Postscript Version

Centering Theory in Discourse Processing

Barbara J. Grosz

Division of Engineering and Applied Sciences
Harvard University

CONTACT INFORMATION

Engineering Sciences Laboratory
40 Oxford Street, Room 413
Cambridge, MA 02138
Phone: (617) 495-3673
Fax : (617) 496-1066
Email: grosz@eecs.harvard.edu

WWW PAGE

http://www.eecs.harvard.edu/grosz

PROGRAM AREA

Speech and Natural Language Understanding

KEYWORDS

Focus
Attention
Reference
Pronouns

PROJECT SUMMARY

In a theory of discourse structure, developed with Sidner (Grosz and Sidner, 1986), we distinguish among three components of discourse structure: linguistic structure, intentional structure, and attentional state. Linguistic structure groups utterances into discourse segments. Intentional structure consists of discourse segment purposes and the relations between them. Attentional state, an abstraction of the discourse participants' focus of attention, records the objects, properties, and relations that are salient at a given point in the discourse. Our current research addresses open problems in all three components; this project focuses on interactions between the linguistic structure and attentional state.

There are two levels of attentional state. The global level is concerned with the relations between discourse segments and the ways in which attention shifts between them; it depends on the intentional structure. The local level is concerned with changes of attention within discourse segments. Centering (Grosz, Joshi and Weinstein, 1995), an element of the local level, pertains to the interaction between the form of linguistic expression and local discourse coherence. In particular, it relates local coherence to choice of referring expression (pronouns in contrast to definite description or proper name), and argues that differences in coherence correspond in part to the different demands for inference made by different types of referring expressions, given a particular attentional state.

This project is concerned with empirical validation of centering theory, with refining the theory to reflect the results of empirical inquiries, and with designing algorithms that embody the theory. In earlier work, we empirically validated several basic claims of the original formulation of centering, and proposed refinements to others. Our current research addresses four central open problems that remain: (1) integration of centering with other processes for interpreting pronouns; (2) application of centering to additional context-dependent linguistic forms (e.g. ellipsis, accentuation); (3) the interaction between centering and discourse segmentation; (4) formulation of computationally tractable rules constraining center transitions. In addressing these issues we are also investigating how to incorporate centering into a general architecture for intra-segment discourse processing. The project is interdisciplinary: it includes the development of algorithms that embody the theory and psychological investigations to determine constraints on parameters of the theory.

We have completed several additional studies concerned with the role of centering in the interpretation of pronouns and have made substantial progress in the study of the role of centering in the interpretation of definite descriptions. In addition, we designed a testbed system for generating short paragraphs using centering rules to constrain the syntactic structure and referring forms used.

More specifically, we have

  • examined the extent to which centering provides a useful framework for analyzing intrasentential constraints on the interpretation of pronouns and names (Gordon and Hendrick, 1997) and the effect of the structure of complex noun phrases (NPs) on coreference by names and pronouns.
  • studied the role of centering in the interpretation of definite expressions. This empirical research involved examining patterns of acceptable coreference involving definite NPs (and also quantified expressions). The results show that patterns of coreference involving such expressions closely match those we had observed for names, and consequently are consistent with tenets of centering theory. A paper describing this work has been submitted for publication (Gordon and Hendrick, 1996).
  • examined the patterns of eye-movements that are observed in reading pronouns and names (Kennison and Gordon, in press); results show that the incoherence caused by failing to refer to the backward-looking center with a pronoun causes readers to engage in regressive eye-movements (movements back to earlier portions of the text).
  • studied the interaction of centering theory with coherence-based approaches to pronoun resolution; this study was done in the context of a larger study of the use of discourse inference in the interpretation of a range of cohesive forms including verb-phrase ellipsis and tense as well as pronouns (Kehler, 1995)
  • taken the results of the psycholinguistic studies and begun implementation of a centering algorithm based on them; this implementation is part of a language generation system that will be used to demonstrate the results of this project (Bodis, 1996).

    PROJECT REFERENCES

    Andrew Kehler, ``Interpreting Cohesive Forms in the Context of Discourse Inference.'' PhD Thesis, Harvard University, Division of Engineering and Applied Sciences (1995).

    A. Bodis, ``Generating Locally Coherent Discourse: A System Implementation.'' Harvard College, Senior Honors Thesis, April 8, 1996.

    P. C. Gordon and R. Hendrick, ``Intuitive Knowledge of Indefinite NP Anaphora.'' Manuscript, (1996).

    B. J. Grosz and C. Sidner, ``Lost Intuitions and Forgotten Intentions." Centering in Discourse, Marilyn A. Walker, Aravind K. Joshi and Ellen F. Prince, eds., [Oxford University Press] 1997.

    B. J. Grosz and Y. Ziv, ``Centering, Global Focus, and Right Dislocation." Centering in Discourse, Marilyn A. Walker, Aravind K. Joshi and Ellen F. Prince, eds., [Oxford University Press] 1997.

    P. C. Gordon and R. Hendrick, ``Intuitive Knowledge of Linguistic Coreference.'' Cognition, Vol. 62 (1997) 325-370.

    S. M. Kennison and P. C. Gordon, "Comprehending Referential Expressions During Reading: Evidence from Eye Tracking." Discourse Processes (in press).

    AREA BACKGROUND

    Centering theory attempts to relate focus of attention, choice of referring expression, and perceived coherence of utterances within a discourse segment. It fits within a larger effort to provide an overall theory of discourse structure and meaning. An understanding of the ways in which focus of attention affects both the production and the understanding of various linguistic expressions in discourse is crucial for the development of natural-language processing systems as well as of theoretical import across the cognitive sciences.

    AREA REFERENCES

    J. Allen, Natural Language Understanding. [Benjamin/Cummings Publishing Company, Inc., Reading, MA] 1995.

    P. C. Gordon, B. J. Grosz and L. A. Gillom, ``Pronouns, Names and the Centering of Attention in Discourse." Cognitive Science, 17:3 (1993) 311-347.

    Barbara J. Grosz, Karen Sparck Jones and Bonnie Lynn Webber, Readings in Natural Language Processing [Morgan Kaufmann Publishers, Inc., Los Altos, CA] 1986.

    Barbara Grosz, Aravind Joshi, and Scott Weinstein, ``Centering: A Framework for Modelling the Local Coherence of Discourse.'' Computational Linguistics, 2(21), June (1995).

    B. J. Grosz, M. Pollack and C. Sidner, ``Computational Models of Discourse.'' Foundations of Cognitive Science, Michael Posner, ed. [MIT Press, Bradford Books] 1989.

    F.C. Pereira, and B. J. Grosz, eds., Natural Language Processing [The MIT Press, Cambridge] 1994.

    RELATED PROGRAM AREAS

    Other Communication Modalities
    Adaptive Human Interfaces
    Intelligent Interactive Systems for Persons with Disabilities