Nathanael Chambers

PhD Candidate
Natural Language Processing Group

Research Interests

Narrative Schemas
My research focuses on learning rich event structures called narrative schemas. Narrative schemas represent event relations, their participants, and semantic roles of the arguments. It is not dissimilar from Schankian scripts or common sense sequences of situations and events. My work is the first to utilize coreference and entity chains to learn a new type of relation based on narrative/discourse coherence.

Temporal Relations
I have several published papers on learning to temporally order events described in text. My supervised learning approaches learn from labelled corpora and I have transferred orderings to order the events in narrative schemas.

Dialogue
Before coming to Stanford, I published several pieces of work on language generation, dialogue, ontologies, and the semantic interpretation of web objects.

Publications (view all 19 with downloads)

Unsupervised Learning of Narrative Schemas and their Participants

ACL 2009

Jointly Combining Implicit Constraints Improves Temporal Ordering

EMNLP 2008

Unsupervised Learning of Narrative Event Chains

ACL 2008

Classifying Temporal Relations Between Events.

ACL 2007

Learning Alignments and Leveraging Natural Logic.

ACL 2007

Using Semantics to Identify Web Objects.

AAAI 2006

One-Shot Procedure Learning from Instruction and Observation.

FLAIRS 2006

Chester: Towards a Personal Medical Advisor.

Journal Biomedical Informatics

view all 19 with downloads

Teaching

Natural Language Processing (cs224n, Spring 2008-9)
Natural Language Understanding (cs224u, Fall 2007-8)

Academic Service

Board on Judicial Affairs (2009-10), Graduate Representative

Judicial Affairs Panel (2008-10), Graduate Panelist

Committee on Research (2008-09), Graduate Representative

Committee on TA Oversight (2008-09), Graduate Representative

Freshman Mentor (2007-08), met weekly with ~15 freshmen

Previous Research Affiliations

[2008 Sum] Google Research
Information Extraction, Relation Learning

[2003-2006] Florida Institute for Human and Machine Cognition
Dialogue systems, semantic web objects, ontology mapping, language generation

[2002 Sum] Lockheed Martin ATL
Language Processing, Dialogue Systems

[1998-2003] University of Rochester
M.S. Computer Science: Statistical Language Generation




WeTheTeachers .com

I created a website in 2005 that encourages teacher collaboration and sharing of resources.
The site was sold to VIP Tone, Inc. in 2008.

School Boards

I created a campaign website: www.chambersforgv.com. My father, Bruce Chambers, is running for the school board in my old high school's district. He continually encouraged and supported me all the way to Stanford...hopefully the Great Valley community recognizes his talents. He's questioned their status quo, and is now being attacked for having the gumption to ask those questions. I credit him with motivating me to pursue my education this far, he loves students, and is honest in everything he does.

Brian Regan and Script Learning

[Observation]

People that play whole game get a whole snowcone.

[Observation]

And people that play half game get a whole snowcone.

[Conclusion]

So it's always whole, always whole snowcone.

[Goal]

So I'd rather play half game.

[Inference]

I'd rather play half game...still get the whole snowcone.

Old Stuff

Kalt Rage Quiz - How Kalt are you?