“Meaning ⇔ Text” Theory and the linguistic universe of Igor Mel’čuk

Full Text

1. Introduction

This issue is dedicated to the internationally renowned linguist Igor Mel’čuk, the founder of the “Meaning ⇔ Text” linguistic theory and one of the pioneers of machine translation. His scholarly life is impressive for its outstanding results and the grand scale of problems discussed and solved. He worked for 20 years at the Institute of Linguistics of the Russian Academy of Sciences, and then another 32 at the University of Montreal, Canada. Since then he has continued to work actively, as Honorary Doctor of the Institute of Linguistics of the Russian Academy of Sciences, as Emeritus Professor at the University of Montreal and a member of the Royal Society of Canada. He lectures, advises young researchers, writes books and articles. So far, he has 305 published articles and 49 books, and work on the next book has begun. Thus, along with a personal anniversary, Igor Mel’čuk celebrates his professional anniversary – 65 years of continuous research and academic creativity.

With his innovative ideas and devotion to science, he has inspired and continues to inspire his colleagues, novice researchers and students both in Russia and internationally, as his linguistic universe has no boundaries, and his human qualities such as kindness, openness and sparkling humour fascinate anyone who has had the privilege of interacting with him.

We invited some of Igor’s longstanding friends, with whom he started and continued his career, his colleagues and students, as well as admirers and disciples of his research work to contribute to this issue. Together with them, we have prepared this intellectual and emotional gift to him. On behalf of the editorial board of the journal, our authors and readers, we wholeheartedly congratulate Igor Mel’čuk on his ninetieth birthday and wish him good health, inexhaustible optimism and new creative achievements.

In this issue, we intend to summarize the contribution Mel’čuk has made to semantics, to the development of the theory of meaning and the methodology of text analysis, as well as to trace the influence of his “Meaning ⇔ Text” theory on the further advancement of text linguistics, the theory and practice of machine translation, lexicography, typology, computational and corpus linguistics. In our introductory article, we will briefly dwell on the main ideas of this theory and outline the directions it has indicated for many fields of language study.

2. Meaning ⇔ Text Theory

2.1. Meaning ⇔ Text linguistic model: its highlights

The Meaning ⇔ Text Theory (MTT) was elaborated by Igor Mel’čuk about 60 years ago and became a primary focus for Mel’čuk as a researcher. Though the theory appeared in the late 1960s, it has not lost its relevance. Since then, different scholars from various schools of linguistics worldwide have been using its postulates, methodology or notions in their research. The theory explains the way the speaker generates the meaning embodied in the text which is understood as a meaningful unity. Many other domains of linguistic research have benefited from MTT, and its findings and approaches are applied in diverse studies.

Since its origins linguistics has sought answers to several major questions: How does language convey meaning? How do linguistic resources combine to produce it? To answer these questions, on the one hand, it is necessary to define what language is, as an entity, and how it is designed. On the other, it is crucial to account for the way meaning is rendered in communication which is viewed, first and foremost, as the transfer of meaning. Mel’čuk’s theory concerns both questions, and builds an integral approach that unites the intrinsic features of the language and the way they are responsible for rendering meaning. MTT combines both strands in an integrated vision of the linguistic construction of meaning via text.

The theory describes a natural language which is understood as a logical device (Mel’čuk 2021: 10), i. e., one designed according to the laws of logic. From our temporal perspective, it seems possible that the idea of such a device may have been influenced by the ideas of the 60s, when a profound interest in cybernetics and the power of machines permeated all branches of scholarly research. This “device” (or machine) provides transition from the meaning that the speaker wants to render to the text which serves as its carrier. Though it is mentioned that the device works both ways – from meaning to text for the speaker and vice versa for the listener, MTT prioritizes text generation (in other words, speaking) over text interpretation (comprehension), reflected in the name of the theory: from meaning to text, which is why we have Meaning ⇔ Text Theory and despite the special sign of a two-way transition, the meaning in the name still comes first.

Another idea popular at the time MTT was born is that of modelling objects to understand how they work. Mel’čuk devised a model of a natural language which is understood as “a set of rules encoded in the brains of its speakers that establish a correspondence between meanings of L(anguage) and their expression, or texts of L(anguage)” (Mel’čuk 2020: 4). Since language is a set of rigorous rules that establish correspondences between meanings and possible connected expressions that could surface in text, this system of rules could serve as a model of language.

This model is formalized via a system of formal, symbolically represented rules. These decode the process of exteriorization of deep level dependences among linguistic units through trees of dependence and the description of lexical functions. Mel’čuk also mentions essential requirements for these rules: they must be logically consistent, unambiguous, and transparent in terms of rendering content (Gladky & Mel’čuk 1969: 9) as well as “parsimonious and elegant in their form” (Mel’čuk 2020: 5). In Mel’cuk’s words, “linguistic meanings (in the technical sense of the term) appear as formal symbolic objects called semantic representations [SemRs], and texts—as phonetic representations [PhonRs]” which could be expressed by way of a formula {SemRi} ⇐language⇒ {PhonRj} | i ≠ j, 0 < i, j ≤ ∞ (Mel’cuk 2021: 9). Several kinds of formalism are used: “semantic networks for representing meanings of sentences and lexical units; dependency trees for representing the syntactic structure of sentences; lexical functions for representing lexical relations; and rules of various types for representing semantic operations” (Mel’čuk & Milićević 2020: xvii).

This model gives weight to the semantic component of meaning. Mel’čuk states that semantics “is not “just another component” of a linguistic system: it occupies a special place because language is above all a communication tool – that is, a means for conveying meaning” (Mel’čuk & Milićević 2020: 6). Thus, the semantic component constitutes the core of language description, as semantic properties of linguistic expressions largely predetermine their syntactic behavior as well as influence their morphology (Mel’čuk & Milićević 2020: 6)¹. Consequently, this model establishes close ties between vocabulary and grammar. According to the Meaning ⇔ Text Theory, semantics is viewed as a system of rigorous formal rules since language per se is represented as a set of rules that establish correspondences between meanings and their possible expressions. Another idea behind the semantic model is the possibility of meaning decomposition, which assumes that complex meanings may be represented via simpler ones (cf. Anna Wierzbicka’s notion of “semantic primitives” [Wierzbicka 2021], see also [Mel’čuk 1989]). Mel’čuk claims that semantic representation is based on a limited number of elementary semes which in some way resemble, and may be called, “atoms of meaning” (Mel’čuk 1999: 58).

As may be seen, this model is multilevel or, in Mel’čuk’s words, stratificational (= multi-stratal) (Mel’čuk & Milićević 2020: 10), for language is understood as a multi-level entity. The author singles out a hierarchy of levels including semantic (meaning), syntactic, morphological, phonological and phonetic / graphic (text). At each major representational level except the semantic, a Meaning-Text Model recognizes two sub-levels which are called ‘deep’ and ‘surface’. Thus, we can speak about a deep-syntactic representation and a surface-syntactic representation, and so on. Mel’čuk provides the following explanation for these sub-levels: “A deep sub-level is oriented towards the meaning – that is, towards the content the Speaker wants to express; its task is to explicitly reflect the relevant informational distinctions. A surface sub-level is oriented towards the text – that is, towards the form in which the content is expressed; its task is to explicitly reflect all relevant formal distinctions” (Mel’čuk 2020 & Milićević: 16).

In terms of the structural features of each layer, semantics is modelled by a network (more than one way in to each point, more than one way out), while the final output, a phonological string, is modelled by a line (one way in and one way out from each point). The intermediate, syntactic layer is modelled by trees, which, appropriately, have just one way in to each point, but more than one way out. In its overall architecture, then, MTT is intended to parallel the way meanings – ultimately related to concepts held in the networked structure of neurons in the human brain – pass via morphosyntactic structures (trees) to the linear string of phonological production.

The model is also structural as language itself and its units are understood as structures. Thus, semantics is described in terms of semantic structures which are represented as networks with nodes. Sentences and their semantics are represented by way of deep and surface syntactic structures (Mel’čuk 2021: 12–13). The same goes for morphology. The structural approach makes the theory applicable for machine translation and text generation.

This model is functional: it is based on MTT, which serves as a framework for functional models of languages with its use of various formal systems: semantic networks, syntactic trees, lexical functions, paraphrasing rules, and so on. The strength of functional models lies in their ability to model an object that is hidden from direct observation. Such hidden properties of language may only be discerned by rigorous analytical processes. Functional models represent the behavior of the object under analysis, which accounts for their usefulness in understanding the inner workings of the brain, appropriately described as a “black box” (Mel’čuk 2012a: 18). In fact, a functional model reveals correspondences between a set of meanings and a set of texts, accounting for the way a speaker expresses meanings via texts and, conversely, extracts meanings from texts (Mel’čuk & Milićević 2020: 8).

This model is global, for it works at all levels of language as a system, showing how language functions from meaning to text. The model’s aim is to make it possible to compute correct expressions for any meaning in English, and vice versa; to establish links not only between linguistic units and their meanings but also between English sentences and their meanings (Mel’čuk & Milićević 2020: 5).

Finally, this approach is synthesis oriented. Though the meaning-text correspondence is bi-directional, Mel’čuk claims that linguistics should study and describe language functioning in the Meaning-to-Text direction, as natural language is mostly about speaking rather than understanding (Mel’čuk 2021: 8–9). Linguistic synthesis, or text production, is more important for linguistics than analysis, or text understanding. This is because text comprehension involves, to a degree, the understanding of the subject matter treated, i.e. it presupposes the understanding of extralinguistic reality, a factor that may distort the whole picture. Mel’čuk posits that “going from a given meaning to all the texts that can express it liberates the researcher from huge “non-linguistic” difficulties leaving him face-to-face with the language in its purest state” (Mel’čuk 2021: 9–10). Another argument in favor of the priority of synthesis over analysis is the idea that before any text is analyzed it should be synthesized, which accords with the idea that text synthesis is primary to analysis

2.2. Meaning ⇔ Text Theory: major guidelines to linguistic semantics

Mel’čuk affirms that meaning and its description will be a considerable task for many generations of linguists (Mel’čuk 1999, 2012, Mel’čuk & Milićević 2020). The Meaning ⇔ Text theory attempted to present his own contribution to the solution to the problem. It offers a system of notions to describe lexical and syntactic semantics. Starting with a general understanding of meaning as “informational content that can be verbalized in the given language” (Mel’čuk & Milićević 2020: 4), Mel’čuk introduces a whole range of notions which have since been used in linguistic semantics to reveal the meaning of a language unit. Such notions as valency, actants, diatese, sentential semantics were born within this theory and have been successfully used by its followers. The correlation of all these terms constitutes an integrated system aimed at the description of linguistic semantics.

Another essential detail for modern semanticists is Mel’čuk’s treatment of meaning. He sees a difference between linguistic (= “shallow”) non-pragmatic, non-extralinguistic, non-encyclopedic meaning and real (= “deep”) meaning (Mel’čuk & Milićević 2020: 73). To explain these contrasts, Mel’čuk analyzes linguistic meaning as including propositional, communicative and rhetorical components (Mel’čuk & Milićević 2020: 77). Propositional meaning is the semantic content of a language unit and can be described by means of logical propositions; communicative meaning deals with the speaker’s communicative intentions, while rhetorical meaning involves the speaker’s stylistic intentions; whether the utterance is neutral, formal, colloquial, poetic, ironic, etc. (ibid.).

Mel’čuk is consistent in his claim that meaning is atomically built and extremely well-structured. Its description makes use of logical atoms, which are predicates and names of objects (Mel’čuk 1999: 62). This approach was used in Lexicography, and the “Explanatory Combinatorial Dictionary of Contemporary Russian” (Mel’čuk & Zholkovsky 1984) serves as a vivid example of its application.

2.3. Meaning ⇔ Text Theory: heuristics and application

The Meaning ⇔ Text Theory paved the way for various avenues of linguistic research. Within MTT an artificial semantic language was elaborated (Mel’čuk 1974) which enriched lexicography as well as provided an integral description of language. This artificial semantic language is abstract, since it was designed for a machine which has no knowledge of what any linguistic expression might mean. Thus, the formal language is strict, representing the structure of meaning. Another outcome of MTT for computational linguistics and machine translation is its building of an inventory of basic semes.

Meaning ⇔ Text Theory serves as a lens for other issues which Mel’čuk studies, and it found application in many areas. When Mel’čuk describes the range of problems that semantic studies border on and affect, he actually outlines the frontiers of his own theory and its impact on various fields of linguistic research. A renewed approach to lexical semantics, as well as producing findings about the semantic and syntactic aspects of linguistic units, contributed to the development of lexicology and lexicography as its applied counterparts. This culminated, on the one hand, in the expansion of modern semantics, which enriched natural language processing (e.g., machine translation, automatic text generation, mathematical linguistics), and on the other, led to advances in the field of language learning and teaching, as in both spheres the main focus is the transference of meaning (Mel’čuk & Milićević 2020: 7).

Another of Mel’čuk’s influential contributions concerns his idea of a language model and its application, not only in linguistics but also in interdisciplinary research. He himself sees a high practical utility of his model in at least three technological and social domains: natural language processing; teaching and learning languages, and manufacturing reference books such as dictionaries, pedagogical grammar books and manuals (Mel’čuk 2021: 28).

To sum up, it is useful to cite Igor Mel’čuk’s description of the perspectives of his theory, when he affirms that “the only reliable key to human thinking, in all its complexity, is natural language. Without a profound understanding of how language is functioning in our psyche, there will be no good understanding of information processing by the human brain. That is why functional models of language, and MTM (Meaning-Text Model) in particular, have nowadays acquired a special significance. Linguistics must take a place of honour among the “hard” sciences, and functional models, which embody the typical scientific approach to complex phenomena, will make their contribution” (Mel’čuk 2021: 28). This theory still serves as a framework and a lens for many other issues which Mel’čuk studies.

3. Outline of contributions to the issue

The issue is opened by Igor Mel’čuk. His article “Russian reduplicative surface-syntactic relations in the perspective of general syntax” is another gift to our readers (see Mel’čuk 2018a), and is part of a larger project aimed at creating an inventory of superficial-syntactic relations in the world’s languages, started 60 years ago (Mel’čuk 1962: 47–49). Since then, several lists of reduplicative surface-syntactic relations have been compiled in Russian, English, French, German, and other languages (e.g., Mel’čuk 2015: 444–453, 2016: 184–194,2018b, etc.). The purpose of this article is to supplement the existing lists and identify special Russian reduplicative superficial syntactic relations that appear exclusively in syntactic idioms based on reduplications. As a result of the study carried out within the framework of the “Meaning ⇔Text” Theory, 7 types of surface-syntactic relations in the Russian language were identified, which belong to the field of phraseology and have not yet received sufficient attention from scholars.

Anna Wierzbicka continues to develop the idea that, despite the phenomenal diversity of languages and cultures, which is often absolutized by researchers, there are basic elements of human thinking that are present as separate words (or signs) in all human languages (see e.g. Goddard & Wierzbicka 1994, Wierzbicka 1992, 2020, 2021). The concepts they designate are equally perceived by people regardless of the language they speak. According to Leibniz, they represent “the alphabet of human thought” (see Wierzbicka 1972: 6). In this issue, she explores the universal concepts of “I” and “you”. She uses an approach to the study of meaning, developed on the basis of “semantic primitives”, known under the acronym NSM (Natural Semantic Metalanguage) (see Goddard & Wierzbicka 1994, Wierzbicka 2021 and others). She argues that “YOU” and “I” (“I” and “THOU”) are fundamental elements of human thought, present as distinct words (or signs) in all human languages.

Wierzbicka's ideas of a universal metalanguage have been successfully tested by a number of scholars in typologically different languages (see Gladkova & Larina 2018) and have proven to be effective in explicating culture-specific words and concepts that are difficult to translate (e.g. Wierzbicka 1997, 1999, 2014, Levisen & Waters 2017, Bromhead & Ye 2020). In this issue, Mark Durie uses the NSM to explicate four Arabic words that name key Islamic concepts without equivalents in English. Emphasizing the importance of understanding the meaning of keywords when studying another culture, the author notes that they are inevitably distorted during translation, as they fall under the influence of another language. The study shows the effectiveness of NSM in explicating Arabic key words and overcoming their resistance to translation into English, as well as their role in providing an adequate understanding of the key concepts of another culture, including religious ones.

Natural semantic metalanguage (NSM) is also used as a method of semantic analysis by Anna Gladkova who considers changes in the modern Russian lexicon, namely, the development of new meanings for the words zhest’ (from the primary meaning of zhest’ ‘tin’ as a type of metal), zhestkij ‘hard/tough/firm’ and zhestko ‘firmly/toughly’. She proposes NSM explications of these words, identifies connection of their meanings with the cultural themes of ‘emotionality’, ‘not being in control’ and ‘strait talk’, and notes the increased salience of these words in modern Russian.

If the semantic metalanguage refers to the study of the internal processes of language, to what is often hidden from direct observation or internalized, the studies of other levels of the language system, also inspired by Mel’čuk’s ideas, refer to the units on the surface of language use, i.e., exteriorized. Thus, Nicholas Evans’ study concerns interactive phonetics, which can be further integrated into the sound system. The latter either expands the phonological system by incorporating new phonemes into itself, or expands the phonotactic possibilities of phonemes occurring in other phonotactic positions. In support of his argument the author draws on phonemes restricted to interactive contexts in the Papuan language Nen, as well as phonemes with restricted phonotactic distributions (English) or morphological distributions (Bininj Kunwok, an Australian language of the Gunwinyguan family). He proposes aims to expand our understanding of this kind of phoneme by considering specific vocal gestures involved in interactional settings.

A wide range of complex issues related to the field of morphology, semantics, pragmatics and typology is considered in the article “Perfective, performative and present: Some non‐standard combinations in Slavic and beyond” (Plungian, Rakhilina & Reznikova). Based on representative empirical material from the Russian National Corpus, which includes the parallel corpora containing examples from literary texts with their translations into other European languages, the authors explore a special type of context in which negative perfective forms acquire the meaning of “prospective present”. The authors argue that in some cases, these constructions correspond to speech acts and can be considered as a kind of performative. Such quasi-performative constructions expand our understanding of the meanings that present perfective forms can acquire.

Several articles in this issue of the journal present the results of research carried out at the intersection of semantics, pragmatics and lexicography. Barrios Rodriguez analyzes the difficulties of lexicographic processing of formulemes that arose when compiling the Spanish electronic dictionary Diretes. Noting that within the framework of the “Meaning ⇔ Text” theory, pragmatems have been formalized on the basis of lexical functions, the author argues the need for their formalization on the basis of pragmatic functions. Drawing on the results of the analysis of 200 formulemes, she concludes that pragmatic functions can be useful not only for formalizing formulemes and pragmatems in other languages, but also for revising the list of illocutionary verbs, from the perspective of Phraseology and Lexicography.

Olga Solopova and Tamara Khomutova present the first results of a larger project of an explanatory combinatorial dictionary of English conflict lexis. It is based on the corpus of texts on political discourse, and is generally consistent with Mel’čuk’s Meaning ⇔ Text theory. The article discusses the principles for creating such a dictionary, and provides an example of a dictionary entry whose microstructure includes semantic, phonological and co-occurrence zones.

Valentina Apresjan and Boris Iomdin explore Russian interrogative-relative pronouns (wh-words) as a lexicographic type which requires a unified description. They draw on the principles of the Moscow School of Semantics (see Apresjan 2005, 2009, Apresjan, Sannikov & Boguslavskii 2010, Apresjan, Iomdin & Boguslavskii 2012, Boguslavskii 1996, among many others), which largely continue and develop the ideas and principles of Mel’čuk’s “Meaning ⇔ Text” Theory. They include a multi-level description of language, a focus on semantic and combinatorial properties of linguistic items, and acknowledge the validity of the dictionary as a tool for linguistic research. The study focuses on semantic, syntactic and co-occurrence properties of a number of Russian interrogative pronouns. The results suggest a deep motivating connection between the semantic properties of interrogative pronouns and their combinatorial abilities.

The category of Russian aspect is the focus of Solovyev, Bochkarev & Bayrasheva, which is not surprising as Mel’čuk has made significant contributions to the formalization and understanding of this complex grammatical category. Drawing on some of his ideas, the authors continue the discussion of whether verbs in aspectual pairs are in inflectional or derivational relations. They propose two new approaches that compare the quantitative characteristics of aspectual pairs of Russian verbs of both types. The results of the study suggest that the suffixal and affixal methods of forming aspectual pairs have an equal status in terms of their inflectional or derivational classification.

4. Conclusion

In this introductory article we aimed to summarize the main tenets of “Meaning ⇔ Text” theory founded by Mel’čuk and outline its contribution to the development of diverse areas of modern linguistics. We must admit that this is a very brief overview which cannot cover the linguistic universe of Mel’čuk. The articles in this issue confirm that the principles of the “Meaning ⇔ Text” Theory, which focuses primarily on language synthesis, speech production and the understanding of meanings that are encoded in text, have been further developed by Mel’čuk and his followers, and continue to be used in all types of linguistic research.

No doubt, the theory will inspire new generations of linguists and acquire new followers, in their eternal quest for the essence of language and the workings of the minds that use it. It is our hope that, after becoming familiar with the articles published in this issue, our readers will turn to the books and articles written by Igor Mel'čuk and enjoy the harmony and depth of his ideas.

Once again, we conclude with wishing Igor many more years of joy, happiness and creativity!

 

^{1 This statement may seem obvious at first glance, but we should remember that one of the most widespread theories of language today, generative grammar and its connected ideologies, consider language in a fundamentally different way: not as a tool for expressing meanings, but as a formal mechanism for generating an infinite set of syntactically correct structures. These can later be “voiced” and “interpreted” with the help of a semantic (or formal-logical) component. In this regard, Mel’čuk’s position has not changed since his earliest works: if, for his generative opponents, language is primarily a “syntactic machine”, the products of which can later be assigned meaning, then in his conception, language has always been seen as a “translating machine”, i.e. a system of multi-level transformations that ensure the transition from meaning to text. This brings MTT closer to what are commonly known as “functional” language models today, although it differs from these latter in a number of respects.}

About the authors

Svetlana Viktorovna Ivanova

Pushkin Leningrad State University

Email: svet_victoria@mail.ru
ORCID iD: 0000-0002-0127-9934

Dr Habil. in Philology, Professor, Head of the Department of Romance and Germanic Philology and Linguodidactics

St. Petersburg, Russia

Tatiana Viktorovna Larina

RUDN University

Author for correspondence.
Email: larina-tv@rudn.ru
ORCID iD: 0000-0001-6167-455X

Doctor Habil., Full Professor

Moscow, Russia

References