Zusammenfassung
Modeling the mechanism of natural communication in terms of a general and computationally efficient theory has a threefold motivation in computational linguistics. Theoretically, it requires discovering how natural language actually works — surely an important problem of general interest. Methodologically, it provides a unified, functional viewpoint for developing the components of grammar on the computer and allows objective verification of the theoretical model in terms of its implementation. Practically, it serves as the basis for solid solutions in advanced applications.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Preview
Unable to display preview. Download preview PDF.
Literatur
The notion task environment was introduced by A. Newell amd; H. Simon 1972. The robot-internal representation of the task environment is called the problem space.
CURIOUS is an advanced variant of the color reader described in CoL, p. 295 ff.
This is in accordance with the approach of nouvelle AI, which proceeds on the motto The world is its own best model. See Section 1.1.
Behavior tests with humans may include the use of language by interviewing the subjects about their experience. This however, (i) introduces a subjective element and (ii) is not possible with all types of cognitive agents.
A classic treatment of artificial vision is D. Marr 1982. For a summary see J.R. Anderson 19902, p. 36 ff. More recent advances are described in the special issue of Cognition, Vol. 67, 1998, edited by M.J. Tarr amd; H.H.Bülthoff
For the sake of conceptual simplicity, the reconstructed pattern, the logical analysis, and the classification are described here as separate phases. In practice, these three aspects may be closely interrelated in an incremental procedure. For example, the analysis system may measure an angle as soon as two edges intersect, the counter for corners may be incremented each time a new corner is found, a hypothesis regarding a possible matching concept may be formed early so that the remainder of the logical analysis is used to verify this hypothesis, etc.
In the following, we will avoid the term ’(non)verbal’ as much as possible because of a possible confusion with the part of speech ‘verb.’ Instead of ‘nonverbal cognition’ we will use the term ’context-based cognition.’ Instead of ‘verbal cognition’ we will use the term language-based cognition.’
An example of a token is the actual occurrence of a sign at a certain time and a certain place, for example the now following capital letter A. The associated type, on the other hand, is the abstract structure underlying all actual and possible occurrences of this letter. Realization-dependent differences between corresponding tokens, such as size, font, place of occurrence, etc., are not part of the associated type.
Instances of the same variable in a concept must all take the same value. Strictly speaking, 3.3.2 would thus require an operator — for example a quantifier — binding the variables in its scope. We use sketchy definitions of tokens and types for the sake of simplicity and in order to avoid discussing the different advantages and disadvantages of logical versus procedural semantics (cf. Chapters 19–22).
In context-based cognition, M-concepts are matched with parameter values, resulting in I-conceptsio, during recognition and realizing I-conceptslo, during action (cf. 3.3.5). In language-based cognition, M-concepts acquire a secondary function as literal meanings attached to word surfaces. During communication, the M-concepts of language are matched with corresponding I-conceptsio, of the context (cf. 4.2.3, 23.2.1).
One may also conceive of handling M-concepts connectionistically.
Our abstract description of visual recognition is compatible with the neurological view. For example, after describing the neurochemical processing of photons in the eye and the visual cortex, D.E. Rumelhart 1977, p. 59f. writes:
For example, C.K. Ogden amd; I.A. Richards 1923 call the use of icons or images in the analysis of meaning ‘a potent instinctive belief being given from many sources’ (p. 15) which is ‘hazardous,’ ’mental luxuries,’ and ’doubtful’ (p. 59). In more recent years, the idea of iconicity has been quietly rehabilitated in the work of W. Chafe 1970, P. Johnson-Laird 1983 (cf. p. 146,7), T. Givón 1985 (cf. p. 189), J. Haiman 1985a,ó and others.
S. Palmer 1975.
C.S. Peirce 1871 writes about Berkeley:Berkeley’s metaphysical theories have at first sight an air of paradox and levity very unbecoming to a bishop. He denies the existence of matter, our ability to see distance, and the possibility of forming the simplest general conception; while he admits the existence of Platonic ideas; and argues the whole with a cleverness which every reader admits, but which few are convinced by.
The program may even be expanded to recognize bitmap outlines with imprecise or uneven contours by specifying different degrees of granularity. Cf. J.L. Austin’s 1962 example France is hexagonal.
In Chapters 23 and 24 the intuitive format of 3.4.2, which uses an arc and a dotted line to indicate intra-and extrapropositional relations, respectively, is replaced by indices in the abstract format of a network database.
For example, in the feature [loc: A2], the value A2 stands for a certain field in the task environment of CURIOUS.
For example, in the feature [loc: Mo 14:05], the value stands for Monday, five minutes after 2 p.m.
In analytic philosophy, internal parameters — such as an individual tooth ache — have been needlessly treated as a major problem because they are regarded as a `subjective’ phenomenon which allegedly must be made objective by means of indirect methods such as the double aspect theory. See in this connection the treatment of propositional attitudes in Section 20.3, especially footnote 9.
See autonomy from the metalanguage in Section 19.4.
The formulation in 3.5.2 assumes that the task environment is divided into 16 fields, named Al, A2, A3, A4, B1, B2 etc., up to D4.
Author information
Authors and Affiliations
Rights and permissions
Copyright information
© 1999 Springer-Verlag Berlin Heidelberg
About this chapter
Cite this chapter
Hausser, R. (1999). Cognitive foundation of semantics. In: Foundations of Computational Linguistics. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-662-03920-5_4
Download citation
DOI: https://doi.org/10.1007/978-3-662-03920-5_4
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-662-03922-9
Online ISBN: 978-3-662-03920-5
eBook Packages: Springer Book Archive