The problem ShieldTalk solves

With the ease of internet access, and the ability to hide behind a computer screen, cyberbullying has given bullies a new platform to harass their victims. According to the National Institutes of Health, Teenagers with underlying mental health issues, when faced with cyberbullying, have been shown to have suicidal thoughts. One in three young people in 30 countries said they have been a victim of online bullying, with one in five reporting having skipped school due to cyberbullying and violence, in a new poll released today by UNICEF and the UN Special Representative of the Secretary-General (SRSG) on Violence against Children.

People can use our product to detect toxicity in text or audio/video. People can also get benefit if the social media or forum company uses it to monitor its users and control the amount of toxicity within its website. The benefits of using our product to detect toxicity in text or audio/video are numerous. By identifying and flagging toxic comments or messages, our product can help individuals and organizations to take action to prevent harassment, cyberbullying, hate speech, and other forms of online toxicity. This can help to create a safer and more respectful online environment, which can lead to better mental health outcomes and increased productivity.

In addition, by partnering with social media or forum companies, our product can be used to monitor user-generated content and control the amount of toxicity within their website. This can help to improve the reputation of the company, increase user engagement and retention, and reduce the risk of legal liability due to online harassment or hate speech. Overall, by providing a reliable and efficient way to detect and prevent online toxicity, our product can have a positive impact on individuals, organizations, and society as a whole.

Challenges we ran into

The challenge we faced was balancing the trade-offs between accuracy and speed. While more complex models may achieve higher accuracy, they may also be slower and less efficient, which can be a problem in real-world applications. To balance the trade-offs between accuracy and speed, we experimented with a range of machine learning models, including deep learning and ensemble models, to find the most effective combination of accuracy and efficiency. We also optimized the code and hardware used to run the models to improve their speed and efficiency.