Friday, March 15, 2013

Web-mining Project: Locate danger spots on a map

I started wondering how difficult it would be to automate the process of collecting information about danger spots on roads and displaying them on a map.  The first step in investigating this was to do a Google search with the query “road accident death Bangalore”. I did get a lot of relevant information, including what appeared to be an NDTV compilation:

I suspected that this compilation is easier for Google to do than for every TV channel.  So, I searched for “topic Google” and found some information about the cool feature named “topics”. (Yes, you can make a topic out of your heart-throb!).
 There was one heart-rending news item about a young neuro-surgeon’s life being snuffed out by a hit-and-run driver:

To make the web “more intelligent”, we will need good techniques to mine its content semantically and present information in a multi-media form. So, I propose a student project which will mine the web daily for information on locations (circular areas on a map, about 100 M in diameter characterized by their latitudes, longitudes and names) where fatal accidents occur. Since news text on the web uses a variety of ways of conveying location information, you will need a variety of techniques to figure out Lat-Long info, necessary for locating the spot on a map satisfactorily. You might need to process multiple sources of information on a single accident and avoid reporting it as multiple accidents. Multiple sources may also make it easier to figure out the location and time of an event more accurately. Forget getting 100% information – there is no such thing in the real world!

Now mark the spots on a map. You might want to remove a spot from the map when a year goes by without another fatal accident there.
Now, figure out how you can make this map available to anyone on the web who wants to see it, without cluttering up something like the Google map for everyone who accesses it. This should be easy.

From a technical point of view this should be fun to work on. Hopefully such a map would be seen once in a while by traffic police officers who might be tempted to visit these spots and report on the map what they intend to do about reducing fatalities at the place visited.

India, unfortunately seems to have a highest number of fatalities on the road (about 140,000 per year) in the world. China has less than half of this; USA has less than 1/4th of Indian fatalities, despite having a lot more vehicles. We would be a stupid lot, if we do not reduce the number of yearly deaths due to accidents by about 50% over the next ten years.

Srinivasan Ramani

No comments: