SLP803 Controlled Languages: Anaphora

Controlled Languages: An Introduction Back|Home|Next

The Role of Linguistic Knowledge in Controlled Languages

5.0 Anaphora and Controlled Languages
5.0 What is Anaphora?
    5.1 Some Examples
    5.2 Definitions in the Literature
    5.3 Anaphora Resolution
    5.4 Anaphora Resolution in Controlled Languages

5.0 What is Anaphora?

5.1 Some Examples
In simple terms, anaphora are generally thought of as pronoun references to objects which follow the first appearance of a noun. In each of the examples below, there is some ambiguity about to what the pronoun in the second half of the sentence refers.

I'm having so many problems with the car trouble, I'll have to take it in.       take in the car, or the trouble?
Ford was in trouble, and he knew it. [Hobbs77]                                       knew trouble?
The Queen sputters a little when she speaks.[Hirst81]                               sputters when who speaks? the Queen, or someone else?
After you fix the fixture with the tool, return it.                                           return what? return the fixture, or return the tool?
My daughters played with their cousins until they were tired.                      untiil who was tired? my daughters, or their cousins?

The clarification of these ambiguities is called resolution. Strictly speaking, the referent of an anaphor is the real-world entity it specifies, and the antecedent is the textual item through which the reference is made [Hirst81].

5.2 Definitions in the Literature
There are a number of different definitions of anaphora. In "Varieties of Anaphora", Krahmer and Piwek [Krahmer00] supply several:

"The term anaphora is used most commonly in theoretical linguistics to denote any case where two nominal expressions are assigned the same referential value or range." [Reinhart99]
"… the special case of cohesion where the meaning (sense and/or reference) of one item in a cohesive relationship (the anaphor) is, in isolation, somehow vague or incomplete, and can only properly be interpreted by considering the meaning of the other item(s) in the relationship (the antecedents)." [Carter87]

Krahmer and Piwek themselves propose the following as characteristic properties of anaphora:

anaphora involve dependency on context for interpretation

anaphora depend on antecedents for interpretation

antecedents generally precede anaphora (but not always)

an anaphor and its antecedent are coreferential as discourse representations [Krahmer00]

5.3 Anaphora Resolution
There are a number of approaches to anaphora resolution. Hirst [Hirst81] sums them up as:

1) Concept activatedness (Kantor) -- an examination of the factors affecting the pronominalizability of a concept
2) Task oriented dialogues (Grosz) -- using a priori knowledge of discourse structure to resolved references
3) Frames as focus (Sidner) -- using discourse clues to choose a frame from a knowledge structure to act as focus
4) Logical formalism (Webber) -- choosing a predicate calculus like representation to handle problems such as quantification in reference resolution
5) Discourse cohesion (Hobbs, Lockman, and others) -- building a focus and resolving reference by discovering the cohesive ties in a text

5.4 Anaphora Resolution in Controlled Languages
Anaphora resolution is a critical issue in controlled languages, since one of the primary objectives of a controlled language is ambiguity reduction. Controlled language tools take a variety of approaches to anaphora resolution, generally depending on the level of sophistication of the tool. Some checkers employ a parser which is able to provide structural analysis of anaphora. Other may be limited to lexically based pattern matching.

The KANT system (Caterpillar Functional English) uses its 'deep structure grammar checker' to check ambiguous attachments and long-distance dependencies, both of which may include anaphora. However, Nyberg, et al. indicate that there is lesser need for anaphora resolution in the checking of technical documentation, as anaphora-related issues are primarily handled by the "limited use of pronouns and relative clauses" .[Nyberg96]

The SECC controlled language tool does not use a parser, and does not resolve anaphora [Bustamente00].

One way in which the Attempto ACE tool [Schwertel00] limits the complexity of anaphora relations is by limiting sentence complexity. Attempto does allow the existence of anaphora, but insists on formalized prolog-interpretable relations between and among objects in its checking tools, including anaphora. Attempto adheres to recency rules to define referents, defining the anaphoric references as the most recent accessible noun phrase that agrees in number and gender .

One of the glaring issues in controlled language tool checking with respect to anaphora is that the majority of such tools function only on an intra-sentential basis. A tool which cannot "look back" to prior text will be wholly incapable of identifying anaphora referents in prior text, a fairly common occurrence [Clemencin96].

Back|Home|Next

Richard Altwarg
Macquarie University Graduate Program in Speech and Language Processing
SLP803 An Introduction to Language Technology

This site last updated November 20, 2000.
Comments and corrections welcome: raltwarg@earthlink.com