My major inspiration is to help the needy in this pandemic situation. In my country like India, many of the poor people don't have proper knowledge of websites, or even access to televisions. Due to this, many of the information regarding the precautions, nearest hospital facility, etc. never reach these people, and due to that many people are dying on the streets just because of either they don't visit proper health facility in time, or which they visit, doesn't have space. But one thing people have is the smartphone, no matter cheap or expensive. And, one app I see people using frequently is Facebook (one of major impact by the company in connecting people). So, I and my teammate, decided to build a bot in Fb Messenger itself, so people can come in contact with this often, and just don't have to learn and search throughout the websites or streets, and all the information are easily accessible with just one question away.
What it does
This bot tells you about the current cases, deaths, and recoveries from COVID-19 countrywise and also statewise but only limited to India, for now atleast. Also, one of the key features is that, if you have symptoms or if you feel you have symptoms, this bot can lead you to various helpline numbers so that help reaches immediately if any serious case is encountered. If your health is deteriorating, this bot also provides information about various hospitals in Indian states treating COVID-19 patients, and also tells you if hospitals have space or not (this solely depends on state), so that you reach the right hospital at right time.
How I built it
We used nodejs, messenger api, and wit.ai, and made them communicate with each other. Questions received from messenger ai were sent to wit.ai for NLP synthesis, and data processed was parsed to nodejs for further actions. After actions, the result was reverted back to messenger api as their possible answer. We used different public websites to scrape information and accelerated their goal of transferring useful information and resources to the people.
Challenges I ran into
The major challenges were caused while collecting data especially live feed data. Collecting this type of data was more important for better analysis of the current scenario. Another major challenges were to extract information from wit's response. For different sentences, wit returns different json structure, for eg., wit identifies some locations but isn't able to resolve it. For handling such scenarios, each possible sentence was carefully tested using Wit in Python. Also, identifying and catching bugs in the questions asked was by far the toughest challenge to tackle. Due to greater efficiency of Wit, it extracted useful information and treated the question with some other similar question's answer. Also, we had some minute issues in communicating with the websites, that was later resolved quickly.
Accomplishments that I'm proud of
I'm proud to build something useful not for the proper earning, educated people, but for the people those don't have access to many priviledges. Through this hackthon, I got an opportunity to build something useful for the community.
What I learned
Through this hackathon, I feel obliged to get introduced to Wit.ai. It is such a nice framework with much accuracy and was impressed by keyword detection and also it predicts the context with much greater accuracy. Also, with this project, I learnt how to detect points or fields to target, or which needs to be addressed to bring a massive change in the community for its betterment.
What's next for Wit COVID
If you think with broader perspective, there is still an important feature missing i.e. identifying COVID hotspots or sealed areas near our houses. We were working on this feature, but due to lack of time, we decided to put a hold on this. Also, we weren't able to add various NGOs working for COVID relief, as these organizations are the fastest to reach an infected person or area, and pay proper medical attention before the actual medical facilities arrive.