from it once you think about it for more than five minutes." https://tinyletter.com/jomc/letters/krav-maga-with-androids

realizing now how much I'm dating myself by posting this but whatever

koko the gorilla's death reminded me of what I think is the galvanizing moment as a kid when I realized I didn't have to—and probably should not—be a conservative: when I saw this episode of Rush Limbaugh's TV show https://www.youtube.com/watch?v=YTOb_KkHkm4 I remember thinking this segment was so mean-spirited, so closed off to everything other than smug self-satisfied ignorance. it embodied everything I didn't want to be. so I guess sorta thanks rush for making it so easy to figure this out

"[T]here have been fewer than 200 cases of false family claims this year — a fraction of 1 percent of the total number of families apprehended at the border — and there is no indication that any of those cases involved MS-13. Of the hundreds of thousands of unaccompanied minors that have come to the U.S. since 2012, Border Patrol says only 56 were suspected of MS-13 ties." https://www.propublica.org/article/ms-13-immigration-facts-what-trump-administration-gets-wrong

random forest is doing wayyyyy better than any of the other classifiers on this data set but it also takes forever to train and adjusting any of the hyperparameters from the sklearn defaults seems to make it worse somehow? 😿

tooting some tabs as I close 'em https://en.wikipedia.org/wiki/Cot%E2%80%93caught_merger https://commons.wikimedia.org/wiki/File:Plan_(top),_and_three-quarter_view_(below),_of_a_type_case,_Wellcome_V0023764.jpg https://books.google.com/books?id=7JfL_0ce6IQC http://www.gtoal.com/wordgames/index.html#foreignscrab https://web.archive.org/web/20120222025924/http://home.teleport.com/~stevena/scrabble/alpha/un_hist.html https://boardgamegeek.com/geeklist/68299/anagram-family-games-non-interlocking-non-juxtapos/ http://thegibsonhousemuseum.blogspot.com/2016/02/logomachy-look-at-nineteenth-century.html https://daily.jstor.org/codifying-what-counts-as-a-word-in-scrabble/ (mostly leftover research from yesterday's talk)

thread on the birdsite about the talk I gave earlier today at #Practice2018 https://twitter.com/metasynthie/status/1010547547419619328

the syllabic nucleus of the vulcan language http://startrekvulcanology.blogspot.com/2010/11/vulcan-language-dice-from-unification.html

loved these poems on tagvverk by Adam Greenberg https://tvverk.wordpress.com/2018/06/21/adam-greenberg/ "Each poem has been assembled using only language from the names of companies found in the Fourtune 500 list from 1989"