कार्तिक रघुनाथन (Karthik Raghunathan)

Ktick
  • Scientist II (Bing Speech and Language Sciences), Microsoft Silicon Valley
  • Alumnus, Stanford Computer Science (Graduate class of 2010)
  • Alumnus, NIT Calicut Computer Science and Engineering (Class of 2008)

[PRINTER FRIENDLY]

Research Interests

Natural Language Processing, Conversational Human-Machine Interaction, Artificial Intelligence


Academics

Stanford University, California, USA

2008-2010
Master of Science (MS) in Computer Science (Specialization: Artificial Intelligence)
with Distinction in Research* in Natural Language Processing
GPA: 3.91/4.00

*granted to 8 students in a graduating batch of 147 students

Masters Research Report: Simple Coreference Resolution with Rich Syntactic and Semantic Features: Is it Enough? [pdf]
Selected Coursework: Machine Learning, Natural Language Processing, Information Retrieval and Web Search, Speech Recognition and Synthesis, Natural Language Understanding, Foundations of Cognition, Seminar in Lexical Semantics: Space and Motion, Seminar in Psycholinguistics: Information-Theoretic Models of Language and Cognition, Research Project in Artificial Intelligence, Designing casual learning software for iPhones

National Institute of Technology (NIT), Calicut, India

2004-2008
Bachelor of Technology (B.Tech) in Computer Science and Engineering GPA: 9.14/10.00

Selected Coursework: Computational Intelligence, Network Security, E-Commerce, Web Programming, Advanced Data Structures, Design and Analysis of Algorithms, Logic for Computer Science, Number Theory and Cryptography, Computer Architecture, Compiler Construction, Computer Networks, Database Management Systems, Operating Systems, Principles of Programming Languages, Software Engineering, Communication and Information Theory, Computational Combinatorics and Graph Theory, Computer Organization, Computer Hardware Design, Computer System Software, Discrete Computational Structures, Theory of Computation, Data Structures and Algorithms, Logic Design, Program Design, Introduction to Computing, Probability and Statistics


Publications

Karthik Raghunathan, Heeyoung Lee, Sudarshan Rangarajan, Nate Chambers, Mihai Surdeanu, Dan Jurafsky and Christopher Manning. 2010. A multi-pass sieve for coreference resolution. In Proceedings of the 2010 Conference on Empirical Methods in Natural Language Processing. [Full Text, BibTeX]

Adam Vogel, Karthik Raghunathan and Dan Jurafsky. 2010. Eye Spy: Improving Vision through Dialog. In Proceedings of the 2010 AAAI Fall Symposium Series. [Abstract, Full Text, BibTeX]


Work Experience

Microsoft Corporation, Sunnyvale, CA

August 2012 onwards
Speech Scientist II  

Microsoft Corporation, Mountain View, CA

March 2012 - August 2012
Research Software Development Engineer II  

Microsoft Corporation, Redmond, WA

July 2010 - March 2012
Software Development Engineer  

Stanford Natural Language Processing Group, Stanford, CA

Sept 2009 - June 2010
Graduate Research Assistant  

Microsoft Corporation, Redmond, WA

June 2009 - Sept 2009
Software Development Engineer (SDE) Intern  

Stanford Natural Language Processing Group, Stanford, CA

Sept 2008 - June 2009
Graduate Research Assistant  

Microsoft Research (MSR), Bangalore, India

April 2007 - July 2007
Research Intern  

Skills

Natural Languages

English, Hindi, Gujarati, Tamil

Programming Languages

Java, C#, Perl, C++, C, SQL, MATLAB

Speech / NLP / AI Tools:

JavaNLP, Berkeley Aligner, Giza++ Word Alignment Tool, Moses Statistical Machine Translation Toolkit, Robot Operating System (ROS), HMM Toolkit (HTK), CMU Sphinx Automatic Speech Recognition System, Festival Speech Synthesis System, VoiceXML

Other tools

Eclipse, Microsoft Visual Studio, Microsoft SQL Server Management Studio, Vim, GIT, SVN, LaTeX, LEX, YACC

Musical Instruments

Keyboard


Projects

Coreference Resolution

Sept 2009 - June 2010

During my second year at Stanford, I conducted research on the problem of coreference resolution, which became the main topic for my Masters research. We started by attempting to replicate Haghighi and Klein's work from EMNLP 2009, and generalize their algorithm to remove the necessity of knowing the test set in advance for directing the semantic bootstrapping algorithm (see research report for more details).

However, from our experiments, we found Haghighi et al. (EMNLP 2009)'s bootstrapping approach to be too unpredicatable in a generic setting and hence explored other ways to improve the system in the absence of a semantic module, while still keeping the overall system deterministic. Using insights gained from error analysis, we implemented a cautious multi-sieve system which globally shared information across entities. Our system, despite being a simple deterministic one outperformed many state-of-the-art supervised and unsupervised models on several standard corpora. We presented our work at the 2010 Conference on Empirical Methods in Natural Language Processing at MIT, Massachusetts.

Our coreference resolution system was implemented using the JavaNLP framework and is available for download as part of the Stanford CoreNLP suite of tools.

[Report, Paper, Talk Slides, Software]

Improving Vision through Dialog

Apr 2010 - June 2010

The research goal behind this project was to enable robots to identify novel objects in new environments, i.e. objects, that their vision systems had not already been trained to recognize. However, instead of the conventional computer vision method of manually collecting new training data for the object of interest, we envisioned having a robotic dialog system that learns names and attributes of new objects on the fly through spoken interaction (modeled on lines of the children's games "I Spy" and "20 Questions") with a human tutor. We presented our work at the 2010 AAAI Fall Symposium Series at Arlington, Virginia.

[Paper, Demo Video]

Bits & Bots

Jan 2010 - Mar 2010

As part of our graduate course on "Designing casual learning games for the iPhone (EDUC 396X)", we developed a shooter game for the iPhone that taught school kids about logic gates and boolean logic. The game involved using bits (0's and 1's) as ammunition to equip a gun that functioned as a logic gate and taking down rogue robots by producing the appropriate output from the gun. Difficulty across levels was varied using factors like the number of different guns being available for use, speed of the marching robots, etc.

[Slides]

SMS Text Normalization

Apr 2009 - June 2009

We developed a system for converting textspeak (language used in SMS communication) to proper English using the Moses statistical machine translation system, as part of the graduate course on Natural Language Processing. We later presented this work in the PhD Poster Session of Stanford Computer Forum's annual affiliates meeting in April 2010.

[Report, Poster]

A Situated, Embodied Spoken Language System for Household Robotics

Jan 2009 - Apr 2009

As a combined class project for the the graduate courses on Speech Recognition and Synthesis and Research Project in AI, we developed a spoken dialog interface to the Stanford AI Robot (STAIR) for giving instructions for simple fetching tasks. Our code is now a part of the SAIL ROS package.

[Report]

TagEz: Flickr Tag Recommendation

Sep 2008 - Dec 2008

We built an automatic tag prediction system for images on Flickr.com using machine learning on both linguistic and vision features, as part of the graduate course on Machine Learning.

[Report]

Automated Receptionist for the Computer Science Department at NIT Calicut

Jul 2007 - Mar 2008

Adapted prior work done at IIIT Hyderabad to serve as an automated receptionist for the NITC CSED Faculty Directory. The system was implemented for Tamil and English languages.


Rapid Protoyping of Spoken Dialog Systems for Indian Languages

Apr 2006 - Jun 2006

As part of my summer internship at the Speech Lab (part of the Language Technologies Research Center) in IIIT Hyderabad, I contributed to the research on methods to rapidly build restricted domain spoken dialogue systems for various Indian languages. Using open source speech tools (Sphinx II Decoder, CMU Statistical LM Toolkit, Festival), we developed a spoken dialog system that functioned as an automated receptionist and handled queries related to the IIIT Faculty Directory. The system was implemented for Telugu, Tamil and English languages.

[Report]


Other Activities

Masters Admissions Committee Member, Stanford Computer Science

Feb 2010 - Mar 2010

Was a part of the Masters admissions committee that screened applications for the MS in Computer Science program (Fall 2011) at Stanford University.

Masters Student Advisor, Stanford Computer Science

Fall 2009

Was one of the student advisors selected for helping out and advising the newly admitted students in the Stanford Computer Science Department's Masters program.

Customer Support Admin, Stanford NLP Group

Sept 2009 - Jun 2010

In charge of providing customer support for the Stanford Natural Language Processing group’s software distributions (nlp.stanford.edu/software/).

Secretary, Computer Science and Engineering Association (CSEA) at NITC

Jul 2007 - Apr 2008

Headed the CSEA (the official association of the Department of Computer Science and Engineering), which plans and organizes all computer science related activities at NIT Calicut.

Participant, MSR-IISc Summer School on NLP

May 2007

Attended the summer school on Natural Language Processing at the Indian Institute of Science (IISc), Bangalore, conducted by Microsoft Research (MSR) India in collaboration with IISc and Department of Science and Technology, Government of India.

Speaker, Seminar-cum-workshop on Spoken Dialog Systems at NITC

Oct 2006

Conducted a seminar-cum-hands-on workshop during Tathva '06 (the annual all-India technical festival of NIT Calicut), guiding the participants to implement restricted domain spoken dialog systems using open source speech tools.

Member of Organizing Committee: FOSS, Tathva and Ragam at NITC

2005 - 2008

Have been a member of the organizing committee in various editions of NIT Calicut’s Free Open Source Software (FOSS) Meet, Tathva (all-India technical festival) and Ragam (all-India cultural festival).


Awards and Achievements


[TOP]