3 - Transformation patterns
"The entry in that dictionary, after not explaining
any of this-I discovered it some time later-
ON THE NET - english
Igor´ Aleksàndrovič Mel´čùk
The study of the transformation of thought into language has always been the main focus of a wide range of researchers. While studying the analysis process, on which we dwelled in the second part of this course, has the evident advantage of dealing with a concrete object - the text - from which inferences are gathered that lead to the formation of thoughts and links, studying the synthesis process, with which we are now dealing, implies the difficulty of discussing an invisible object, not easily objectifiable: thought, sense.
One research trend, wishing to describe the process of synthesis for its hypothetic use in machine translation in objective terms, takes its cue from the Homskian2 theory in order to describe the passage from sense to text as a pure transformation, without going into details as far as the thought formation is concerned, with all the psychological consequences that such an approach would imply. The main advocate of such approach is the Russian researcher, French first, and then Canadian by adoption, Igor´ Aleksàndrovič Mel´čùk.
The part of Mel´čuk's thought that interests us here, from the huge quantity published by this linguist, is relatively tiny and concerns what the researcher calls "sense<=>text system". Our interest is merely theoretical because, as we'll see, Mel´čuk's system is based on the notion of "synonym" and "synonymic paraphrase", which finds a very little practical applicability to the reality of translation.
The main thesis is as follows:
|Natural language is a peculiar type of transformer reprocessing given sense into corresponding texts and given texts into corresponding senses3.|
The notion of "sense" is limited to its informational aspect, and forcefully ignores all the sphere of connotation and meaning nuances. At the basis of such conviction is the notion of "meaning equality " [ravnoznačnost´], and at the center of Mel´čùk's interests is "synonymic transformation" of one text into another text "of equal meaning". Such an attitude implies some limitations, which the Russian researcher himself acknowledges:
|meaning equality must be understood within the limits of a given conventional precision: we have the right to agree to ignore sense nuances that are too subtle for our objectives4.|
It is clear that such a way of dealing with the study of meaning and its transformation into text is, by approach, close to the notion of translation: the very notion of "sense" is defined on the basis of the notion of "translation", even if by "translation" we mean something very simplified when compared to what we have come to understand it as in this course:
|for us sense is inextricable from synonymic paraphrase, particularly from translation (which is simply an interlingual periphrasis). Such view of sense is very outdated, for example, compared to R. O. JAkobson: "The signifier is what is perceived, while signified is what is understood or, ion more concrete and operative terms, what is translated"(Jakobson 1959, p. 62)5.|
The model is, therefore, simplified, but the advantage of such simplification lies in its applicability, in its operational functionality. The main purpose of the research on the sense<=>text model lies in the taxonomical classification of the modes of transformation so as to provide an objective logical descriptions of individual shifts that a computer can accept as an input. Returning to Chomsky's theory-contemplating a competence that is the single speaker's expressive potential and a performance that is the actual linguistic practice of that speaker in a concrete situation-, Mel´čùk contents himself, for the time being, with being able to make a computer work on competence.
In other words, the aim is not to have computers translate real sentences, speakers' speech acts, texts that a translator encounters in her everyday practice. It is, by contrast, to try and transform normal, regular sentences, formed according to acquired norms and not deviating from these by any formal exception.
The description of sense-text correspondences must incorporate, in Mel´čùk's opinion, three components:
1) an inventory of the elementary meaning units (semes) and the rules for their combination into more complex sense units (semantic representations);
2) an inventory of the elementary text units (morphs) and the rules for their combination into more complex text units (composed words, word combinations, sentences);
3) rules for comparing every sense unit against the corresponding text units6.
One of the problems with Mel´čùk's theory lies in the presumption of describing "passages from complex senses (obtained through combination operations) to equally complex texts [...]"7, while we know that the thought generating a text is more complex than the text. First, inner language, "mentalese" in other words, is much faster than external verbal language, and it is impossible for the actualizer -writer/speaker- to be able to produce text at the speed it is produced. Second, the organization of verbal language is linear. Words are combined according to the rules active on the syntagmatic axis and selected according to the rules active on the paradigmatic axis. From this junction, a "line" of speech is born. Mental language is, on the contrary, closer to a hypertext than to a text: it contains links beyond physical contiguity (preceding word, following word) or logical connection (affinity of function, sense, sound, etc.). It also implies at first glance arbitrary links of a mnemonic or affective character, free associations, sudden fulgurations. Third, we know that information theory always indicates the presence of a loss in every communication act. Mel´čùk's theory, speaking of complex thought matching an equally complex text, does not allow for the loss implied in thought verbalization, simplifying also in this sense.
Like in Chomsky's theory, Mel´čùk's as well divides into two main elements: a linguistic part, that informs on the functioning of a language in a both descriptive and prescriptive sense, and an algorithmic part, in which mechanisms or procedures are devised that serve for making use of information about language. "Algorithm" is a technical term of mathematics deriving from the Western distortion of the name of the VIII Century Arab mathematician al-Khuwarizmi. It is a procedure for resolving problems that has the aim of sorting out the smallest possible number of rules with precise and punctual indications for executing them. Algorithms are the basis of the compilation of instructions for computers; this is why Mel´čùk wants to create them for verbal synthesis.
While Chomsky's system is generative (its aim is to describe text generation based on sense), Mel´čùk's system is transformational or translational:
|Un MST [sense-text model] essaie de se comporter comme un locuteur, qui ne passe son temps ni à générer des ensembles des phrases grammaticalement correctes ou à distinguer entre les phrases correctes et incorrectes, ni à transformer des structures abstraites ; un locuteur parle, c'est-à-dire qu'il exprime, au moyen de textes, les sens qu'il veut communiquer. Un MST doit faire la même chose: «traduire» un sens donné en un texte qui l'exprime (voilà pourquoi ce modèle est qualifié de «traductif»)8.|
We are not going to further follow the sense-text theory, for the time being. However, I think it important to highlight some aspects that I consider important for translation science, that are synthesized here:
- speaking of "putting into words" mental material or inner language as a translation process, which at the beginning looked like a somewhat heretical statement, the more we consider the thought of researchers in the divergent sciences is increasingly appropriate. Mel´čùk has as a major focus what we would call "translation process" from mental into verbal (and vice versa). In this he has a similar attitude to that of behavioral psychologists: he deliberately chooses to ignore what happens in the "black box" of mind and concentrates on the outer results of such functioning;
- Mel´čùk's attitude toward the stated problem is admirable on the methodological plane. To look for a taxonomy of phenomena involved in the translation process is a potentially very useful stimulus for translation science. In order to avoid studying a theory that is not easily applied to everyday translation practice, I think it advisable not to use simplified models that deal with a simplified language made only of the most important and best known rules of a language. This is the reason why I do not further investigate the diagrams created by Mel´čùk and used by him to experiment machine-assisted translation. I think it essential to work with connotation, with which virtually any real speech act is dense. But probably one day, when research is able to describe language in greater details, the taxonomic approach described in this unit will be at its methodological nucleus.
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, translation from the Spanish by Esther Allen, ISBN 0-8112-1466-4.
MEL´ČÚK I. A. Opyt teorii lingvističeskih modelej «smysl<=>tekst». Semantika, sintaksis. [Experiment of theory of the linguistic "sense<=>text" models. Semantic, sintaxis]<. Moskvà, Nauka, 1974.
MEL'CUK [MEL´ČÚK] I. Vers une linguistique Sens-Texte. Leçon inaugurale. Paris, Collège de France, 1997, 78 pages.
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. 130. «La nota de ese diccionario, tras no explicar cuanto acabo de contar y descubrí algo después, terminaba así [...]». Marías 1998 (2000), p. 161.
2 By Noam Chomsky. Since in this name the pronunciation of the first sound is like in "hooligan", or the Spanish «Javier», the French «haut», the Tuscan «casa», for an easier pronunciation the adjective war formed on the basis of its sound and not of the way the name is written in the West.
3 Mel´čuk 1974, p. 9.
4 Mel´čuk 1974, p. 10.
5 Mel´čuk 1974, p. 11.
6 Mel´čuk 1974, p. 18.
7 Mel´čuk 1974, p. 18.
8 Mel´čuk 1997, p. 7.