Hacker Newsnew | past | comments | ask | show | jobs | submitlogin
Using Mathematica to generate Web 2.0 company names (collison.ie)
26 points by mqt on April 10, 2009 | hide | past | favorite | 4 comments


I don't know Mathematica, so I can't quite tell exactly what algorithm is being used in Patrick's code, but I do know that Markov chains are a useful way of generating similar names and are easy to program in common languages. (Patrick may be using Markov chaining - I see a RandomChoice from CharFreqs.)

Basically, analyze a word corpus and create frequencies for letter pairs, considering word start and end as letters for convenience. Then, starting with a word start, choose letters randomly based on the frequency of the pairs where the first letter in the pair is the last letter in your current word. Continue until a word end is chosen.

Building frequencies using letter triplets rather than letter pairs can get better results, where pairs may choose unlikely match-ups such as "cth", taking e.g. a"ct" as common, and "th"e as common. Longer tuples can be chosen but a slight adjustment of word start and end strategy is needed.


I don't know Mathematica either, but here's a Markov-model hack as barrkel suggested: http://github.com/darius/languagetoys/blob/505309d5900ac49c6...


This is a nice idea, but I was hoping for a big text dump of names I could look at. Maybe you could turn this into a web app.


iCantThinkOfBetterName

... (iPhone app)




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: