18 Jun 2020, 06:37

Meet the winners of the #ZindiWeekendz COVID-19 Tweet Classification Challenge

Zindi is excited to introduce the winners of the #ZindiWeekendz COVID-19 Tweet Classification Challenge. In just 60 hours, the virtual hackathon attracted 216 data scientists from across the continent and around the world, with 142 placing on the leaderboard.

The objective of this challenge was to develop a machine learning model to assess if a Twitter post is about COVID-19 or not. This model will help gather tweet data about the epidemic without relying only on key words like ‘covid’ or ‘coronavirus’ being present, allowing researchers and engineers to gather a more comprehensive dataset for sentiment analysis.

ML-driven sentiment analysis is an important tool to understand communities’ feelings around major issues such as COVID-19. Gathering comprehensive social data for sentiment analysis can be limited, however, if data collection relies only on keywords such as ‘coronavirus’ or ‘covid’. This model could be put into practice as part of a larger effort to understand online sentiment around COVID-19, and inform future communications and public interventions by governments and non-government public health organisations.

The winners of this challenge are: Krishna_Priya from India in 1st place, HaythemTELLILI from Tunisia in 2nd place and Iheb_Bouzayani also from Tunisa in 3rd place.

A special thank you to the 2nd and 3rd place winners for sharing some insights into how they succeeded in this challenge. You can see the winning solutions below.

1st place solution

3rd place solution

This hackathon will be re-opened as a knowledge competition, you can join here: #ZindiWeekendz Learning: COVID-19 Tweet Classification Challenge

Haythem Tellili (2nd place)

Zindi handle: HaythemTELLILI

Where are you from? Tunisia

Tell us a bit about yourself?

I am a second year Telecommunication Engineer Student at National Engineering School of Tunis and a machine learning enthusiast with interest in computer vision and natural language processing. I always aim high and keep myself motivated by participating in hackathons.

Tell us about the approach you took.

I used transformers (Roberta-large) and fastai which is a framework built on top of PyTorch to solve this challenge.

What were the things that made the difference for you that you think others can learn from?

A good cross-validation strategy is always the key to succeed in any kind of competition because without this, competitions are more or less like gambling. Also ensembling/blending is an important last step which helps me cover that extra mile at the end.

What are the biggest areas of opportunity you see in AI in Africa over the next few years?

Year after year, Africa shows that it can be a leader in many areas such as Artificial Intelligence. We see that companies in Africa such as Instadeep are among the top 100 start-ups in the world which is a great achievement. I see a brilliant future and many opportunities for Africa and for African data scientists in a lot of domains such as computer vision, reinforcement learning and so on.

What are you looking forward to most about the Zindi community?

Zindi is growing up very fast and I am so proud because we have such an amazing community in Africa. A huge thanks to them for trying to make Africa much better.

Iheb Bouzayani (3rd place)

Zindi handle: iheb_bouzayani

Where are you from? Tunisia

Tell us a bit about yourself?

I am a final year student in computer science. For me, data science is the fuel of the next coming years. I'd like to work in more projects related to data science and machine learning.

Tell us about the approach you took.

I used ensemble learning (two Roberta-large pretrained models with fastai)

What were the things that made the difference for you that you think others can learn from?

The data was clean. So, preprocessing data was not a weapon for anyone to win. I think for me, I took my time to tune parameters that's why I got a higher score than others.

What are the biggest areas of opportunity you see in AI in Africa over the next few years?

I see AI in every area of Africa. We have the potential to do something great. We have clever engineers and creative ones. To be precise, the biggest two areas are Education and Health.

What are you looking forward to most about the Zindi community?

Talking to the CEO :D

What are your thoughts on our winners' feedback? Engage via the Discussion page or leave a comment on social media.