bully buster - the tweet cop

• 

  Precision Recall curve to evaluate model performance

• 

  code snippet 1

• 

  code snippet 2

Inspiration

With the increased use of Social Network, cyber bullying has become a growing menace. The three popular Social Networks are Facebook, Twitter and You Tube and unsurprisingly, they lead the charts harboring the greatest number of cyber bullying incidents. In light of this, I propose this work as a base for realizing and envisioning harassment on Twitter.

What it does

This project explores the ability of natural language processing (NLP) to give a computerized approach to gauge bullying on social media and identify abusive posts. The code was implemented in Python and developed especially for the Twitter feed. However, the concepts can readily be extended to other social media platforms, including Facebook and You Tube.

How I built it

The project primarily focuses on the backend algorithm. Existing methodologies on the topic was researched and the mathematical nuances were studied. After having developed a sound algorithm, it was implemented on python as a classifier. The precision recall curves were plotted to evaluate the performance of the classifier.

Challenges I ran into

The biggest challenge was to obtain the data set within such a short interval. I tried streaming the twitter feed, but the task was time consuming. Hence, I had to resort to using existing data set from previous research work on the topic.

Accomplishments that I'm proud of

The model achieved a sound accuracy and precision of almost 75%, bettering existing works on the topic.

What I learned

1. The mathematics behind Natural Language Processing tools
2. Evaluation and testing of a model using accepted methodologies
3. Time management of adoption of strategies that work to build a model in such a short time

What's next for bully buster

1. Further testing and development of the algorithm
2. Develop a front end app or a website that uses this algorithm to test abusiveness of a particular tweet

Built With

• github
• machine-learning
• mathematics
• python