Program of the Tenth Annual Conference of the Cognitive Science Society: 17-19 August 1988, Montreal, Quebec, Canada

By Cognitive Science Society | Go to book overview

FLEXIBLE NATURAL LANGUAGE PROCESSING AND ROSCHIAN CATEGORY THEORY

Sandra L. Peters, Stuart C. Shapiro, and William J. Rapaport

Department of Computer Science State University of New York at Buffalo


1. INTRODUCTION.

Artificial intelligence systems typically hand-craft large amounts of knowledge in complex, static, high-level knowledge structures. These systems generally work well in very limited domains, but are simply too rigid to support natural language understanding in general. This is due in part to the fact that artificial intelligence, natural language processing (AI/NLP) systems have not taken seriously the principles of categorization first set forth in the seminal work of Rosch [ 1976, 1978, 1981], and later extended by researchers including Barsalou [ 1987], Murphy & Medin [ 1985], Lakoff [ 1987], and Neisser [ 1987]. Thus, AI/NLP systems have very shallow representations of generic concepts and categories, employing either static, simple, featural models of concepts based on necessary and sufficient criteria in uniform taxonomies where no level is distinguished [e.g., KLONE, Brachman 1983], or passive data structures with slots and explicit default values, such as frames, schemata, and scripts [e.g., KRL, Bobrow & Winograd 1977ab; NETL, Fahlman 1979]. We have previously shown that systems that use the former representations are unable to model human category systems [ Peters & Shapiro 1987ab]. In this paper, we discuss the inadequacy of systems based on the latter types of representations, arguing that frames, schemata, and scripts lack the flexibility, generality, and adaptability necessary for representing generic concepts in memory. We present alternative "active" representations, in which frames or schemata do not reside in semantic memory, but rather are constructed as needed from a less organized semantic memory. Their construction can, therefore, be influenced by the current task and context.

In addition, our processing and representations are based on a Roschian model of categories, i.e., on (1) a recognition of the unique nature of basic level categories within natural category systems, and (2) prototype theory. We will discuss some of the current research that supports our representations and processing, and show that our system's performance is enhanced by taking these principles of categorization seriously. Our implementation uses the SNePS knowledge representation and reasoning system, including a generalized ATN parsergenerator [ Shapiro 1979, 1982; Shapiro & Rapaport 1987]. We present a detailed example that shows the use of our representations and processing strategies in the task of discourse comprehension. In particular, our example will concern implicit focusing in natural language comprehension, i.e., the implicit activation of thematic associates and salient attributes of concepts during discourse comprehension.

1.1. Roschian Model: Basic Level Primacy . Psychology, linguistics, and anthropology have produced a variety of measures of perception, behavior, and communication showing a convergence of cognitive tasks at the basic level. Not all levels of a taxonomy are equally used and useful: for taxonomies of common objects and organisms, the basic level, the level of table and bird, is the most informative and useful [ Rosch et al. 1976; Berlin 1978; Tversky & Hemenway 1984]. Our knowledge is organized at this level (i.e., most attributes of category members are stored at the basic level) and visual imagery is particularly strong for basic level concepts. The informativeness of the basic level originates from the amount of knowledge stored at this level and the rich perceptual component of basic level categories. Basic level concepts trigger many reflex inferences; i.e., they routinely activate many satellite concepts. Where informativeness is the greatest, so too is inferential power.

We have previously presented a representation for natural category systems that uses distinct representations for basic and non-basic level concepts, and discussed empirical evidence supporting basic level primacy and the need for AV/NLP systems to recognize the uniqueness and importance of the basic level [ Peters & Shapiro 1987ab]. The work described in this paper builds on, and presupposes that earlier work.

1.2. Roschian Model: Prototype Theory . Although categories have been viewed traditionally as concepts established by necessary and sufficient criteria, and many Al systems continue to model natural concepts in this way [e.g., Brachman 1983], recent categorization research does not support this view [ Rosch 1976, 1978; Mervis & Rosch 1981; Murphy & Medin 1985; Barsalou 1987; Lakoff 1987; Neisser 1987]. Rosch has suggested that

-125-

Notes for this page

Add a new note
If you are trying to select text to create highlights or citations, remember that you must now click or tap on the first word, and then click or tap on the last word.
One moment ...
Default project is now your active project.
Project items
Notes
Cite this page

Cited page

Style
Citations are available only to our active members.
Buy instant access to cite pages or passages in MLA 8, MLA 7, APA and Chicago citation styles.

(Einhorn, 1992, p. 25)

(Einhorn 25)

(Einhorn 25)

1. Lois J. Einhorn, Abraham Lincoln, the Orator: Penetrating the Lincoln Legend (Westport, CT: Greenwood Press, 1992), 25, http://www.questia.com/read/27419298.

Note: primary sources have slightly different requirements for citation. Please see these guidelines for more information.

Cited page

Bookmark this page
Program of the Tenth Annual Conference of the Cognitive Science Society: 17-19 August 1988, Montreal, Quebec, Canada
Table of contents

Table of contents

Settings

Settings

Typeface
Text size Smaller Larger Reset View mode
Search within

Search within this book

Look up

Look up a word

  • Dictionary
  • Thesaurus
Please submit a word or phrase above.
Print this page

Print this page

Why can't I print more than one page at a time?

Help
Full screen
Items saved from this book
  • Bookmarks
  • Highlights & Notes
  • Citations
/ 778

matching results for page

    Questia reader help

    How to highlight and cite specific passages

    1. Click or tap the first word you want to select.
    2. Click or tap the last word you want to select, and you’ll see everything in between get selected.
    3. You’ll then get a menu of options like creating a highlight or a citation from that passage of text.

    OK, got it!

    Cited passage

    Style
    Citations are available only to our active members.
    Buy instant access to cite pages or passages in MLA 8, MLA 7, APA and Chicago citation styles.

    "Portraying himself as an honest, ordinary person helped Lincoln identify with his audiences." (Einhorn, 1992, p. 25).

    "Portraying himself as an honest, ordinary person helped Lincoln identify with his audiences." (Einhorn 25)

    "Portraying himself as an honest, ordinary person helped Lincoln identify with his audiences." (Einhorn 25)

    "Portraying himself as an honest, ordinary person helped Lincoln identify with his audiences."1

    1. Lois J. Einhorn, Abraham Lincoln, the Orator: Penetrating the Lincoln Legend (Westport, CT: Greenwood Press, 1992), 25, http://www.questia.com/read/27419298.

    Cited passage

    Thanks for trying Questia!

    Please continue trying out our research tools, but please note, full functionality is available only to our active members.

    Your work will be lost once you leave this Web page.

    Buy instant access to save your work.

    Already a member? Log in now.

    Search by... Author
    Show... All Results Primary Sources Peer-reviewed

    Oops!

    An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.