Pages

Sunday, 4 March 2012

Google Prediction APIs

Whatever the millions of Google Plus detractors might be saying around the world, Google Plus still is an incredible social platform for anyone with a flair for anything techie. To be more generic, it's incredible for any person who wants to know something new. Simply put, you come to know of newer and awesomer things happening around the world, on G+, which you otherwise wouldn't. You circle the right people, get into the right conversations, and before you know it, you're wiser than all the fools who are guffawing at you for doing the same.

But relax, this isn't another of those "Google Plus is awesome" posts of mine. This is something far more interesting.

Enter, Google Prediction APIs.

I came to know of this barely an hour or so back, on (no points for guessing) Google Plus. It pretty much turned the horrible day that I had been having till then into one that was worth living through.

So what are Google Prediction APIs? They are a handy set of cloud based machine learning tools that you can use in your web applications to analyse data. Using these, you can add the following features to your application - (copied from the developers page)

  • Customer sentiment analysis
  • Spam detection
  • Message routing decisions
  • Document classification
and a whole lot more.

Now, a brilliant application developed by Marc Cohen which uses the Google Prediction APIs is this.

It is essentially an application that takes a specific geographical location as input and gauges the weather that is there at the moment. Now, the catch is, that it's not just another of the hundred million weather apps on the internet. The source of its data is Twitter. And it uses Sentiment Analysis from Tweets sent from that area to estimate the weather there. What is even more awesome is that its been neatly integrated with Google Maps to provide a brilliant interface for the input!

So here's how it works:
  • You choose the geographical region of interest from the Google Maps frame that comes up.
  • The Google Maps APIs sync with Twitter APIs to generate all the tweets that have been posted from the specified geographical region.
  • The Google Prediction APIs kick in and scan the whole lot of tweets and look for weather related data, matching anything found against a training set of data.
  • It combines its scan results to give a fairly accurate prediction of the current weather in the region of interest.
I feel that this is one incredible application that showcases some seamless integration with Google Maps APIs, Twitter APIs and the Google Prediction APIs used in conjunction with the CrowdFlower crowdsourcing platform.

I have always wanted to get my hands dirty with Google APIs and this seems to be an awesome place to start. Let me know what you think in comments.

Cheers!
Post a Comment