Yuchen Zhang's Home Page

Last update: Dec 27, 2012

Yuchen Zhang

Email: yuchenz(AT)cs(DOT)cmu(DOT)edu

I started working at Nuance Communication Inc. as Language Modeling Researcher since August, 2013.

I was a Masters student at Department of Computer Science, Carnegie Mellon University.
My division is the Language Technology Institute (LTI), and I was working in the InterACT lab as a research assistant.

My work here was generally about natural language processing, computational linguistics, speech recognition and understanding, etc.
I enjoy jogging regularly and have participated in several local 5K events back in Ohio.
I'm also a big fan of Haruki Murakami (村上春树)'s novels, do check out "Pinball, 1973", "A Wild Sheep Chase", and "The Wind-Up Bird Chronicle" if you have time.

My resume. [PDF]

* * * *

Education

M.S. in Language Technologies, LTI, School of Computer Science, Carnegie Mellon University, Pittsburgh: 2011.8 - 2013.8.
B.S. in Computer Science with cum laude Honor, Miami University, Oxford: 2008.8 - 2011.5.

[Top]

Research

School of Computer Science, CMU: Graduate Research Assistant (2011.8 – Now)

Aimed to develop rapid methods of keyword spotting among large collections of audio.
Sped up the Keyword Spotting system by 15 times by implementing a parallel processing framework using client-server model.
Implemented a script that automatically optimizes a set of parameters during confusion-network generation. Improved the accuracy of the Keyword Spotting system by 16% on average.

[Top]

Projects

hmmpytk (Dec 2012 - )

A generic HMM (Hidden Markov Model) implementation in Python.

Features:

Observations and states can be anything (primitive types, objects, tuples ... )
Does not depend on any external packages
Achieve reasonable efficiency
All the numeric calculations are done in the log domain (prevent underflow)

hmmpytk comes with an example script that does POS (Part-of-speech) tagging.

To check out the repository:

git clone https://code.google.com/p/hmmpytk/

Or visit the project hosting on Google Code

* * * * * * * * * * * * * * * * * * * * * * * * * * * *

Miso the Ontology Learner (May 2009 - August 2009)

Automated knowledge extraction system that used NLP techniques such as POS-tagging, parsing and regular expression to extraction knowledge from a textbook to build an ontology.

Download the research report here.
Source code will be given upon request.

* * * * * * * * * * * * * * * * * * * * * * * * * * * *

ThinkPad Anti-Thieft Monitor (No longer maintaining)

Application runs in background on ThinkPad computers that detects sudden movement of the laptop and send out the alarm.

You are more than welcome to take over this project and port it to Windows Vista/7/8.

Project webpage on SourceForge.net.

* * * * * * * * * * * * * * * * * * * * * * * * * * * *

Kiwi the Language Guesser

Language classifier that detects 17 natural languages in the world.
Trained the classifier with language specific features extracted from character-level N-Gram of text from each language.
Converted the testing language text into one-dimensional N-Gram vectors and then classify.

Download the executable (Java 1.6)
Source code will be given upon request.

[Top]

Travel

Road Trip to the West (December 2010 - January 2011)

Liyang Zhao, Han Zhu and I drove for 17 days and completed the road trip to California and back.
(K, L, M are places we visited in another trip in 2010, and N is Oxford, OH - where we started).

* * * *

Road Trip to the East (Fall 2010)

Pictures coming soon!

[Top]

Fun Stuff

Some journals I wrote (sorry they are in Chinese)

迈大散记 [PDF]
黄石的孩子全跳舞 (All Yellowstone's Children Can Dance) [PDF]
草原日记 (2010.5 - 2010.8, summer intern at Project Vote Smart) [PDF]

* * * *

咆哮体生成器

猛击这里开始咆哮!

* * * *

Some Links

Ryan's home page (this guy runs like crazy!) [Link]
Chinese American Culture Society (Miami University) [Link]

[Top]