|
Research
|
Current Research Projects
| Vandalism Detection in Wikipedia |
|
My
research on vandalism detection focuses on identifying elusive vandal
edits that cannot be easily detected by either simple text features or
simple edit patterns. I have identified a set of novel features that
combine both stability of text and editing patterns of users. Using the
PAN vandalism corpus published by Bauhaus-Universität Weimar, I have
demonstrated the effectiveness of my approach to detect elusive vandal
edits.
|
Data Lineage Management on Collaborative
Documents
|
|
My
research on data lineage focuses on developing efficient data
structures to index and query document metadata at fine granularities
(sentence or word). The metadata describes who contributes which part
at what time. Current techniques either manage metadata at the document
level or assume oversimplified editing models. I have designed a
persistent data structure to efficiently index the metadata of
collaborative documents at the word level. Furthermore, my data
structure supports different kinds of editing operations including
inserts, deletes, undoes, and reverts. The system I have built is being
used to manage the metadata over seven millions of Wikipedia
articles.ities.
|
| Data
Consistency
Control
in
Collaborative
Editing
Systems |
|
My
research on data consistency control focuses on adapting transactional
techniques in replicated database management systems to build a robust
and scalable framework for the development of various collaborative
editing systems. The framework provides suitable primitives to model
diversified consistency models found in collaborative environments.
These primitives consider four aspects: granularity of sharing, time to
release local edits of users, notification of editing conflicts, and
conflict reconciliation. I have implemented this framework over
Berkeley DB High Availability, a replicated database engine.
|
Past Research Projects
Dependency Management in Workflow
Management Systems
|
|
My
research on dependency management focuses on designing suitable
primitives to modeling dependencies as well as applying the
aspect-oriented programming paradigm to automate the lifecycle of
workflow specifications and executions. I have designed a unifying
framework to model four types of dependencies including data, control,
service, and cooperation dependency. These dependencies are created
either due to the parallel interactions of agents or the interactions
with remote services. In my framework, dependencies are
declaratively specified and formally analyzed to remove redundancies.
They are eventually weaved into the specification of business
processes.
|
| Elba
Project: Automated N-Tier Application Deployment |
|
Cooperated
with
HP
Labs
in
automation
of application staging. This research focuses on
translating service deployment workflow specified by a resource
management system to a target deployment language. The research
challenges include synchronization constraint description, validation,
and code weaving.
|
Work Experience
May 2008 – Aug 2008
|
Research Intern, IBM T.J.
Watson Research
Center, Hawthorne, NY, USA
Mentor: Dr. Arun K. Iyengar
Topic: Combining Quality of Service and Social Information for Ranking
Services
|
May 2006 – Aug 2006
|
Research Intern, IBM T.J.
Watson Research Center,
Hawthorne, NY, USA
Mentor: Dr. Chitra Venkatramani
Topic: a Stream Filter Algorithm in Distributed Streaming Systems
|
| May 2005 – Aug 2005 |
Intern,
Internet
advertising
applications,
Microsoft
Corporation,
Redmond,
WA,
USA
|
Teaching Experience
Aug 2003 -
Present
|
Graduate
Teaching
Assistant
CS4400: Introduction to Database Design
CS4365/6365/8803: Introduction to Enterprise Computing
CS4220/6235: Real-Time Embedded Systems
|
Research Presentations and Posters
2010
|
Conference talk at CollaborateCom'10
Title: Modeling and Implementing Collaborative Text Editing Systems
with Transactional Techniques
Slides: [PDF]
Conference talk at ICDE'10
Title: A Partial Persistent Data Structure to Support Consistent Shared
Access in Collaborative Editing Applications
Slides: [PDF]
Poster at CIKM'10
Title: Elusive Vandalism Detection at Wikipedia: A Text Stability-based
Approach
Attended by Danesh Irani
Poster: [PDF]
|
2009
|
Conference talk at ICSOC'09
Title: Quality of Service and Social Information for Ranking Services
Presented by Revathi Subramanian
Slides: [PDF]
|
|