It first pulls Wikipedia’s list of No Such Thing As A Fish episodes, pulls out all the headline facts, and feeds them into my Markov chain implementation. Every few hours, it picks a fact from the real list, and uses the first two words as a seed to generate a sentence. (If it generates one of the original facts verbatim, it rejects it and generates another — although it does sometimes rephrase them.)
The source code is on GitHub, and it generates things like:
- A kangaroo’s tail is used as an apartment block.
- Kaiser Wilhelm II once lost a valuable arms contract for Germany because he was feeling irritable.
- The Aztecs wore necklaces made out of marzipan and got angry when people ate them.
- David Frost used to scare woodpeckers.
- Desperate Dan stopped eating cow pies because of a pork carcass, and Rosemary Einstein who came up with glowing mushrooms.
- Tomorrow will be the same time, they would know who it came from.
Deck The Halls
Deck The Halls turns tweets with the correct meter into an endless and bizarre version of the classic carol:
Stop assuming Santa's gender.— Jesse (@jessetrowe) December 10, 2016
Falalalala, la la, la la! https://t.co/IGII54aVqa— Deck the Halls (@falalala_la) 10 December 2016
What the fuck is wrong with people— т૨אℓℓ รтα૨ (@TrxllStar) December 11, 2016
Falalalala, la la, la la! https://t.co/cd3fvCRmnW— Deck the Halls (@falalala_la) 12 December 2016
Freezer & refrigerator https://t.co/b4EttqF8tl— 1⃣6⃣ (@_teegram) December 11, 2016
Falala, lalala, la la la! https://t.co/7FXCSDtrvC— Deck the Halls (@falalala_la) 11 December 2016
Oral Sex Can Lead to Cancer https://t.co/AIuEg8ZYJW— daiana bridge (@daianabridge) December 13, 2016
Falalalala, la la, la laaaa! https://t.co/CvAmGqG9oR— Deck the Halls (@falalala_la) 13 December 2016
To do this, it listens to the public Twitter feed. This is a tiny, random subset of all tweets. It then strips out @names at the start of tweets, and URLs at the end, and feeds the rest through Carnegie Mellon University’s pronunciation dictionary which is an invaluable resource for automating puns. It ignores everything except the lexical stress pattern, and it ignores the difference between primary and secondary stresses. If what’s left is eight syllables with the stress on numbers one, three, five and seven, the tweet fits into the song.
If it’s been less than 20 minutes since the last tweet, then the matching tweet is quoted, with the appropriate line of “falala”s — that is, “falalala la, la la, la la!” for the first two tweets, “falala, lalala, la la la!” for the third, and “falalala la, la la, la laaaa!” for the fourth. After each such verse, it starts again.
People seem interested by unusual-looking dates. This meme, for example, gained a surprising amount of traction considering how strangely confused it is about how two-digit years correspond to four-digit ones:
At exactly 06 mins and 07 seconds after 5 o’clock on Aug 9th 2010, it will be 05:06:07 08/09/10. This won’t happen again until the year 3010.
And then there’s this one, which is just mad.
AN INTERESTING FACT ABOUT AUGUST 2010. This August has 5 Sundays, 5 Mondays, 5 Tuesdays, all in one month. It happens once in 823 years.
All Augusts do that. Not always with Sunday, Monday and Tuesday, but near enough, and the exact combination isn’t rare. Even if it were, the calendar loops every 400 years so nothing like this could possibly happen every 823. It would be an astonishing coincidence if the 1-in-823 freak August contained the 1-in-1000 freak county date thing.
But how impressive is the county date thing really? I mean, yes, it won’t happen again until 2110, but if you use the UK date format it happened again in September. If you use the US date format, but write the time after the date, then it happened again later that morning.
And you don’t have to start at five. You could start at six, and it happened in 2011. If you start at eight, and it happened in 2013.
And you don’t have to count in ones. You could do just the odd numbers. Or just the primes. Or…
So I did what anyone would do: I fired up Python, scraped the Online Encyclopedia of Integer Sequences, and set my Raspberry Pi tweeting every interesting date it could find.