Recently the world has been going through a lot of disasters. Some natural and the other man-made. Some people unite at these times but are unable to manage their will to help. Twitter has been really helpful in making this happen. However, this activity is not managed properly and many times tasks are overlooked or are looked by many people at once. Example: In Chennai Floods, many people tweeted about the help they required, however, sometimes the help was not given or was given by 2 sets of people. The motivation is to do a 1 to 1 matching of people who want to help and people who are in need of help with no conflicts and better management.
What it does
It takes in twitter data for particular HashTags and creates a stream classifying relief required into various categories. Currently the categories are predefined. But intelligent systems can be used to mature the set of categories in the future. Each relief request is converted into a task which can be seen by everyone in the form of tabular format. Once someone takes up a task, they are given 1 day to work on it and if they are unable to finish the task, their trust level decreases. Once they complete the task, their trust level increases. This ensures that no task is taken by multiple users and that no task is overlooked.
How I built it
We built the application on HTML and used firebase for Database. We used twitter API to get all the relief request, for login purpose and also for showing tweets from a particular hash tag in real time. We also used Facebook login API for alternate login. We used Spark and Kafka for natural language processing and classification of tweets.
We started by dividing tasks on various levels, like database, frontend and backend. Then as the app kept going we kept taking up tasks and working on it. We used slack for communicating important information and API keys.
Challenges I ran into
The main challenge was to be able to built a working application using new technologies in such sort span of time. But that was the fun of it too. Hence we love hackathons ;) :P .
Since none of us had worked on UI, we faced quite some challenges on that end as well.
Accomplishments that I'm proud of
To be able to come here even after having so many assignments' overhead ;) Frankly it would be to able to mock about how we were pathetic at different aspects, still overcoming them and coming up with an awesome app which has a scope of lot of improvements and can help people.
What I learned
We learned that what makes a great app is an awesome and nobel idea and its not so difficult to built on a new technology. You always end up learning from them.
What's next for Rescue Rangers
There is a lot to our Rescue Rangers in the future.
- Classification of twitter data is not matured enough. We can do a lot of natural language processing to eliminate rumours and classify data on the basis of tasks
- Tasks are not prioritized currently, however we are various algorithms in mind to prioritize. a. On the basis of keywords(Emergency, SOS) b. On the basis of upvotes given by humans(Sometimes computer is unable to understand exact sentiment of a request)
- Currently we determine the trust level on the basis of tasks taken by the user. We are not sure about whether the task is completed. We can have a mechanism of assuring that task marked completed is completed.