Spark that evoke during chennai’s flood

Due to very heavy and unprecedented rain, Chennai was submerged . Many areas was drastic hit with huge inflows of water.

We thought of helping Chennai. Tweets were flooding, we thought of using social media as disastrous management.

Quickly gained spark equivalent to 100 redbull energy drink decided to do a night out.

Mr. @Chelladurai Pandian was grasping the trending tweets with hashtags #chennairains, #chennaifloods, #chennairescue, #chennaivolunteer soon after we got that spark of connecting help providers with help seekers.

Here we go to hatch the rough idea

  • Fetch the tweets
  • Do a text analysis
  • Decide whether the tweet belongs to help seeker, Help provider and of which category( food, items, place offered to stay, doctor required etc )and their place of availability ( among various areas of chennai)
  • classification algorithm was discussed
  • how to relate the Helper vs Seeker on which terms and properties
  • Area names of Chennai was pre-populated ( but there was a hiccup people wont tweet exact text of area there was short forms, abbreviation, eventually we decided to fetch geocode as well)
  • spatial data was obtained if it exists
  • Ahhh ! ready to dirty the hand

A game between Helper vs Seeker begins!

opened up the ear to listen on twitter

twitterStream.filter(track=[“#chennaiRainsHelp”,”#Chennaivolunteer”,”#chennairescue”])

on data classification begun. Simple text analysis was did

To identify seeker’s :

around.*now.*\?
now.*around.*\?
now.*around
need.*help
help.*need[ed]?
need.*help.*to.*contact
needed
volunteers?
volunteer?
pregnant.*lady.*stuck
need[s]?.*help
need.*without.*if
How.*is
any.*information
any.*info
Does.*anyone.*-.*.*need
can.*someone.*
someone.*can
available\?

To classify as Helper’s :

if.*you.*need
Ready.*to.*help
available\s+
available.*contact
available.*call
\bif\b.*need
if.*call
if.*contact
if.*needed
food.*pockets+.*call
volunteers.*\?
volunteers.*\?
needed\?
need.*\?
available
give

To identify which category both belongs to

medicine
blanket
milk
medical
napkin
Nightiees 
t-shirts 
saree
doctor
mosquito coil
tooth paste
biscuits
breads
pai
blankets
gynecologist
volunteer
donation
water bottle 
drinking water
sintex tank
power bank
mobile recharge
petrol
oxygen cylinder

and one more last property is to identify the area so that we connect them first based on proximity

Vadapalani
Valasaravakkam
Vallalar Nagar
Vanagaram
Velachery
Veppampattu
Villivakkam
Virugambakkam
Vyasarpadi
Washermanpet
Kummidipondi
West Mambalam
openarea ………………..

ah ! classification algorithm worked we tamed it by adding few more text manually, data the gold kept growing …more interesting now how to relate them ?. Our machine learning algorithm needs to be tamed for this as well…interesting really is it ….

discussed it ! Google it ! go-ogled it ! gooooooogled it ! nailed it !

the tool is graphDB made the task easier

prioritize the connection with the proximity and add weight-age if both the parties has symmetrical category . Eliminate the retweets and ensure if a seeker is once connected with helper avoid redundant connection as helpers are priceless and more humanness.

So, it came to the end by connecting the Helpers and seekers about sharing their contact.

That’s the game between Helper vs Seeker !!