Whereas both these approaches apparently draw on the traditional rationalistic paradigm of mind-matter-duality- static the former, dynamic the latter-in presupposing the external world structure and an internal representation of it, the third and fourth category do not:the cognitive approaches presuppose the existence of the external world, structured by given objects and properties and the existence of representations of (fragments of) this world internal to the system, so that the cognitive systems' (observable) behaviour of action and reaction may be modelled by processes operating on these structures; the associative approach is described as a dynamic structuring based on the model concept of self-organization with cognitive systems constantly adapting to changing environmental conditions by modifying their internal representation of them.
the enactive approaches may be characterized as being based upon the notion of strcutural coupling. Instead of assuming an external world and the systems' internal representations of it, some unity of structural relatedness is considered to be fundamental of-and the (only) condition for-any abstracted or acquired duality in notzions of the external and internal, object and subject, reality and its experience; the semiotic approaches focus on the notion of semiosis and may be characterized by the process of enactment too, supplemented, however, by the representational impact. It is considered fundamental to the distinction of e.g. cognitive processes from their structural results which-due to the traces these processes leave behind-may emerge in some form of knowledge whose different representational modes comply with the distinction of internal or tacid knowledge (i.e. memory) on the one hand and of external or declarative knowledge (i.e. language) on the other.
According to these types of cognitive modeling, computational semiotics can be characterized as aiming at the dynamics of of meaning constitution by simulating processes of multi-resolutional representation [5] within the frame of an ecological information processing paradigm [18].
As we take human beings to be systems whose knowledge based processing of represented information makes them cognitive, and whose sign and symbol generation, manipulation, and understanding capabilities render them semiotic, we may do so due to our own daily experience of these systems' outstanding ability for representing results of cognitive processes, organize these representations, and modify them according to changing conditions and states of system-environment adaptedness.
Following this line, however, will necessitate to pass on from traditional approaches in competence oriented linguistics analysing introspectively the propositional contents of singular sentences as conceived by ideal speakers/writers towards a new understanding of meaning constitution as a dynamic process based upon the semiotic cognitive information processing the traces of which are to be identified and systematically reconstructed on the basis of empirically well founded observation and rigorous mathematical description of universal regularities that structure and constitute different levels of representations in masses of pragmatically homogeneous texts produced by real speakers/writers in actual situations of either performed or intended communicative interaction. Only such a performance oriented semiotic approach will give a chance to formally reconstruct and model procedurally both, the significance of entities and the meanings of signs as a function of a first and second order semiotic embedding relation of situations (or contexts) and of language games (or cotexts) which corresponds to the two-level actualisation of cognitive processes in language understanding [18].
In terms of the theory of information systems, texts-whether internal or external to the systems-function like virtual environments4.
the spaciotemporal identity of pairs of immediate system-environment coordinates which will let the system experience the material properties of texts as signs (i.e. by functions of physical access and mutually homomorphic appearance). These properties apply to the percepts of language structures presented to a system in a particular discourse situation, and the representational identity of pairs of mediate system-environment parameters which will let the system experience the semantic properties of texts as meanings (i.e. by functions of emergence, identification, organisation, representation of structures). These apply to the comprehension of language structures recognized by a system to form the described situation
Hence, according to the theory of information systems, functions like interpreting signs and understanding meanings translate to processes which extend the fragments of reality accesssible to a living (natural and possibly artificial) information processing system. This extension applies to both, the immediate and mediate relations a system may establish according to its own evolved adaptedness or dispositions (i.e. innate and acquired structuredness, processing capabilities, represented knowledge).
This in a way experimental quality of textual representations which increases the potentials of adaptive information processing beyond the system's lifespan, is constrained simultaneously by dynamic structures corresponding to knowledge. The built-up, employment, and modification of these structural constraints5 is controlled by procedures whose processes determine cognition and whose results constitute adaptation. Systems properly attuned to textual system-environments have acquired these structural constraints (language learning) and can perform certain operations efficiently on them (language understanding). These are prerequisites to recognizing mediate (textual) environments and to identify their need for and the systems' own ability to actualize the mutual (and trifold) relatedness constituting what Peirce called semiosis 6. Systems capable of and tuned to such knowledge-based processes of actualisation will in the sequel be referred to as semiotic cognitive information processing systems (SCIPS) [17, 19].
Representation, therefore, has to be considered fundamental to the distinction of the processes of cognition from their results which may emerge-due to the traces these processes leave behind-in some structure ( knowledge). Different representational modes of this structure not only comply with the distinction of internal or tacid knowledge (i.e. memory) on the one hand and of external or declarative knowledge (i.e. texts) on the other7, but these modes also relate to different types of formats ( distributional vs. symbolic), modeling ( connectionist vs. rule-based) and processing ( stochastic vs. deterministic). It is this range of correspondences that Fuzzy Linguistics is based upon and tries to exploit to come up with a unifying framework for most of the different approaches followed sofar.
Soft categorising appears to be a prerequisite for fuzzy linguistic modeling examples of which will illustrate the notion of dynamic structures emerging from corpora of natural language discourse.
As as some of these procedural characteristics have also be claimed by cognitive linguistic approaches and computational models of language understanding, their main traits may help to illustrate the different positions of semiotic modeling in fuzzy linguistics.
In one of the rare ventures on discussing of how cognitive, i.e. knowledge based information processing mechanisms may be provided with the knowledge bases they are meant to operate on, and how these knowledge structures may be related to observable language data, Bierwisch (1981) sketches a hierarchy of information processing mechanisms whose representational format (sets of rewrite rules operating on structured data) allows algorithms be formulated and implementations be found to guaratee their computability. According to this schema (Fig. 1) and starting with the morpho-phonological level, an information processing mechanism M1 is postulated which receives utterances as input and produces some associated structures as output. In doing so, however, the mechanism's performance will be determined not only by the external input strings but also by some internal knowledge of elements and rules which allow to agglomerate the structures identified. The acquisition and representation of this internal knowledge is hypothesized as resulting from a process M2 which also includes a multitude of rudimentary, incomplete, and tentative M1-kind processes. M2 is assumed to be a complex information processing mechanism whose inputs are corpora of utterances together with some environmental information, and whose outputs will be the grammars underying these utterances. Again, this mechanism's results will not only and completely be determined by the external inputs but also by some internal structures which are believed to control the human language faculty in a fundamental way as so-called linguistic universals. These may (or may not) be assumed to be derivable as results of an information processing mechanism M3 whose input is as comprehensive (or unspecified) as the term languages might allow.
Taking the relation of inclusion for M1 Ì M2 to hold also for M2 Ì M3, and considering M1, M2,M3 computationally specifyable procedures of language analysing processes instead of mere metaphors for some (more or less plausible) mechanisms of the human mind, then it appears reasonable to consider M3 a collection of all the processes of methodical analysis, representation, and comparison of structured sets of utterances from different languages, including the processes in M1 as a device that explicidly specifies an utterance's structure relative to a given grammar, and the processes in M2 as a system that generates a grammar from a corpus of utterances relative to the given set of universals. This modeling view allows for the notions of Universals Þ Grammar Þ Structure to be understood as variables of theoretical constructs hinged on empirical regularities observed in Languages, Corpus, Utterance respectively. Whereas the latter are external representations,the former are internal to any SCIP -system and considered external representations only under the competence linguistic approach to cognitive modeling. As such they are hypothesized to form a hierarchy of linguistic -not language -entities which formally specify a class of other linguistic entities (following the double arrows in Fig. 1).
The model theoretical and operational problems inherent in this setup concern the (non universal and highly restrictive) representational format which is assumed to enable the denotation of universals, grammar and structure, and the essentially top-down, non-recursive propagation of externally presented but internally processed results of these mechanisms. Thus, M3 whose performance in identifying universals and representing them externally depends crucially on the efficient performance of M2 which is said to employ these universals as internal procedural constraints in order to identify syntactic regularities and represent them externally in a rule based format as grammars. Grammars, in turn, have to be employed as internal procedural constraints by M1 if this mechanism's identification processes and the external representation of their findings is meant to be successful.
Distinguishing between these two kinds of structures either external or internal to the mechanisms M introduced so far, is indicative of the systems theoretical view proposed in semiotic modeling. It easily allows to translate these mechanisms as sets of procedures which allow to describe and simulate a living systems' abilility to process environmental input (external structures) according to procedural constraints known to the system (internal structures) in order to produce some results of this processing. However,it appears not at all conclusively compelling to assume that these procedural constraints and the processing results need to be represented in a rule-based format. According to an ecologically motivated systems theoretical view, systems enacting these processes under boundary conditions as determined by their surrounding environments, or their internal structuredness, or both, will have to process certain inputs to produce specified output structures. But identifying their status of being at the same time internal and external to the processing system is tantamount to the methodological dilemma which can solely be solved on the grounds of revising the representational mode and the formatting constraints which the model construction has to be decided on to allow.
Following Chomsky these modes have been restricted to abstract principles of language competence by processes [2] whose assumed rule-based determinacy consequently led to formal representations of these rules giving rise to the above model hierarchy of discrete strata [1]. In trying to relate these strata to observable performative language data structures in order to mediate observable language regularities with theoretical constructs supposedly representing principles underlying these constructs, the methodological shortcomings of the cognitive linguistic approach are revealed. It suffers from competence theoretically inspired idealisations of regularities and theoretical abstractions (like universals, grammars, sentences) whose symbolic notations and formal expressions may be scrutinized for their syntactic correctness but lack empirically observable and experimentally testable procedures of language representation which are independant from competent speakers' understanding of that language.
This is achieved by analysing the linear or syntagmatic and selective or paradigmatic constraints which natural language structure imposes on the formation of (strings of) linguistic entities on whatever level of entity formation. It has been shown and illustrated elsewhere [15], [9] in some detail, that fuzzy linguistic modeling allows to derive the representational means (e.g. soft categories, continuous gradation, variable granularity, flexible plasticity, dynamic approximation, etc.) which crisp categories and competence theoretically inspired idealisations of performative regularities lack. The (numerical) specificity and (procedural) definiteness of sub-symbolic, distributed formats in entity formation appear to provide for higher phenomenological compatibility and more cognitive adequacy than traditional levels of categorial representation whose symbolic mediation and syntactic correctness could only formally be scrutinized but not empirically or experimentally be tested [11].
For written German discourse analysed on type-setting level with m discernable types of signs (letters) and maximum lengths n of strings there are quite a number of theoretically possible (Tab. 1, col. Tn) crisp n-ary relations Tn = Zn, i.e.
Out of these, however, only those have to be computed
which are not only actually possible (col. An) but which have
indeed been observed to factually occur, i.e. Fn Í Fn-1 ×Z
(Tab 1, col. Fn), i.e.
The fuzzy relational modeling (Eqns. 3 and 2) shows that even for higher n only bi-grams have to be traced and computed due the (n-1)-ary relations computed on the previous level of representation. It is this principle of procedural self-similarity of n-ary agglomerative steps which allows for the trie-like representation [3] of entities that are labeled (by soft categorial n-relative letter transitions) and are an outcome of procedural constraints (over n levels of processing) which produce a dynamically structured system of fuzzy relations that depicts the overall transition tendencies of signs. For the letter Z this structure is given in Fig. 4 illustrating sub-regularities of morphic word formation. bzw. des Verständnisses dies erzwingt.
The core of the representational formalism can be characterized as a two-level process of abstractions. The first (called a-abstraction) on the set of fuzzy subsets of the vocabulary provides the word-types' usage regularities or corpus points, the second (called d-abstraction) on this set of fuzzy subsets of corpus points provides the corresponding meaning points as a function of all differences of all usage regularities which a set of word-types may produce by its word-tokens' frequencies as observed in pragmatically homogeneous corpora of natural language texts.
The basically descriptive statistics to specify intensities of
co-occurring lexical items in texts is centred around the
correlational measure
To specify these correlational value distributions' differences,
a measure of similarity (or rather, dissimilarity) is used
The consecutive application of (Eqns. 7) on input texts and (Eqns. 9) on the output data of (Eqns. 7) allows to model the meanings of words as a function of differences of usage regularities (Fig. 6).
Thus, ai,j allows to express pairwise relatedness of word-types (xi,xj) Î V ×V in numerical values ranging from -1 to +1 by calculating co-occurring word-token frequencies (Eqn. 5) for pairs of items.
As a fuzzy binary relation, [(a)\tilde] : V×V® I can be conditioned on xn Î V which yields a crisp mapping
Considering C as representational structure of abstract entities constituted by syntagmatic regularities of word-token occurrences in pragmatically homogeneous discourse, then the similarities and/or dissimilarities of these entities will capture their corresponding word-types' paradigmatic regularities calculated by d Eqn. 6 serving as second mapping function, As a fuzzy binary relation, [(d)\tilde] : C ×C® I can be conditioned on yn Î C which again yields a crisp mapping
Identifying zn Î S with the numerically specified elements of potential paradigms, the set of possible combinations S ×S may structurally be constrained and evaluated without (direct or indirect) recourse to any pre-existent external world. Introducing a Euclidian metric
Weighted numerically as a function of an element's distance values and its associated node's level and position in the tree, Cr(zi) either is an expression of the head-node's zi meaning-dependencies on the daughter-nodes zn or, inversely, expresses their meaning-criterialities adding up to an aspect's interpretation determined by that head [15]. To illustrate the feasibility of the D-operation's generative procedure, the substructure of relevant constraints (related meaning points) DDS(zi) Í áS,zñ anchored with the lexical item xi, i = COMPUTER is shown in Fig. 2.
- [1]
- M. Bierwisch: Logische und psychologische Determinanten der Struktur natürlicher Sprachen. In: J. Scharf (Ed): Naturwissenschaftliche Linguistik, Nova Acta Leopoldina, NF 54, Nr.245, pp. 176-187; 841-851 (J.A. Barth) Halle, 1981.
- [2]
- N. Chomsky: The Logical Structure of Linguistic Theory. (Plenum) New York, 1975.
- [3]
- D. E. Knuth: The Art of Computer Programming. Vol.3, (Addison-Wesley) Reading, MA, 1973.
- [4]
- D. Marr: Vision. (Freeman) SanFrancisco, 1982.
- [5]
- A. Meystel: Semiotic Modeling and Situation Analysis: an Introduction. (AdRem Inc) Bala Cynwyd, PA, 1995.
- [6]
- C. Peirce: Pragmatism in Retrospect: A Last Formulation. In: J. Buchler (Ed): The Philosophical Writings of Peirce, pp. 269-289. (Dover) New York, 1906.
- [7]
- B. Rieger: Bedeutungskonstitution. Einige Bemerkungen zur semiotischen Problematik eines linguistischen Problems. Zeitschrift für Literaturwissenschaft und Linguistik, (27/28):55-68, 1977.
- [8]
- B. Rieger: Unscharfe Semantik. Die empirische Analyse, quantitative Beschreibung, formale Repräsentation und prozedurale Modellierung vager Wortbedeutungen in Texten. (Lang) Frankfurt a.Main/ Bern/ Paris, 1989.
- [9]
- B. Rieger: Unscharfe Semantik: zur numerischen Modellierung vager Bedeutungen von Wörtern als fuzzy Mengen. In: H. Friemel, et.al. (Eds): Forum-90: Wissenschaft und Technik. (Informatik-Fachberichte 259), pp. 80-104. (Springer) Berlin/ Heildeberg/ NewYork, 1990.
- [10]
- B. Rieger: Fuzzy Modellierung linguistischer Kategorien. In: W. Feldmann, H. Hinrichs (Eds): Text und Lexikon, [in print] (Niemeyer) Tübingen, 1996.
- [11]
- B. Rieger: Warum Fuzzy Linguistik? Überlegungen und Ansätze einer computerlinguistischen Neuorientierung. In: D. Krallmann, W. Schmitz (Eds): Intern. Gerold Ungeheuer Symposium, [in print] (Nodus) Münster, 1996.
- [12]
- B. B. Rieger: Feasible Fuzzy Semantics. On some problems of how to handle word meaning empirically. In: H. Eikmeyer, H. Rieser (Eds): Words, Worlds, and Contexts. New Approaches in Word Semantics, pp. 193-209. (de Gruyter) Berlin/ New York, 1981.
- [13]
- B. B. Rieger: Fuzzy Representation Systems in Linguistic Semantics. In: R. Trappl, et.al. (Eds): Progress in Cybernetics and Systems Research, Vol. XI, pp 249-256. (McGraw-Hill), Washington/ New York/ London, 1982.
- [14]
- B. B. Rieger: Lexical Relevance and Semantic Disposition. On stereotype word meaning representation in procedural semantics. In: G. Hoppenbrouwes, et.al. (Eds): Meaning and the Lexicon, pp 387-400. (Foris), Dordrecht, 1985.
- [15]
- B. B. Rieger: Distributed Semantic Representation of Word Meanings. In: J. D. Becker, et.al. (Eds): Parallelism, Learning, Evolution. Evolutionary Models and Strategies WOPPLOT-89, [Lecture Notes in Artificial Intelligence 565], pp. 243-273. (Springer) Berlin/ Heidelberg/ New York, 1991.
- [16]
- B. B. Rieger: On Distributed Representation in Word Semantics. ICSI-TR-91-012, (ICSI), Berkeley, CA, 1991.
- [17]
- B. B. Rieger: Meaning Acquisition by SCIPS. In: B. M. Ayyub (Ed): ISUMA-NAFIPS-95, [IEEE-Transactions: Joint Intern. Conf. on Uncertainty Modeling and Analysis, North American Fuzzy Information Processing Society], pp. 390-395, (IEEE Computer Society Press), Los Alamitos, CA, 1995.
- [18]
- B. B. Rieger: Situation Semantics and Computational Linguistics: towards Informational Ecology. A semiotic perspective for cognitive information processing systems. In: K. Kornwachs, K. Jacoby (Eds): Information. New Questions to a Multidisciplinary Concept, pp 285-315, (Akademie), Berlin, 1995.
- [19]
- B. B. Rieger: Situations, Language Games, and SCIPS. Modeling semiotic cognitive information processing systems. In: A. Meystel, N. Nerode (Eds): editors, Architectures for Semiotic Modeling and Situation Analysis in Large Complex Systems, pp. 130-138, (AdRem), Bala Cynwyd, PA, 1995.
- [20]
- B. B. Rieger / C. Thiopoulos: Semiotic Dynamics: a self-organizing lexical system in hypertext. In: R. Köhler, B. Rieger (Eds): Contributions to Quantitative Linguistics. Proceedings of the 1st Quantitative Linguistics Conference - QUALICO-91, pp. 67-78. (Kluwer Academic) Publishers, Dordrecht, 1993.
- [21]
- H. A. Simon: The Sciences of the Artificial. (MIT Press), Cambridge, MA, 1982.
- [22]
- F. Varela / E. Thompson / E. Rosch: The Embodies Mind. Cognitive Science and Human Experience. (MIT Press), Cambridge, MA, 1991.
- [23]
- L. Zadeh (Ed): Fuzzy Sets and their Application to Cognitive and Decision Processes. (Academic Press), New York/ San Francisco, 1975.
- [24]
- L. A. Zadeh: Fuzzy Sets. Information and Control, 17(8):338-353, 1965.
- [25]
- L. A. Zadeh: Fuzzy Logic, Neural Networks, and Soft Computing. Comm. of the ACM, 37(3):77-84, 1994.
1Published in: Meystel, A. (Ed.): A Learning Perspective. Proceedings of the 1997 International Conference on Intelligent Systems and Semiotics (ISAS-97) [NIST Special Publication 918], Washington (US Government Printing Office) 1997, pp. 541-551; also in: Albus, J./Meystel, A. (Eds.): Intelligent Systems: A Semiotic Perspective (Vol. I: Formation of Symbols), Washington/New York/London (Wiley) 1997 [in print]
2There were only the first three of these four approaches distinguished by Varela/Thompson/Rosch (1991).
3According to standard theory there is no direct genetic coding of experiencial results but rather indirect transmission of them by selectional advantages which organisms with certain genetic mutations gain over others without them to survive under changing environmental conditions.
4Simon's (1982) remark ''There is a certain arbitrariness in drawing the boundary between inner and outer environments of artificial systems. ... Long-term memory operates like a second environment, parallel to the environment sensed through eyes and ears'' (pp. 104) is not a case in point here. Primarily concerned with where to place the boundary, he does not seem to see its placing in need to be justified or derived as a consequence of some possibly representational processes we call semiotic. As will become clear in what follows, Simon's distinction of inner (memory structure) and outer (world structure) environments is not concerned with the special quality of language signs whose twofold environmental embedding (textual structure) cuts accross that distinction, resolving both in becoming representational for each other.
5What Simon (1982) calls memory in accordance with his questioning of the inner-outer-distiction of cognitive systems and their environments.
6By semiosis I mean [... ] an action, or influence, which is, or involves, a coöperation of three subjects, such as sign, its object, and its interpretant, this tri-relative influence not being in any way resolvable into actions between pairs. (Peirce 1906, p. 282)
7Whereas tacid knowledge cannot be represented other than by the immediate system-environments' corresponding states, explicit knowledge is bound to acquire some formal properties in order to become externally presented and thereby part of mediate system-environments. Natural languages obviously provide these formal properties-as partly identified by research in linguistic competence (principles knowledge and acquisition of language)-whose enactment-as investigated in studies on natural language performance (production and understanding of texts)-draws cognitively on both bases of (explicit and tacid) knowledge.
8The Trier dpa -Corpus for instance comprises the complete textual materials from the so-called basic news real service of 1990-1993 (720.000 documents) which the Deutschen Presseagentur (dpa), Hamburg, deserves thanks to have the author provided with for research purposes. After deletion of editing commands the Trier-dpa -Corpus consists of approx. 180 Mio. (18 ·107) running words (tokens) for which an automatic tagging and lemmatising tool is under development. It is this corpus which provides the performative data of written language use for the current (and planned) fuzzy -linguistic projects at the our department.
9In subscribing to a structuralistic view of natural languages, the distinction of langue-parole and competence-performance in modern linguistics allowes for different levels of language description and linguistic analysis. Being able to segment strings of language discourse and to categorize types of linguistic entities, however, is but making analytical use of structural couplings presented by natural language discourse to semiotic systems properly attuned.