Dvesh Prahari : Attackers in Disguise

Deals with detecting hate speech and sarcasm in Indian Political memes in context of code-switched languages
Curated and annotated an IPM (Indian Political Memes) image dataset consisting of political memes
Proposed CNN-LSTM model that processes image and text separately for state-of-the-art results in this domain
Applied the model to several twitter pages to filter out the Indian twitter pages that provoke vindictive sentiments during elections
Improvised further on the model accuracy and pre-processing techniques post thesis and working on a paper
Tech Stack: Python, NLTK, Keras