Research
My primary research interest is in natural language processing. Most of my recent work has been on Machine Translation. As a Masters student at CMU-LTI, I am working on leveraging linguistic syntax to improve the quality of automatically translated text. Currently, I am exploring how syntactic information can help improve word order in the output of phrase-based MT decoders.
My advisors: Alon Lavie, Stephan Vogel
Education
MS (Language Technologies), Carnegie Mellon University
2007--2009. Current QPA: 3.84
B.Tech (Info Technology), Indian Institute of Information Technology
2002--2006. GPA: 9.83/10 University Gold Medalist, 2006.
Publications, Coursework, Research Experience, Achievements:
Refer to Full CV
Tools Developed
TRAMBO (Fall 2008)
Trambo is a framework that allows any MT decoder to be run inside the Hadoop framework. Running an MT decoder (Tuning/Decoding) can often be very time consuming. Trambo enables faster experiments by distributing the task over a Hadoop cluster.
Trambo currently supports these decoders: Moses, CMU-STTK, CMU-XFER, Treegraft. Adding support for more decoders is fairly easy. Trambo's Minimum Error Rate Training is due to Ashish Venugopal.
Trambo has been set up on the Yahoo! M45 cluster. If you have access to the cluster, and would like to use this framework, please send me an email.
ATAVI (Fall 2007)
ATAVI is a tool that visualizes parse trees of bilingual parallel sentences along with word alignments.
