Logos Multilingual Portal

6 - Text generation - third part

"[...] to tell what happened, give an account of what
took place [...] is, in fact, a mere illusion or

Natural languages, as opposed to artificial languages, are very flexible. The different components (conceptual, lexical and syntactic) are highly interdependent, each component possibly influencing the others. The advantage of such heterarchical architecture is that it allows for various orders of data processing. For example, lexical choice may precede the choice of syntactic structure and vice versa2.

Another argument for hypothesizing that there is no rigid chronological dismantling between the mental processing and verbal draft stages is that the order in which thoughts occur does not necessarily coincide with the order in which they are formulated in verbal terms. This is also because verbal language proceeds along a line (words are in sequence) or segment (whence "suprasegmental traits"), while thought is more similar to a hypertext in which every element is related not only to the previous and the next one according to their actualization; but also to other elements which are contiguous for reasons of sound, paradigm, sense, experience, affect, accident etc.

When choosing the thoughts to be expressed, we navigate in a multidimensional hyperspace (long-term memory). The objects of this space are indexed along various dimensions (from, function, part of, etc.). For example, a rose can be seen as a flower, a gift or a symbol, a car may be indexed according to its origin (France, Italy), producer (Renault, Fiat), function (private, public transportation), type (sport, 4-wheel drive), price (luxury, economic), year (old-timer, modern), etc. This multidimensional classification (cross indexing) allows objects to appear in several categories or at different levels of a tree3.

  Many experiments were carried out analyzing the trivial mistakes we all make when we speak (slips of the tongue, substitution of a word for another, inverted elements etc.). From these experiments one can try and abductively reconstruct which were the mental processing stages that determined the analyzed formulations. The prevailing view among researchers examining scientific literature about involuntary verbal mistakes embraces the view of step-by-step processing in stages; in a first stage actions or events are located in their outline, then the elements composing them, then again the interrelations, according to a model like this one:

Figure 1: the tree that, with its branches, represents the mental processes necessary to formulate a sentence (from Zock 1996). Temporal progression is represented moving from top to bottom. Horizontally the various part of the sentence are extended.

This tree represents the process through which the sentence «When the old man saw the little boy drowning in the river, he went to his canoe in order to rescue him»4 is generated.
  As you can see moving from top to bottom, from a cognitive point of view at first the two major events around which the sentence is created are identified (the old man sees, the old man goes). In a second stage, the two events are specified in their outline: perception, finalized action. In the third stage specification is greater, and so on, until the details of the final version are reached (lower part of the tree).
  One of the elements favoring the mixed-generation view (implying approximate but global macroplanning first, while then, the more the time to actualize the single parts into words is closer, single microplannings and detailed verbal actualizations occur) is the presence of elements making sense only contemplating the whole speech act like, in Zock''s example, "When". Such a word, at the beginning of the sentence, would be unpredictable and insensible if the writer (or the speaker) did not know the global progression of the speech act, composed by a main clause and a subordinate, temporal clause.
  On the other hand, short-term memory is too little to allow for the planning of the whole sentence in all details from the time the first segment is actualized. It is, therefore, necessary to set up the sentence planning question like Zock does, in two stages: a general plan in which the skeleton of the sentence is present, and many specific micro-plans in which single components are actualized.
  When we speak of metatext drafting in terms of translation from the mental into the verbal, we want to give an approximate, general idea. Going into details one realizes that, actually, there is a series of micro-translations that, like shuttles, move in the two ways from verbal to mental and vice versa. After the first approximate translation of mental material into lexicon, the semantic fields of the selected words, together with their syntactical combinability, the connotative meanings of these words for the writing person, the cognitive experiences connected to the use of such words by the subject determine an informational feedback (from verbal into mental) that influences the selection of other words and the completion and/or modification of syntactical structures (from mental into verbal).
  Both the stage in which the text of a translation is drafted, and the stage in which the metatext is revised, add up to a continuous work of micro-translation from the mental into verbal and vice versa that ends only in the moment when the decision not to intervene on the text any longer is made (when the draft is considered as ultimate).


Bibliographical references

BATEMAN J. & ZOCK M. Natural Language Generation, in R. Mitkov, editor, Handbook of Computational Linguistics, Oxford University Press, 2001, ISBN

MARÍAS J. Negra espalda del tiempo, Punto de lectura, 2000 (original edition 1998), ISBN 84-663-0007-7.

MARÍAS J. Dark Back of Time, New York, New Directions, 2001 (translated by Esther Allen), ISBN 0-8112-1466-4.

ZOCK M. Holmes meets Montgomery: an unusual yet necessary encounter between a detective and a general, or, the need of analytical and strategic skills in outline planning, in VI Simposio Internacional de Comunicacion Social, Santiago de Cuba, 1999, p. 478-483.

ZOCK M. The power of words, in Message Planning, 16th International Conference on Computational Linguistics (COLING), København, 1996, p. 990-5.

ZOCK M. Sentence generation by pattern matching: the problem of syntactic choice, in R. Mitkov & N. Nicolov editors, Recent Advances in Natural Language Processing. Series: Current Issues in Linguistic Theory, Amsterdam, Benjamins, 1997, ISBN p. 317-352.

1 Marías 2001, p. 8. «[...] relatar lo ocurrido, dar cuenta de lo acaecido [...] es una mera ilusión o quimera [...] » Marías 1998 (2000), p. 10.
2 Zock 1997, p. 326.
3 Zock 1999, p. 481.
4 Zock 1996, p. 990-991.