Національна Академія Мистецтв України Інститут культурології

Although appositive (listing, enumerative) representation eliminates the motivational filament encircling predicates it is here to stress that the transformation of textual segments into thesauri and the removal of predication do by no means entail randomization. In opposite to the seeming randomness thesaurus serves to disclose latent motivational ties providing textual coherence and lexical compatibility. As the result of textual self-interpretative procedures it reveals the order of taxonomic classes concealed and latently present within textual borders. All transformations of text into listing structure (indices) come to textual compilation that can be conceived as the disclosure of textual paradigmatic (and subsequently numerical) aspects and of textual taxonomy in broader sense. Thus it goes again about invisible satellite permanently accompanying text and present therein as its possibilities. This latent virtual taxonomic and numeric aspect of text gives grounds against any conjecture as to the randomization (taken also as the opposite to redundancy) of text in its accompanying glossary. The text being transformed into a listing structure, it doesn’t become random enumeration as well as the entries of compilation’s indices don’t get similarity to the row of occasional numbers.

There is still another reason for the preference to thesaurus in textual description and data representation in general. It is here to remind that textual quality of coherence is based upon the manifestation of inference (syllogism). This inferential nature of text predestinates its irreducibility to a set of propositions and the inapplicability of predicates’ calculus to its description. It is in particular to stress the properties of density and continuity that impart to thesaurus the capacity of displaying the function of mediation. Between each pair of separate elements of thesaurus a third can be inserted as the necessary intermediary element that would motivate their confrontation. Be such element absent in the text represented with the thesaurus it will be found in the wider lexicon that this thesaurus belongs to. For example one can find between “to go” and “to memorize” an intermediary element “the way” while supposing one to retain in memory the wayside details during the movement along the way. Thus the very formation of thesaurus poses the question of transgressing textual borders and supplementing the proper textual lexical stuff with the outer elements. Respectively already the compilation as the interpretative procedure includes commenting textual data with the elements taken from an observer’s experience & competence. Then compilation can be said to presuppose always the procedure of anamnesis. Therefore textual description always is initiated with the corollaries produced by an observer. It is no need to remind that anamnesis presupposes always amnesia to erase textual chunks (exclusion or elimination being the constant satellite of selective processes). Each separate text may be said to become the result of the amnesia of other opportunities. In this respect competence is comparable to the image of the Bottom Book («голубина книга» in the sense of «глибинна книга»). One can say of inclusions & exclusions as the initial steps at each text’s exploration with the subsequent procedures of data accumulation & elimination. This case is to be discerned from usual informational excess vs. deficit or pleonasm vs. ellipsis as it goes here about mental experiments with the text.

This procedure of reminding and supplementing textual elements with those missing and taken from memory gains still more importance with summarizing a text for its convolution and compression. As it has already been mentioned such procedure represents one side of the textual vital “breathing” where growth vs. shrinkage, expansion vs. compression alternate. To cope with compressing and summarizing problem one should bear in mind such an obvious triviality that the limit of each summary is a title. It is title that signifies the ultimate border for folding a text and for conceiving the respective categorical or typical situation as its prototype. The selection of the title for a work can’t be arbitrary or relative; it needs special substantiation and represents preponderated interpretative decision: one would only imagine replacing “Othello” with “Iago”. Respectively title is identical also with the ultimate limit of each textual paragraph’s compression (that’s of a strophe, of a segment, of a passage). Then the whole textual description can become the disclosure and interpretation of a title’s folded contents. Title is selected from the designations of the textual passage’s key detail as a prototype in preliminary scanning & skimming the text so that the choice of title reveals the interpretation of the passage.

Thus it is the intitulation that is to be regarded as the ultimate degree of textual compression and the first task for textual interpretation. This task is too often regarded as trivial and becomes ignored; meanwhile the very approach to intitulation becomes the complicated interpretative problem. Titles also are not to be blended with what is recently called concepts. Rather they designate riddles and puzzles to be solved: N.V. Gogol’s “Inspector” implies the latent inferences to be guessed and not the depiction of the respective serviceman as such. As a puzzle this title is to be compared with what is known as emblem so that the whole narration can be said to represent a title’s interpretation giving the sought solution. Thus title initiates interpretation as generative process in evolving lexical attraction. In this respect it serves also as the device for semantic condensation and accumulation that is disclosed in such generative process. In particular title becomes also the initial mnemonic prompt. It is to select and substantiate preference for the title that could be defined as the genuine one and would correspond to adequate interpretation. That depends upon the problem to be found out where the title would become its designation. Intitulation is also the limit for a textual indexation procedure as the core of compilation: in a way one can entitle each simple sentence in converting it in a nominative sentence so that an index appears that will take into account the utmost density of text with entitling the slightest details. There can be a set of possible titles for each passage, and they can be regarded as a kind of partitive synonyms. In its turn the entries of the compiled textual index can be regarded as virtual titles. It ensues evidently that compressions build a kind of invisible satellite permanently accompanying the text itself. Such satellites can be found in proverbs as moral conclusions for fables’ epic narratives. In particular proverb can be regarded as the compression of a narrative plot entailing its development from the compressed state. Besides, there is always such invisible satellite of a text represented in its convolution as a plot.

These procedures of compressing data in a summary presuppose the circumstance that entails the opportunities of the reverse movement of expanding the obtained data. Meanwhile such opportunities are not determined with one – to - one correspondence. While excluding (eliminating) a part of information within te procedure of compressing a text one obtains at least the two kinds of omitted information: at one side, it is textual latencies to be disclosed and reinterpreted, and at another side it is lacunas to be refilled. Respectively one can say of intensive and extensive procedures of textual expansion that are associated with generation of textual organic continuation or mechanical additions (insertions).

Here again the fundamental antinomy cone into play as far as the problem of reproducibility arises together with the task of regenerating a text from the summary (as well as of memorizing the whole from a prompt). The very opportunity of such procedure implies the conjecture about the existence of a built-in algorithm within summary (“internally stored program”) that would enable expanding a text from a summary and respectively about summary as the element of code vs. text. The existence of the codified structures (as satellite programs) in each text can be most evidently demonstrated in the constantly returning syntactic structures of sentences with their “subject – predicate – complement – attribute – circumstance” frameworks. Meanwhile it can’t concern summaries where inner semantic references are still to become the object of researches. The implicit “built-in mechanism” of unpacking the contents and rebuilding the text is obviously that of reflection as the constantly present in compressions. Due to it a proverb can become a summary of a novel or an anecdote, and as a prompt it helps in recollecting the whole course of events. At the same time these recollections can differ one from another essentially attesting textual interpretability & transformability so that the very possibility to reproduce the original text remains out of the question, therefore compression remains irreducible to code.

It becomes from here obvious that compilation & compression are reciprocal procedures. As far as each entry of indices can become a potential title for summary the compiling procedure can be conceived also as the intitulation of text (the ultimate limit of convolution). In its turn convolution as the element to be reproduced in textual expansion needs its indexation. The more succinct is a description, the more it approximates code (without being reduced to it) and the more it moves away from the described text. A title as the point of convergence of propositional and appositive descriptive structures becomes at the same time the point of reduction to code. Compilations & compressions can be said to become a special case of the well known procedure of data representation (or knowledge representation) as the product of textual interpretation. The dependence of representative forms upon the interpretative approach can be demonstrated with the results of the explorations of memory. It has turned out that the representation doesn’t only eliminate “inessential details” but it also adds those absent in the primary contents. It is something integrated as a whole that is reproduced and represented in memory⁵³⁶. Thus it is integration that plays here decisive role so that the data representation of a text must first of all take into account the functions of the described passage. The task of data representation presupposes functional approach from the beginning onwards.

Any text is accompanied with the shadows (epiphenomena, satellites) of indices. At the same time it always presupposes propositional form due to its integration. The appositive structure remains then an abstraction extracted from the text as its collateral product. It impoverishes text essentially; therefore it can’t replace text in its descriptive representation being only auxiliary analytical abstraction & extraction. It is worth observing that in appositive structures (enumerations, compilations) the ambiguity grows due to the indefiniteness resulted from the interpretative textual distortion. In particular an index’s entries are to be regarded as the bunches of homonyms so that the necessary meaning can be detected due to the coexistence of these entries within the borders of the same place describing the same textual passage with its measure of semantic compatibility (the “isotope”). In the same way summary as textual compression is connected with the formation of homonyms (whereas expansion presupposes the dissociation of homonymous bunches). Such homonymous ambiguity becomes especially clear in titles that can become puzzles to be solved in the whole text!

Together with such collateral products of analysis another kind of description is to bear in mind, that’s of propositional compression with the ultimate limit of title being an implicit proposition. It is the participation of an observer that is needed for compiling such propositional textual summaries. Observer becomes an indispensable agent of descriptive procedures conducting dialogue with an author. In this respect incomplete sentences (as titles and subtitles) become compressions to be evolved due to the retention of the inner ties that enable their reversible transformations into propositions. Therefore it becomes evident that propositional transformations of text must have preference in descriptive procedures in comparison to indexation of appositive structures. Accordingly one should follow the mentioned homology between artistic text and a single sentence or a single word (A.A. Potebnya, O.E. Mandelstam, V.G. Admoni) with the correction that it must go about integrative conclusion (and not separate propositions or notions) that can represent textual entity. Besides, indexation itself causes cumulative effect entailing the necessity for the compiled enumeration to return to propositional structure.

The technique of textual elaboration aiming at data representation has been earlier developed with the view of automatic summarizing scientific texts. It gave earlier in particular the results of the “quotation + comment” model that the procedures were supposed to be reduced to⁵³⁷. Meanwhile we have just seen that such approach presupposes only extensive opportunity of preparing lacunas to be filled up later whereas the intensive way of description always gains priority. Paradoxically it is comment and not quotation that takes the initial place in any descriptive procedure. It comes from the above discussed that one always needs an observer’s attendance to obtain the necessary supplements for textual quotations because the epiphenomena of implicit textual data are at hand to be taken into consideration. Be indexation & intitulation sufficient for data representation of textual disposition, they obviously would be insufficient to disclose the theme or plot of the described situation. One needs therefore the disclosure of textual latency of inner form with its locutions as the situational prototypes (to be found in particular in the generic textual register).

The existence of latent data within the original text as the object of convolution reminds the reciprocity of compression and expansion. Subsequently the text to be compressed can be taken as incomplete that has first to be expanded. In particular itv can be expanded in extensive way as far as everywhere one can find lacunas due to the universal property of incompleteness. Such extensive expansion can be exemplified with the already mentioned rhetoric device of amplification where textual construction always admits supplementing with decorative additions. Still of more importance is a more difficult intensive way of expansion as the disclosure of implicit contents. Thus the descriptive paradox arises: expansion precedes compression because each text can be taken for a summary of some more extent text. The universal textual incompleteness results in the necessity of preliminary expansion. Text must generate further text before being described and summarized!

The development of the approaches to the tasks of the kind has accumulated enormous experience. After the inventions of Immediate Constituents (IC) analysis in descriptivism and Derivative Trees (DT) devices in generative linguistics the program of General Problem Solver (GPS) initiated the development of the so called artificial intelligence (AI) researches that became the source for computational linguistics. All these achievements have become implemented for common language analysis. One of the most eminent achievements of Artificial Intelligence studies was the discovery of the opportunity to generalize paradigmatic aspect of a language as algorithm and in its turn to describe algorithms (as well as calculations as a whole) as ordered sets of elements and prescriptions. It is worth reminding that the concept of generative grammar (GG) introduced by N. Chomsky as the generalization of paradigm bears the same definitive features that algorithms (and calculations) do. For GG it sounds as the sequence <{V_T}, {V_A}, I, S> where V_T and V_A denote terminal and auxiliary lexicons respectively, I is for initial or axiomatic rules (relations) and S denotes set of relations to build the “correct” sequences⁵³⁸. Meanwhile the similar modernized definition is valid for algorithms (as the implementation of the ideas of Turing and Church) where one finds <{E}, {A}, {B}, n, C>, the E, B denoting input and output alphabets with A as intermediary one, C again being the set of relations and n designating dimension. The same concerns the objects of GG and algorithms that are sequences⁵³⁹. As to the linguistic consequences of the mentioned novelties, the most important of them consist in the opportunity of representing paradigmatic and taxonomic aspects of a language with the aid of listing structures (indices) where the idea of an ordered set and its subspecies (vectors, tensors, matrices, strings, at last, arrays as the most universal kinds of them) would be generalized.

The principle that enables such algorithm’s description is based upon the so called Erbran’s universe (from the name of the mathematician that had suggested it still in 1920-s) where a string of propositions can be replaced with the apposition (enumeration) of names⁵⁴⁰. Thus concept of enumerative structure or listing (register, array) elaborated in the AI becomes a generalization of the concept of algorithm that includes paradigmatic and taxonomic aspects of language (with such subspecies as thesaurus, etymological nest, semantic field). It permits to replace the dualistic opposition (paradigmatic vs. syntactical aspects) with new non-dualistic opposition (listing vs. textual aspects of language). The eviction of dualistic approach is here achieved due to the fact that lists (registers, massifs) can be regarded as a special kind of text with loose inner connections. Text differs from list on the grounds of reciprocal references that unite its units in an inseparable structure. As a result an opposition [List vs. Text] replaces the previous [Language vs. Speech] one. Here the presence of a referential net of relations becomes a definitive feature. In contrast to a usual ordered set or sequence that meets the demands for listing it is such hidden net that determine special places reserved for the allocation of verbal substance. To use a text’s definition let be one of the most frequently cited books by Yu.A. Shreider be quoted where the stress has been put just upon this relative structure of these situational places⁵⁴¹. It is the referential net that turns out to become the most essential textual feature. Let the syntactical scheme be defined as S = <{M}; A₁, …,A_n> then text will be where  denotes mapping from M into an ordered set [Шрейдер, 1971, 199]. In other words it is these latent invisible references that transform a plain list (sequence, array, ordered set) into text. To go to more recent results it is to underline that within the AI there is the concept of structure differing from lists (arrays) just due to the refusal from homogeneity⁵⁴².

An essential property of the structures approximating that of a text is the definitive role of positions (and particularly of the way of searches) for the identification of their elements occupying these positions on the branches of derivative trees in GG⁵⁴³. As an example of textual distinctive features that enable distinguishing them from lists and other usually ordered sets may be mentioned the so called nodes’ method widely used in expert systems. It goes about the nodes of a semantic net arranging the decision-making procedure, a node bearing a set of rules. The effects of these rules are to be compared to a bunch of references uniting inputs with outputs⁵⁴⁴.

To sum up one can say that it is the listing procedures that remain the worthiest for the proper linguistic aims from the obtained results of AI. The development of computational devices builds perfectly autonomous sphere of knowledge that has on reality much lesser common points with linguistics as one had relied earlier, but the listing procedures belong to that field where such development would bring mutual effects. “Nests” in etymology, for example, represent a very persuasive sample of listing structures where enumerations in one language are continued in another. Another case of the usage of listing procedures is associated with the so called frames.

The concept of frame has been introduced within the artificial intelligence studies as a generalization of statistical description of an object through a selected set of parameters or attributes. Such a set represents an ordered sequence and corresponds to that of vector or tensor as a similar ordered set of coordinates in an abstract space. One uses the term array to designate all the ordered sets of data of the kind. The peculiarity of frame consists in it obtaining the structure of a questionnaire so that each notation could be identified through its place (respective number)⁵⁴⁵. Such places for notations are called slots that represent a questionnaire’s questions to be answered and filled with descriptions⁵⁴⁶. A special subspecies of slots is designated as terminals that are regarded as those to be filled with hyponyms of the names selected for the slots. The concept of a slot has come from cinematic practice where different “windows” enable showing the object from different viewpoints thus representing its peculiarities. This attempt to treat verbal material as visual determines also obvious limitations when it becomes impossible to fulfill descriptive tasks with the selected number of slots, and one needs to build a net of several frames⁵⁴⁷. Such cases demonstrate inefficiency of frames’ descriptive approaches, so it is assumed as a preferable to develop those semantic nets which would be void of the mentioned limitations⁵⁴⁸.

As to the use of frames for the description of narratives, one can distinguish a particular type of the so called predicate frames that can represent plot in the way that Julius Caesar’s phrase veni, vidi, vici (that exemplifies actually the construction of