Department of Cognitive Science
Johns Hopkins University
3400 N. Charles Street
Baltimore, MD 21218
Research Area
Cognitive Science of Language.
The relationships between representation, computation, and computational architectures.
Research Interests
I am currently working on my dissertation under the supervision of Paul Smolensky.
The dissertation explores Map Seeking Circuits as a computational
architecture suitable for bridging different levels of theory in
Cognitive Science. In this context, I design and implement simple
models of human speech perception concerned with how discrete,
categorical linguistic objects are perceived from an infinitely
gradient and variable acoustic stream. Specifically, I am looking at
the perception of dialect variants of English vowels. I am
particularly interested in how high-level theories of the organization
of mental representations and processes evolve when faced with the
issues that arise when these theories are fleshed out and implemented
in a well-defined and interpretable computational architecture.
Please contact me if you are interested in reading the proposal.
Some of my previous projects include:
I began my Hopkins career looking at how recursive syntactic processing may be instantiated in neural networks.
My first qualifying project
(under the supervision of Bob Frank) was
an examination of how a specific type of neural network could instantiate novel
symbolic models of computation called Deque automa, and how these could be used
to compute the various dependencies found in syntax. Predictions
about how computation breaks down when the limits of these networks are reached was compared
to psycholinguistic data from German and Dutch.
I briefly collaborated on a project initiated by Melanie Soderstrom, Paul Smolensky, and Don Mathis, in which an "abstract genome" is specified that is capable of growing abstract neural networks with layers of interconnections and cell types that allow it to (approximately) implement an OT system for a CV theory of syllabification.
With Bill Badecker, I conducted
an investigation of the modularity of linguistic processes involved
in pronoun resolution. This involves psycholinguistic experiments
designed to detect interactions between syntactic influences
attributed to discourse structure, and semantic influences of verbs
described as Implicit Causality effects, on pronoun referent
resolution. Our results were inconclusive, but consistent with a
model in which discourse structure has a strong early influence on
pronoun resolution, while the semantic factors associated with implicit
causality verbs are integrated later in the comprehension process.
An unpublished manuscript describing the work completed to this point is available here (pdf).
Along similar lines, but focusing on linguistic theory rather than experimental investigation, I participated in a collaboration with Adam Buchwald, Amanda Seidl, and Paul Smolensky on an attempt to extend the Optimality Theory framework to examine the interfaces between syntax, pragmatics, and prosodics using a bidirectional architecture.
We called this Recoverability Optimality Theory because the essential notion is that some conception of the recoverability of the meaning of an utterance has an effect on the speaker's choice of the form for that utterance.
We focus on how speakers choose what form to use for referring expressions cross-linguistically.
Blutner, Reinhard, Petra Hendriks, Helen de Hoop, and Oren Schwartz. (2004). When Compositionality Fails to Predict Systematicity. In Simon D. Levy and Ross Gayler Program Cochairs, Compositional Connectionism in Cognitive Science, (Tech Report of AAAI Fall Symposium), AAAI Press, 2004.
Schwartz, Oren and Ciprian Chelba. (1998). Chapter 4: Chelba-Schwartz's Parser.
Center for Language and Speech Processing Summer Workshop 1998 Core Natural Language Processing Technology Applicable to Multiple Languages:
Final Report,
CLSP Tech Report,
Johns Hopkins University, 1998.
Posters and Presentations
2004 - Schwartz, Oren. When Compositionality Fails to Predict Systematicity. Presented at AAAI Fall Symposium, Compositional Connectionism, Washington DC, October 24.
2004 - Schwartz, Oren. CVNet - on Neural Network Interpretations of OT. Presented at ILLC-Day 2: Language, Bonn, Germany, June 7.
2004 - Schwartz, Oren. Recoverability Optimality Theory as an embodied bidirectional model. Presented at Szklarska Poreba Workshop 5, Poland, Feb. 27 to Mar. 1.
2003 - Badecker, William and Oren Schwartz. Implicit Causality and Focusing in Pronominal Reference.
44th Annual Meeting of the Psychonomic Society, Vancouver, November 6-9.
2002 - Buchwald, Adam, Oren Schwartz, Amanda Seidl and Paul Smolensky. Recoverability Optimality Theory: Discourse Anaphora in a Bidirectional Framework. Paper presented at 6th annual Workshop in Optimality Theoretic Syntax (WOTS-6), Potsdam University, October 18-19, 2002. (Paper presented by first two authors)
2002 - Buchwald, Adam, Oren Schwartz, Amanda Seidl and Paul Smolensky. Discourse Anaphora and Bidirectional Optimization: Recoverability Optimality Theory. Paper presented at EDILOG 2002, University of Edinburgh, September 4-6 2002. (Paper presented by first two authors)
WOTS-6:
SIXTH WORKSHOP ON OPTIMALITY THEORY SYNTAX
October 18-19, 2002
EDILOG:
6th workshop on the semantics and pragmatics of dialogue
The University of Edinburgh, Sept 4th-6th, 2002
ACL-02:
Association for Computational Linguistics
40th Anniversary Meeting
July 6 - 12, 2002
ESSLLI 2001:
13th European Summer School in Logic, Language and Information
HELSINKI, FINLAND
13-24 August, 2001
CLSP WS98:
Center for Language and Speech Processing Summer Workshop 1998:
Core Natural Language Processing Technology Applicable to Multiple Languages
Baltimore, MD
June 29 - August 21, 1998
