Methodological corpora toolkit and its possibilities for modelling cognitive and semantic matrices

In linguistics, the end of the XX c. was generally marked by the reorientation of research attention from grammar (morphology and syntax) to the lexical-semantic system, and the beginning of the twenty-first century, by the active involvement of corpus technologies that have helped to revise traditional views of the language as a whole. As a consequence, it has been suggested that linguistics “outlines the contours of a new language model, which is significantly and fundamentally different from the earlier models postulated in the last quarter of the twentieth century” (Plungyan, 2008: 7-20).

In order to correctly understand in what scientific interpretation Plungyan uses the term “new language model”, we briefly formulate our own vision of the concept of “language model”, as well as try to understand the cognitive nature of such a model and its mechanisms, which are obviously based the ability of a native speaker of a particular language to reflect the ontology of that entity, using its multiple levels, and, to be more precise, to perform various combinations of linguistic characters (both in terms of expression and in terms of content) to provide efficiency of communication.

And precisely for the solution of such linguistic problems in the dynamic peak of the development of structuralism, when the procedures of modelling process in phonology and in syntax were actively developed (and semantics, on the contrary, remained the most debatable sphere at that time), Z. Harris introduces the term “model” in the scientific apparatus of linguistics, borrowing it from philosophical works devoted to the problems of mathematical logic (cit. Apresyan, 1966: 99-100). These priorities are fundamental for the development of new directions in structuralism - mathematical linguistics and quantitative linguistics.

Among the many definitions of the term “model” that exist today, the most optimal is philosophical interpretation of this term, which means an “object, artifact (artificially created scheme), which serves to present reality, reflects a set of features of existing or hypothetical objects and depends on the tasks formulated and solved by a researcher (its designer)” (Lukach, 2013: 144). The above-mentioned definition does not contradict the views of Losev (2004) on the methodological process of modelling, which in his understanding involves the construction, first, of simulation models of really existing objects and phenomena, and, second, of analytical models of hypothetical objects for predicting their functional signs.

A deeper study of modelling problems in the field of artificial intelligence has directed the attention of scholars to discovering the mechanisms of human consciousness. And designing models of virtual realities has changed in general scientific views about the limits of reality and led to a revision of the principles of construction of phenomena of model nature. As a result of these observations, the term “model / modelling”, and most importantly, the role of language in the processes of constructing model objects were redefined.

Actually, the hypotheses, which assumed that language is the mediator between the world and man, were discussed in the XIX c., beginning with the works of Humboldt and Potebnya, but they gained theoretical justification only in the XX century, and first of all, when a new understanding of language in cognitive science appeared in connection with the study of cognitive models - structures that reflect the perception, storage and transmission of information about the world (Kubryakova, 1996: 56-57).

Directly in science, a cognitive model is studied as an epistemological construct - a hypothesis about the structure of human consciousness and thinking. But in case of empirical confirmation, this model can acquire ontological status.

Representatives of cognitive semantics (J. Lakoff and others) distinguish two types of basic models: models of identification (or models of categorization) and models of mentalisation (i.e. conceptualisation) (Korolyova, 2012). Cognitive models of the first type reflect the process of segmentation from the holistic image of the world of individual most significant objects in order to further categorize them. Mentalisation models are closely linked to linguistic structures because they are a direct reflection of cognitive and purely linguistic semantics.

The question of how knowledge is represented in human memory is related to both processes: categorization and conceptualization, where the latest symbolic models provide the storage of knowledge of language, while the first models allow us to re-explain the essence of lexical and grammatical phenomena (Piaget, 1983).

Taking into account the new methodological provisions of the theory of categorization as a process of modelling the phenomena of objective reality, Lewis's statements that the language is a grammaticalised lexis, not a lexicalised grammar, seem to be correct (Lewis, 1999). In this case, cognitive-grammatical ability is considered, in particular, as an opportunity to form unique combinations of lexical units to describe and interpret non-frequent situations / cases, while the ready-made vocabulary already stores in the memory lexical combinations to designate standard and highly probable events (see more: Horina, 2014: 20). Of course, grammatical and lexical combinatorial models are reduced copies of two basic cognitive models.

These issues have recently been actively discussed by representatives of interpretive or cognitive semantics, who suggest that the combinatorial nature of character compatibility is most effectively studied with the involvement of corpus managers / corpus tools.

Compatibility of words, as Horina (2014) writes in her thesis (p. 17), citing Hvishiani (1979), should be studied on the basis of the cognitive unity of collocation and colligation, since the correct construction of speech is impossible only with the observance of grammatical valence, or, in terms of the corpus linguistics, - the colligative correctness.

Hvishiani (1979) explains his assumption by the fact that “the idea of the areas and rules of word usage is realized by a speaker under the influence of functioning of a certain number of phrases with that word. And in terms of a speech community, these phrases meet all the criteria of the language norm (including the sociolinguistic aspect), that is, their language status is recognized” (p. 172).

This methodological principle has led to a new vector in the study of problems of semantics: from the analysis of the meaning of an individual word (in explanatory dictionaries) to its compatibility with other words (both content words and function words) in phrases (including phrasal verbal complexes in English), called multiword expressions (MWE) in the corpus linguistics.

Due to such a turn, studies of the action of two processes in their unity were updated: colligative (a set of morphological-syntactic conditions that ensure the compatibility of linguistic units) and collocational (lexical-phrase and lexical-phraseological) combinatorics (semantic syntagmatics). Combinatorial syntagmatics becomes the leading methodological principle of cognitive-interpretive semantics. Ter-Minasova (2007; 2008) points it out in her writings and states that at present “the study of a word in only colligative manner, that is, mainly from the standpoint of its grammatical valence, is insufficient” in order to explain communicative failures and even the phenomenon of cognitive shock.

Sinclair (1996), the developer and manager of the first corpus programs for the presentation of speech samples, which were part of COBUILT (Collins Birmingham University International Language Database) project, was of the same opinion. We fully share his view that units of meaning are already embedded in ready phrases, meaning that mini-models of potential combinations of characters due to their cognitive character have the nature of ready phrasal formations reflected in consciousness of native speakers of every language (such as phrase verbs in English). That is why corpus tools include a specially designed mechanism for the automatic separation of collocations and colligations from texts.

It has long been proven, and mentioned above, that words do not function on their own, out of context, unless, of course, they are very rare names or terms. The vast majority of words are used in an already similar environment and are therefore schematised / stereotyped in the mind of a speaker and await the recurring events of their further connectivity, which are called collocations in the corpus linguistics (Sinclair, 1996). And, as experience shows, native speakers much more commonly use ready-made, commonly used templates, stencils, clichds, fixed phrases in everyday speech, rather than generate unique, unprecedented (in lexical and grammatical structure) phrases (Zymnya, 1972).

In order to understand the cognitive nature of combinatorics as a phenomenon of compatibility in the system of each language, it is necessary to find out the scientific scope of the terms “colligation” and “collocation”.

The first term “colligation” usually refers to the grammatical rules of compatibility, and therefore is used to designate the grammatical environment to which the word / sequence of words refers (grammatical company the wordkeeps) (Vlavatska, 2011, Flowerdew, 2012), or vice versa (avoids keeping). At that time, it was introduced exactly in this sense into the scientific circulation by the representatives of logical grammar to study the combinatorial capabilities of any part of language (this is what Katsnelson called valence (see: Bober, 2018)). In the most recent works on combinatorial syntagmatics, it was used the aforementioned Vlavatska (2011) when performing a valence analysis of verb compounds in English. As a result of this analysis, she developed a classification of combinatorial models of English verbs based on their valencies.

In the methodological tradition of the London Structural School, its founder, Firth (1957), actively used two terms: collocation and colligation, and believed that any meaning of a word depends on its contextual environment. However, he mostly studies the meaning of a word in collocations.

Since then, the concept of collocation has been linked to the textual influence on the choice of vocabulary to form a compound or a phrase. From this consideration, it follows that the very context of a particular situation in lexical terms can also be referred to as “collocation” in the sense of the typical and constant environment of a particular word. As a rule, collocation is considered through the relationship between individual lexical elements, and collocation is considered as the relationship between elements of a particular construction. According to Vlavatska (2011), collocations, that is, word compounds occupy an intermediate position and are at the intersection between lexicology and phraseology, as they are peripheral units both “for lexicology, which deals with mostly free lexical compatibility (syntagmatics), and for phraseology”, the object of which are connected units, that is, idioms” (p. 19-26).

But there is another opinion on this matter by Ter-Minasova (2009), who is worth agreeing, and who considers that all phrases in languages are motivated sociolinguistically, and denies the existence of absolutely free phrases that mechanically form a productive abstract construction, i.e. a superficial grammatical model of their creation, or a block diagram.

Corpus Methods and Their Effectiveness for Construction of Cognitive -- Semantic Matrices