top of page

Research

Kalpayita - A Machine Learning approach to Interior Designing

International Journal for Research and Scientific Development

Researched on Text to 3D scene conversion methods and developed a desktop application for Interior Designers. The code is in Java and utilizes Stanford CoreNLP pipeline for processing input text. After the text is processed, a scene graph is created and rendered on a desktop application using JMonkey.

Publication: http://ijsrd.com/Article.php?manuscript=IJSRDV6I30847

Harnessing label semantics to extract higher performance under noisy label for Company to Industry matching

Workshop on ‘Small Data, Big Opportunities: Making the most of AI’ at 3rd ACM ICAIF 2022

Utilized minimum labeling strategy (with only ~15% labelled data) to train a Sentence Transformer model on regression objective. Achieved 30% average precision and 99% average recall on cases where noisy labels were completely incorrect and overall 8% gain in precision and 53% gain in recall over noisy labels.

Publication: https://arxiv.org/abs/2212.01685

bottom of page