I have always wanted to do a project using Machine Learning but didn't know where to begin. I seeked help from one of my professors who gladly took me under her and became my guide for a ML project. The topic of the project was "Cyberbullying Detection on Twitter against various personalities and hateful topics".
I used Twitter API to fetch tweets by creating an application in Twitter developers. The tweets collected were targeted by famous personalities or had to have some hateful topics. I worked only on Hinglish data.
I identified the sentiments of English and Hindi words/pharases. The tweets were then classified as bullying or non bullying based on sentiments. Then, I manually classified all the bullying tweets into direct or indirect bullying.
Once the training set was ready, I employed various machine learning, deep learning and hybrid algorithms to my datasets. I analysed the performance of each algorithm using metrics like Accuracy, Precision, Recall, F1 score and ROC AUC. I chose accuracy and ROC-AUC to identify the best performing algorithms.
I also calculated the credibility of twitter users by using a set of twelve rules mentioned in the research paper [1].
It was a research based project and as a beginner of Python at the time, my skills definitely improved. I also came to understand various machine learning and Natural Language Processing concepts.
References:
[1] Geetika Sarna and M.P.S. Bhatia, “Content based approach to find the credibility of user in social networks: an application of cyberbullying,” International Journal of Machine Learning and Cybernetics, Vol. 8, Issue 2, pp. 677 – 89, 1 April, 2017.
Top comments (0)