Ontology learning process involves identification of concepts and the relationships between these concepts. Automated ontology learning based on ML/DL techniques identifies these triples but suffers from the problem of duplicate or similar relationships. We propose a solution to identify similar relationships so that the ontology can be fine-tuned.

We investigate the disparity between English and non-English language models and show that detection frameworks based on pre-trained large language models like BERT and multilingual-BERT systematically perform better on the English language. We demonstrate the promise of incorporating the information contained in images via multimodal machine learning to address this disparity.

We investigate the robustness of Birdwatch against adversaries and show that the current Birdwatch system is vulnerable to manipulation attacks. To overcome this vulnerability, we develop HawkEye, a cold-start-aware graph-based recursive algorithm, and show that it is more robust to such attacks.

We use statistical analyses and regression-based techniques to predict the resolution times of incident tickets. We employ techniques like dynamic rolling window, auto-regressive window-flip and artificial data creation to handle data eccentricities.