social media search engine question

I came across this site called social mention and am curi开发者_Python百科ous about how applications like this work, hopefully somebody can offer some glimpses/suggestions on this.

1. Upon looking at the search results, I realize that they grab results from facebook, twitter, google.... I suppose this is done on the fly, probably through some REST api exposed by the mentioned?

2. If what I mention in point 1 is probably true, does that means sentiment analysis on the documents/links return is done on the fly too? Wouldn't that be too computationally intensive? I am curious because other than sentiments, they also return the top keywords in the document set.

3. They have something called the "trends". They looked like the trendingtopics in twitter, but seems like they also include phrases >3 words long. Is this relevant to nlp's entity extraction or more to keyphrase extraction? Is there apis other than that of Twitter that provides this? Is "trends" generally done on search queries submitted by users or do the system actually processes the pages?

A curious man.

sentiment can be fast and on the fly, if it is for example rule-based and the dictionaries are in memory. Curious? Get in touch