Refugee Crisis

UN Refugee Agency

UNHCR, the UN Refugee Agency, is a global organization dedicated to saving lives, protecting rights and building a better future for refugees, forcibly displaced communities and stateless people. USA for UNHCR supports UNHCR and its partners, providing lifesaving essentials including shelter, water, food, safety and protection. Around the world, they help refugees survive, recover and build a better future.

A critical aspect of USA for UNHCR’s operations is the collection of donations, specifically over phone calls, which are integral to the work that the organization does. These phone calls currently exist in unharvested audio files, and USA for UNHCR would like to use these audio files to discover ways of improving their operations to better serve donors.

The Project

Thus, your project will be to develop a system that converts these audio files into text and then, using a combination of NLP/general machine learning plus data visualization tools, analyzes these phone calls to better understand how USA for UNHCR can best serve donors over the phone. We propose the project with the following milestones:

Milestone 1: Base Deliverable
Create a system/script that when fed phone calls in the form of audio files, can convert these audio files into text to be analyzed. The system should be able to accurately transcribe the audio so that we do not lose or misrepresent any important audio data. This task should be done using Python, and you are encouraged to use any existing libraries to aid in this deliverable.

Milestone 2: Text Analysis
Once you have successfully transcribed these audio files, your next task will be to apply NLP/general machine learning to conduct topic modeling on these texts, with the goal of modeling trends in how the USA for UNHCR interacts with phone donors.

Milestone 3: Data Visualization via Tableau
Your final task will be to take the results from Milestone 2 and visualize it in some useful form using data visualization tools, preferably Tableau. This visualization should make clear to a general user what trends resulted from the topic modeling you conducted, and it should be easy for the USA for UNHCR to understand what to focus/not focus on in terms of bettering their operations in working with phone donors.

Recommended Skils (not all required):
  • Python
  • Machine Learning/NLP
  • Data Visualization/Tableau Experience