[framschwartz]

Very cool! And yes, all the data is out there and is public, it's just a problem of filtering through the noise to get the signal. Because < 0.01% of topics discussed online are sustainable trends, it can take a bit of engineering work to do this all...

1. at a non exorbitant cost 2. in a time period where it takes less time than the cycle of information otherwise it's useless by the time the trend is surfaced.

[chx]

Yeah, as I said, this was primitive: Stanford NLP was used to extract keywords out of tweets and then it was down to measure the velocity of keywords.