: research


I'm a Ph.D. student in CSAIL at MIT. I work with Prof. Michael Collins in the areas of non-parametric methods and acoustic modeling.

My current works involves developing non-parametric methods to perform acoustic modeling to minimize word error rate in large-vocabulary speech recognition.


Natasha Singh-Miller
Doctoral Thesis: Neighborhood Analysis Methods in Acoustic Modeling for Automatic Speech Recognition

Natasha Singh-Miller and Michael Collins
In NIPS 2009.
Learning Label Embeddings for Nearest-Neighbor Multi-class Classification with an Application to Speech Recognition

Natasha Singh-Miller, Michael Collins, and T. J. Hazen.
In Interspeech 2007.
Dimensionality Reduction for Speech Recognition Using Neighborhood Components Analysis

Natasha Singh-Miller and Michael Collins
In ICASSP 2007
Trigger-based Language Modeling Using a Loss-Sensitive Perceptron Algorithm
(extended version for RQE)

Natasha Singh
Master's Thesis: The Use of Syntactic Structure in Relationship Extraction