Національна Академія Мистецтв України Інститут культурології

The ultimate degree of generalization can be represented with pronouns and pure auxiliary verbs (be, have, make, do etc.). While describing a text one has also to use such abstract predicative and pronominal auxiliaries that would designate “terra incognita” of the presumed categories. They can also be represented as the elements of comments. Thus one has abstract objects designated with auxiliaries (that include vacuous contents) and the concomitant circumstances verifying or falsifying such abstract categories. These verbal auxiliaries and pronouns with indefinitely wide meanings and vacant abstractions are to be made precise with the specialized elements supplementing them. One can cite here a very apt L. Martynov’s verse to elucidate the ubiquity of such widened pronouns of predicative nature: “There are the same and only dramatis personae: I, Thou, He, We, You, They”⁵⁶². The necessity of applying descriptors (or indicators) built in the manner of such generalities was substantiated by S.I. Gindin in view of the impossibility to reproduce the contents of propositional structures unequivocally⁵⁶³. Respectively the method of indexation that’s of textual representation with indices was acknowledged as the appropriate tool for semantics⁵⁶⁴.

Indices (listing structures of textual glossaries or thesauruses) as linear structures can be divided in those of chaining (concatenation) and branching (ramification) the lexical stuff. Separate sentences being turned into indices (lists), the whole text acquires an outlook of ramifications and concatenations of the sets of subordinate indices so that the respective textual database becomes a cluster compiled of lists. Such clusters can easily be transformed into graphs and respective matrices. In its turn indices obtained immediately with the respective transformations of separate propositions into appositions can be regarded as files (bound structures with their elements’ interdependence). To detect and display the ties inherent for the semantic references of a text one must thus loosen these files and build a wider listing structure that would represent ties inherent for the whole textual entity and not those of separate utterances. Such primary strings (listing structures) with an outlook of files demonstrate actually the well-known IC – trees of syntactic subordination reflected in their cluster - like structure. Respectively these lists include at least two levels: those built of formal predicate nodes (represented usually with verbs) and those of subordinate complements (agents) and circumstances and other attributes restricting the upper designated categories. Thus in reality two lists are at hand: that of titles of predicates with their abstractions and that of their restrictive attributes, so that the scheme “object - attributes” returns. These lists represent generalities vs. particulars of the text respectively that are confronted here as those ruling vs. subordinate (though generalizing and ruling functions don’t coincide). While decomposing the primary indices one has to separate also predicative elements from those of the so called agents (or “actants” of L. Tesniere) together with circumstances as the further details⁵⁶⁵.

In its turn be the procedure of the kind carried upon till its limits, then the abstract schemes of the pronominal predicates appear that are to subsume the generalities to the respective abstract classes⁵⁶⁶. Being consequently applied such method of description would come to full substitution of textual lexical units with the abstract descriptors of artificially built metasystem. Such attempts have already been undertaken by V. V. Martynov still at the eve of AI researches⁵⁶⁷. Textual representation then must necessarily transgress the limits of pure indices of textual stuff and involve the descriptors of the metasystem. Not need to say that such description precludes any possibility of obtaining new knowledge and of solving problems as far as it deals with the limited set of ready descriptors.

From another side the compilation of details of concomitant circumstances as opposed to the predicates shows semantic compatibility (sometimes called lexical “isotopes”) that include mainly particulars of the text as far as they serve to diversify, differentiate and restrict the abstract categories to be made concrete. The formation of isotopic rows is regarded to be connected with textual mental integration that provides the entirety of textual semantic space⁵⁶⁸. To continue such statements one has grounds to remark that compatibility is also connected with motivational net as the foundation of textual integration and respectively as the necessary attributes of the categories represented in text.

At the same time one has reasons to doubt as to the validity of such “isotopes” in rendering the textual glossary: even the simplest enumeration is not a mere chaotic “heap of words” already while it evokes the background experience ordering the interrelation of lexical units. In every case such “horizontal” series of details presumes grouping in semantic fields, in taxonomic classes (as etymological nests), in groups of lexical attraction. All it precludes the simple heaping of words only on the grounds of their presence within the limits of the same place (“isotope”) of a taken textual passage. Another objection concerns the homonymous dissociation of the generalities so that the vertices of such nets can’t represent elementary units. Each abstract category represented with a vertex of the net splits into bunch of homonyms without attributes of concomitant circumstances that make its meaning precise. It is one of the reasons why a summary can’t be built of generalities as such in “vertical” dimension only. For instance, in the description of a tale or a short novel the predicate “to advise” (as applied to the protagonist of the hero) remains too ambiguous to be apt for a summary: the meanings of advices as such differ essentially in regard to their contents so that it goes about the homonymous dissociation of such predicates. It already prevents descriptors from being applicable for artistic text. One can see also the obvious limitations of such approach that restrict textual stuff with the level of potentialities. Thus one must renounce such attempts of reaching the ultimate edges of description to come to actualities. Together with these objections the confrontation of the indices of generalities vs. particulars discloses very essential deficiency. It looks out as if there is the predestinated vertical abstract scheme of predicates to be filled with diversifying horizontal details. In this case all specialized meanings look out as if they become the exemplification of generalized abstract scheme or its embellishing ornament. In its turn generalizations are indispensable already within the pure enumeration of details as far as the argumentation (or objections for refutation) for the coherently motivated integral text become necessary (that are out of the question to be withdrawn in its description). All it betrays the obvious scheme of the field structure with its “center vs. periphery” opposition, vertical generalities getting the centralizing role with subsuming textual predicates to abstract categories. Thus from one side details can’t be reduced to simple additional examples as well as generalizing moments are necessary to represent details in the adequate compilation. It means that the dissection of listing structure in the two vertically opposed levels of generalized and specialized entries would fail to represent textual contents adequately.

The device of such dissection is not the artificial explorative device only. There existed in the history of poetry the bright example of such confrontation of generalities and particulars in the so called versus rapportari. Here the rows of predicates and of the respective completive members were situated in separate lines so that the reader had to correlate the words from the preceding and the succeeding lines. Bright examples of the kind are to be found in Latin poetry of Mediaeval Ages according to the rule of the correlation of separate singular words (in Latin sive applicati, sive singular singulis [Курціус, 315-316]). They are built as the rows of the homogenous members of sentences written separately where the ruling member (predicate) is situated on the place in one line that corresponds to the place of the subordinate member in another line so that a series of couples arises. Especially it is to be stressed that such devices have become folklore proverbial texts as the examples from J. Werner’s collection bear witness: “Gaudens gaudenti, flens flenti, pauper egenti, prudens prudenti, stultus placet insipienti” (it befits for those being glad to rejoice, for those weeping to weep, for pauper to beg, for wise to inquire, for stupid to remain stupid), “Rex Famuli Catulus Classis Ferrum Leo Serpens – Dat Cupiunt Sequitur Sulcat Scindit Facit Angit – Dona Datum Lepores Mare Terram Vulnera Gentem – Ore Manu Pedibus Vento Bove Dente Veneo” (king, servants, puppy, fleet, plough, lion, serpent – gives, want, pursue, ploughs, cuts, causes, tortures – gifts, grants, hares, sea, soil, wounds, living beings – with the voice, in the hands, with legs, leeward, with an ox, with teeth, with poison) [Werner, G-3, R-69]⁵⁶⁹. Meanwhile it is easily noticeable that the weights of the members of such rows are not equal, and the field structures of such texts can always be displayed that in its turn is not stable and permits different ways of centralizations, and the same concerns referential nets of the locutions. Enumerative structures of such texts represent only potential structures that are still to be actualized. In particular it goes here about the separation of the layers of generalities vs. particulars that conceals some latent deep structure that is still to be detected with the means off attributive analysis.

Textual expansion (as opposed to compression) is usually correlated with the process of restriction, differentiation and diversification of abstractions, with making them concrete. Even a single word’s expansion demonstrates such differentiation as the aspect of organic growth. In its turn differentiation arises as the consequence of motivated textual growth and is determined with the referential net representing the motivation. Ultimately wide, indefinite and actually almost vacuous categories designated in pronominal manner are thus supplemented with the concomitant details that restrict their volume. Here one deals with the strategy of ultimately wide but indefinite abstract categories that transforms itself into the tactics of building their expansion. Auxiliary “pronominal predicates” designate seemingly vacant categories, they demand details for their restrictions, and this restriction generates text with its concrete images. In particular, amplification as the repletion of a text’s convolution in the series of generalities with the diversifying details can be mentioned among the known rhetoric device. At the same time one has to warn against conceiving it as a simple process of supplementing abstract skeleton with facultative decorative embellishments. Furthermore the reciprocal interdependence of generalities and particulars as the representation of problematic contents becomes especially clear in respect to the titles. It is not the known “sense” to be represented in “text” but the problem depicted with the attributive features to be explored that a title designates and the initial scheme of the text suggests. Title (and initial moment in textual generation and description at all) can’t be substituted with abstract generalities: it would become absurd to replace “Hamlet” or “Faust” with the pronoun “He”. In the same way separate particulars (as well as the proper names without peculiar implications as “Franz” instead of “The Robbers” in F. Schiller’s drama) are not sufficient for representing a title’s contents. Neither could title coincide with initial lines of the sedgment: it is not “to be or not to be” that determines the essence of Hamlet’s monologue, “to die as to sleep” or “to put an end to life’s calamities” being also possible versions for intitulation. Therefore title must necessarily refer to the problem described with its essential attributes as its prototype.

The analytical task is in no way to be reduced to the searches for the supposed generalities that would be hidden behind the particular cases demonstrated in the text in the manner of Shakespearean words from the 91st sonnet “But these particulars are not my measure / All these I better in one general best”. How very seductive such reduction would look out, a text can’t be reduced to the exemplification of generalities in the manner of allegories of the Jesuit “school theatre” or amplification of the prepared rhetoric scheme. It is implicit and latent problem and not generalized abstractions that are to be looked for. Action is not abstract operational procedure performed with a doer that could be designated with pronominal particle to exemplify the abstract didactic morale. Iago is not the personification of abstract Calumny as Hamlet is not the example for the “concept” of Hesitation. Neither is a lyrical text generated as an amplification to diversify the preponderated abstract scheme of abstract pronouns and predicates in the manner of coloring a ready design or filling an earlier prepared scheme with additional embellishments.

Here also a remark is to be made: although textual generation makes abstractions concrete with textual expansion and semantic diversification of categories, it is neither generalities as such nor separate particulars that initiate the generative process. It is not the vacuous abstractions neither the details in the manner of proper names that stand at the initial point of textual expansion: it is the problems to be solved and not abstract schemes to be exemplified. Therefore it is often the concomitant circumstances and not predicates that can grow to separate autonomous plots and give thus the prototype for conceiving the situation with the respective categories. For instance it is not the departure of the hero as such but the concomitant ominous event as the prediction of misfortune that gains importance. That is why it is neither general category as such nor special detail that becomes essential in summarizing the whole narration. In terms of functional approach it goes here about the transition from potentialities to actualities where actual predicates (the rhemes) appear. In any text the relationships of generalities and particulars can’t be reduced to that of obligatory skeleton of predicates and facultative embellishments of details in the manner of amplification. It is not ornamental decoration or allegorical examples because one deals with actualities where circumstances become predicates (rhemes). Textual perspective is constantly rebuilt with actual predicates taken from circumstantial details of horizon that replace potential predicates. Together with actualities the centralizing mode of textual field structure proper for potentialities is to be replaced with referential filament. The transformation of textual field’s structure with the transition to actualities does not only mean that generalities loose their centralizing function but also that the whole process of textual generation must be represented differently in comparison to that within potentialities. As far as it can’t go about the ways similar to rhetoric amplification (embellishments or exemplifications of abstract scheme as the “decoration vs. construction’s” model) one can’t also regard governing relations as the single generative mechanism. Thus centralized subordination as the generative mechanism should be replaced with more widely conceived motivation presupposing also spontaneity as its counterpart.

With the transition to actualities still one deficiency of the existent methods becomes obvious. It is to be stressed that attributive relations can by no means be reduced to those of inclusion or subordination. Data representation presupposes the “object - attribute” aspects of text that do not coincide with those of inclusion of hyponyms into class and subordination of circumstances to predicates: it can be easily exemplified with known cases of G. Peano (from the statements “Peter and Paul are apostles” and “There are twelve apostles” doesn’t follow the conclusion that *apostles are twelve) and E. Husserl (from the statements “The paper is red” and “Red is a color” doesn’t follow the conclusion that *paper is color)⁵⁷⁰. In terms of such distinctions one can notice that it goes only about attributes within the scope of potentialities. In particular one can say that objects disappear in textual representations of potentialities so that only abstract attributes remain. The both dissected indices of generalities and particulars belong to attributes and can’t represent the object without special references. It determines the principal deficiency of the above described analytical approaches that are closed within these attributive abstractions together with potential textual structures. Thus the task arises of uniting these attributes anew in such a way that it would become possible to reproduce the object as in actual message. It is important that attributive relations irreducible neither to generalization (and logical inclusion) nor to specialization (and reciprocal negation of differential features) are correlated with textual referential relations. Then attribution is to be correlated with implication as the source of the inferential foundation of textual coherence.

There must be made an important warning: the task of reproducing the compiled text from its compilation is to be esteemed as the unsolvable one because the incompleteness as the immanent textual property is here still reinforced and increased to the degree of indefiniteness. In the same way no summary can possess any ambition of being equal or similar to the original, neither can it become a germ to restore this original. One could only compare the novels from Decameron with the preceding descriptions of their plots. Compilation and convolution obviously don’t retain the whole textual functional structure comprising separate locutions and lack a program (algorithm) to restore exactly the text; vice versa, a set of the possibilities for new derivative texts is opened. Thus the demand of the original’s reproducibility is out of question in compilation & convolution. The tasks of analytical procedures consist in exploring possible derivative texts that can be generated from the data representation. The text (as the message) becomes only a text (without the message’s contents) while being compiled and compressed so that it is to be conceived as one of such possible versions.

The procedure of building relational semantic net (in opposite to frame) consists in the standard data representation of text without preliminary premises of frame questionnaire: the represented text is transformed in a listing (enumerative) structure that is in an index of locutions disclosing their semantic connections (both contact and distant) instead of purely syntactic scheme retained in frames. Thus it is with producing a compilation of phrases that the