Google Rambler

...a.k.a. "google ipsum"

Google Rambler attempts to construct, from a single seed word, a body of text that looks like it makes sense. The words come from common search queries recommended by Google, which seem to be popular search queries from its users.

Check out the source: github.


Wait, what?

So, say you're typing in a Google search and start off with the word "house". Google's Autocomplete feature might recommend something like "houses for sale in Boston". Rambler stores the phrase, then takes the last word, "boston", and puts that back into Google's Autocomplete "API" (it's not documented, but it's open). From supplying "boston", it might get back something like "boston red sox suck". Then it'll lookup the word "suck" and so on and so forth.

Rambler tries to improve the quality of the text generated by doing a bit of lexical analysis on it, or sometimes just picking the longest recommendation.


I was curious to see if this would be an effective way to generate sensible text. I thought this might work like a Markov chain based on Google's dataset of popular search queries, which have a very human element. Turns out the text is really meaningless. So I guess this might be useful as a lorem ipsum generator.

Created by Gordon Zheng (@capable_monkey).

Try me!