shaanchandra Goto Github PK

I’m Shantanu, a AI Research Scientist 🧠📈 at ZS. I have done my MS in AI from University of Amsterdam, thesis supervised by Ekaterina Shutova at ILLC (Institute for Logic, Language and Computation). My research focuses on graph neural networks, multi-modal AI, GANs and natural language processing 💬.

My major work so far are:

• SAFER (📚Socially-Aware Fake News Detection Framework ): We propose a novel social context-aware fake news detection framework, SAFER. It aggregates information with respect to: 1) the nature of the content disseminated, 2) content-sharing behavior of users, and 3) the social network of those users. We perform a systematic comparison of several GNN models for this task and introduce novel methods based on relational and hyperbolic GNNs, which have not been previously used for user or community modeling within NLP. We empirically demonstrate that our framework yields significant improvements over existing text-based techniques and achieves state-of-the-art results on fake news datasets from two different domains.

• meme_challenge (📚A Multimodal Framework for the Detection of Hateful Memes - NeurIPS 2020 competition track): UNITER-based architecture for detecting hateful, multi-modal content. Won 4th place of the Hateful Memes Challenge by Facebook.

• Multi-object detection and tracking for farm animals : The repository contains object tracking using 3 methods: a) Custom BoundingBox + tracking, b) Autonomous detection and tracking with unique ID tags, c) Background Segmentation. We generate the processed video with the tracked BoundingBox markings, a track plot and heat map of all the tracked objects and a .csv file with the x and y co-ordinates of each uniquely tracked object in each frame to aid any form of further processing (for instance, behavorial analysis, clustering, etc).

• Pose estimation in animals: Pose estimation in animals using DeepLabCut

• [HyGT] (coming soon): A hypergraph transformer architecture utilizing contrastive loss to process heterogeneous hypergraph structures.