Ph.D. candidate in Computer Science with eight years of industry experience. My research interests are grounded language learning, natural language processing, and machine learning. 

• Explored design factors in grounding vision-language pairs, and presented probabilistic interactive learning algorithms that outperformed baseline by at least 5% in reducing the annotation costs in highly noisy, complex, and varied datasets.
• Proposed an unsupervised semantic model using statistical language similarity metrics and overcame limitations of finding negative grounding samples in the corpora of perceptual and linguistic data. This model outperformed baselines by 20% in learning the semantic representation of object traits. 
• Generalized grounded language acquisition by moving away from predefined categories to category-free learning by using a deep generative neural variational autoencoder. It guaranteed a minimum linguistic grounding score of 0.45 compared to minimum baseline performance of less than 0.1.
• Proposed statistical metrics to automatically analyze the complexities in the multi-modal, multi-lingual sensor data in order to reason the quality of the grounded learning systems.  It will assist grounded language learning researchers with improved design decisions. 

Published eight research papers, including an AAAI publication on grounded language learning with perceptual robot sensor data.

Measuring Perceptual and Linguistic Complexity in Multilingual Grounded Language Data. Nisha Pillai, Cynthia Matuszek, and Francis Ferraro.
In FLAIRS, 2021 

Sampling Approach Matters: Active Learning for Robotic Language Acquisition. Nisha Pillai, Edward Raff, Francis Ferraro, and Cynthia Matuszek.
In IEEE BigData (special session on machine learning in big data), 2020
Paper   Slides     

Building Language-Agnostic Grounded Language Learning Systems.Caroline Kerry, Nisha Pillai, Cynthia Matuszek, and Francis Ferraro.
In IEEE International Conference on Robot & Human Interactive Communication (Ro-Man), 2019
Paper

Deep Learning for Category-Free Grounded Language Acquisition. Nisha Pillai, Cynthia Matuszek, and Francis Ferraro.
In NAACL Workshop on Spatial Language Understanding & Grounded Communication for Robotics (NAACL-SpLU-RoboNLP), 2019
Paper   Slides   Poster 

Optimal Semantic Distance for Negative Example Selection in Grounded Language Acquisition. Nisha Pillai, Francis Ferraro, and Cynthia Matuszek.
In Robotics: Science and Systems (R:SS) Workshop on Models and Representations for Natural Human-Robot Communication, 2018
Paper   

Unsupervised Selection of Negative Examples for Grounded Language Learning.. Nisha Pillai and Cynthia Matuszek.
In the 32nd Conference on Artificial Intelligence (AAAI), 2018
Paper   Slides     

Identifying Negative Exemplars in Grounded Language Data Sets.. Nisha Pillai and Cynthia Matuszek.
In Robotics: Science and Systems (R:SS) Workshop on Spatial-Semantic Representations in Robotics, 2017
Paper   Slides   Poster

Improving Grounded Language Acquisition Efficiency Using Interactive Labeling.. Nisha Pillai, Karan. K. Budhraja, and Cynthia Matuszek.
In Robotics: Science and Systems (R:SS) Workshop on Model Learning for Human-Robot Communication, 2016
Paper   Slides   Poster