Ping

Global directory of areas where the demand of help of organizations is in real need

Comment 2

Inspiration for Ping

The Challenge

Charity organizations and helpful institutions receive many issues reported by hyperlocal area representatives (city majors, district leaders, village leaders, tribal leaders, …) every day.

Main issues are reported in the categories of: Health (nutrition, child mortality) Education (years of schooling, schooling attendance) Standards of living (electricity, housing, internet connectivity …)

Because of the large amount of incoming data it is hard to process and identify the ones that are more relevant over the others. Some inquiries manage to stand out by requesting more often, state issues better but may not be as relevant as other issues that are not well stated or requested less frequently. Emotional factors play an additional role and weigh in on making the right decisions.

The Solution

“Ping” shall provide support to identify currently relevant issues by processing incoming requests automated and funneled into a chatbot that is wired to a natural language processing infrastructure. Incoming requests with information like location, situation descriptions, and attached media (photos, videos) are measured against existing data and indexes backed by research of large institutions like the United Nations.

And Ping also takes unusual anomalies in consideration. An example would be issues reported in an area that actually reflects wealth via its indexes.

We wanted to create a supportive platform that makes use of Facebook technology infrastructure, automated processes and reference data to identify requests and areas in accurate, true and real need with support of machine learning and factual automated ways.

All incoming requests are being classified via Natural Language Processing and entries are shown filtered by categories to match organizations with topic overlaps.

Example:

A request which is identified around child starvation or child mortality is automatically shown in Ping and going forward makes automated decisions via machine learning optimizations for UNICEF, to be able to instantly reply to the issuer. As first version a simple email is being sent to the matching representative.

To get more context about incoming requests they are being visualized on a heatmap filled along with existing research information about the area (like the Multidimensional Poverty Index, Wealth Index, GDP, Politics, Urgency Index) where it is an ease to identify the urgency by the requested area.

Interactive On-Site Field Guides

AR visualizations additionally help in providing an overview to identify the relevant areas in true demand of help and give more data insight about specific areas. Applied with the Multitracker Example (Tracker 1: Field Map, Tracker 2: Objects in the surroundings) below these can act as interactive on-site field guides. Make sure to try the two trackers in the comments to get the experience.

Credit:

SDG9, SDG11 and SDG8 experts (Jamie Yang, Yatin Mana from Harambee, Samhir Vasdev and Michael Trucano, Jason Whittet, Albert Chan from the Facebook Staff provided us requesting access to the Whatsapp API via Twilio as Messenger alternative, because locals tend to use Whatsapp (input via disaster expert John Crowley at NetHope) provided invaluable feedback to provide an accurate solution that can be mapped to real world challenges.

Getting the word out

First we need to let people know that Ping exists as an app:

Initially organizations need to get contacted by us that Ping exists as a free tool that they can use to accurately identify areas in urgent or more relevant need (we ideally get in touch with global leading mother organizations) Then organizations reach out to local representatives and institutions that Ping exists and that they can submit their relevant issues from now on without any hassle.

What it does

1. For helpful organizations and institutions it is a supportive platform to identify requests and areas in true need

It is a collection of requests that have been submitted by hyperlocal representatives of areas in claimed need of support.

Local representatives can submit their inquiries via Facebook Messenger Chatbot on the Ping Facebook Page.

Organizations then get a “Ping” via Facebook Messenger or Whatsapp Notification about an area that is being identified as in true and accurate demand of help.

2. For local representatives it’s easy to submit and they know that their request is being processed, responsibility is being taken and that it is not pointless

The barrier to submit an issue via Messenger/Whatsapp is very low and local representatives receive an answer (immediately) by a dedicated agency to know that they are being heard and that their request is being processed.

3. Measure change via automated “Pings” to local representatives

After a certain period (like 6-12 months after support has been issued) local representatives automatically get asked to provide information via answering questions on the current state to measure change through the preferred communication channel (Facebook Messenger Bot or Whatsapp).

4. (Future) Automated reports to measure change

Going forward we plan to add the feature of automated reports via parsing live machine learning data to measure the state of improvement or regression.

How we built it

Facebook React, Facebook Login, Javascript, .NET Core, LiteDB, HTML, CSS. Facebook Messenger Chatbot API. Photoshop for coherent guides and image assets. Python and pytorch for the machine learning computing for the identification of poor areas.

We use NLP and NER to automatically add tags based on the description text

We initially wanted to implement the 2018 Global Multidimensional Poverty Index (MPI) to provide a real world indicator into the Google Map that is visually represented.

But since the MPI breaks it down to nation wide data only we used an alternative dataset from Kaggle which reflects poverty rates broken down on the US states

We added a visual layer of the local infrastructure development to get a better understanding of the level of the claimed need by area.

We received the input by Albert Chan that Whatsapp is often preferred over Messenger in rural and urban areas and therefore we requested access to the Whatsapp API via Twilio (thanks to Albert Chan of Facebook Staff) to provide support for submissions and notifications via Whatsapp as well.

We open sourced this project

The code and files of the web app, chatbot and Spark AR effect are open sourced. You can find it here.

Challenges we ran into

We initially wanted to create an educational version of Ping which provides a collection of resources and content to educate areas via dynamic content and media like AR Effects, Chatbots and printed educational guides in the areas of nutrition, child mortality, internet connectivity, electricity, transportation, housing …

But via checking the idea back with SDG Experts it turned out that local representatives mostly run into the issue of something that broke which worked earlier or where more resources are required to take full action.

We spent most of the two days understanding the real world issues better to provide a solution that hopefully helps to improve the current state in the day to day of areas with needs.

Accomplishments that we're proud of

After our long ideation sessions throughout the two days and receiving so much helpful feedback and just starting developing in the early evening on Day 1 we were happy to roll out the first version of the Webapp its API’s and Chatbot at around 9pm :)

Vision

Our Vision with Ping is to be able to predict possible occurrences of incidents and issues to prevent and secure them before they actually happen.

What's next for Ping

Initially get in touch with representatives of organizations and polish the idea to help improve the support for real world issues and improve the product on the technical side.

In addition to the Multidimensional Poverty Index we want to add more indexes (like the Wealth index, Health index) to identify even more accurate and relevant needs.

Planned Updates:

  • Improve the processing of incoming messages by NLP and machine learning (by measuring against data sets to identify the urge of a request or area)
  • Automated decision making per case
  • Improve the UX with Whatsapp
  • Add support for further messaging apps in addition to Messenger
  • Automated Reports via live machine learning data to measure the state of improvement or regression

Built With

Share this project:

Updates