I love this! What's great to me about this article (and so many of your other articles) is how you're taking an already-interesting idea like probabilistic sentence fragment generation using Markov chains and then recontextualizing into the world of surrealist art. There's so much to be gained by reading this through!
I made a similar thing in Ruby, but focused on individual words instead of sentences: rubygems.org/gems/markov_words/ . I was interested in generating passwords that sound like English so that they're easy to remember but hard(er) to guess with dictionary attacks. I discovered a similar thing about prefix length, although with single-word Markov-generators what you get baby talk on one end, and complete words with strange endings on the other.
It's pronounced Diane. I do data architecture, operations, and backend development. In my spare time I maintain Massive.js, a data mapper for Node.js and PostgreSQL.
Thanks! That's an interesting use for Markov chains, although I think it's a war password managers have won overall due to the sheer volume of accounts people have to make everywhere. You can't get them in everywhere though so there'll always be a call for backup strategies.
I remember reading ages ago on The Daily WTF from someone who implemented a password generator with a similar thought about memorability, using romanized Japanese phonemes due to their consistent pronunciation -- only to discover well after deployment that the Anglophone userbase found the occasional juxtaposition of pairs like "shi" and "te" memorable for less than entirely desirable reasons. As far as I can recall that was purely random, although the Scunthorpe problem could come up with predictive output too.
The thing I didn't mention is that I may have also implemented my own password manager for my org which uses my password generator API :) The service fees on password managers are outrageous for group usage! But mine is decidedly less well-integrated into every single app platform. Still, it works well for sharing passwords.
I also definitely have the problem with inadvertently generating words that are not exactly things you'd want people shouting in the computer lab. I hadn't heard about the Scunthorpe problem, but that makes a lot of sense. So what we've learned is that if you give an army of monkeys typewriters, they won't write Shakespeare very often, but they'll spout out something profane every few minutes :)
For further actions, you may consider blocking this person and/or reporting abuse
We're a place where coders share, stay up-to-date and grow their careers.
I love this! What's great to me about this article (and so many of your other articles) is how you're taking an already-interesting idea like probabilistic sentence fragment generation using Markov chains and then recontextualizing into the world of surrealist art. There's so much to be gained by reading this through!
I made a similar thing in Ruby, but focused on individual words instead of sentences: rubygems.org/gems/markov_words/ . I was interested in generating passwords that sound like English so that they're easy to remember but hard(er) to guess with dictionary attacks. I discovered a similar thing about prefix length, although with single-word Markov-generators what you get baby talk on one end, and complete words with strange endings on the other.
Thanks! That's an interesting use for Markov chains, although I think it's a war password managers have won overall due to the sheer volume of accounts people have to make everywhere. You can't get them in everywhere though so there'll always be a call for backup strategies.
I remember reading ages ago on The Daily WTF from someone who implemented a password generator with a similar thought about memorability, using romanized Japanese phonemes due to their consistent pronunciation -- only to discover well after deployment that the Anglophone userbase found the occasional juxtaposition of pairs like "shi" and "te" memorable for less than entirely desirable reasons. As far as I can recall that was purely random, although the Scunthorpe problem could come up with predictive output too.
The thing I didn't mention is that I may have also implemented my own password manager for my org which uses my password generator API :) The service fees on password managers are outrageous for group usage! But mine is decidedly less well-integrated into every single app platform. Still, it works well for sharing passwords.
I also definitely have the problem with inadvertently generating words that are not exactly things you'd want people shouting in the computer lab. I hadn't heard about the Scunthorpe problem, but that makes a lot of sense. So what we've learned is that if you give an army of monkeys typewriters, they won't write Shakespeare very often, but they'll spout out something profane every few minutes :)