The matching of AGFL subcategories to Russian lexical and grammatical groupings Irina V. Azarova Department of Mathematical and Applied Linguistics Saint-Petersburg University (Russia) azic@bsr.spb.ru Abstract The problem of enumerating a set of grammatical categories and arranging their taxonomy is a classical task for any linguistic description, the formal grammar representation being the most strict. It is vital for the formal description to characterise the scope of its relevance to language expressions: valid grounds for grammar organisation from one side, and the number of grammatically acceptable language expressions from the other side. Concerning the latter, it is well known that a lot of expressions in such languages as Russian with its free word order are constructed syntactically according to excessively loose rules. The common attitude to such situation in grammatical theory is to distinguish two layers: syntactic structure and sentence functional perspective one, each being described by its own grammar component. The weak point of such position is that there is a huge intersection between devices in both. So the only possible way is to choose the paradigmatically fixed constructions for grammar incorporation, but this may have subjective colouring. The conventional lexical groupings such as animate nouns, action verbs, act verbs are considered concerning AGFL syntactic categories: verb personality/ impersonality and transitivity/ intransitivity, substantive cases. The grammatical description of a language, especially syntactic one, may vary greatly in that what set of language features are described and for what objectives. Thus a number of descriptions are possible at once: constructive, communicative, semantic, logical, functional, expressive and so on. We choose, however, one or several perspectives in accordance with our main goal. The problem of defining the main syntactic item simple sentence is rather typical. In the AG-80 [5] at the beginning of the sentence section the objective is postulated to describe all grammatical units constructed as relatively independent of linguistic and extralinguistic context, well-formed grammatically, predestined to communicate messages. These special items are called sentences, they are opposed to the wide stock of any possible linear speech fragments executing the communicative function in some situation. Trying to describe what distinguishes sentences and fragments of speech, AG-80 comes to the conclusion that the former are relatively context-free while the latter are context-dependent; the former, therefore, are well-formed grammatically, that is, constructed after some limited in number valid grammatical patterns. The following material, however, shows us that sometimes contextually dependent sentences may be considered grammatically well-formed. For example, in the case of the elliptic realisations, which are conceivable in the context. 1
It became clear that the division of expressions into grammatical and situational ones may be caused by another reason the opposition between two different strata of sentence formation: constructive versus communicative. Many linguists hold this position [4]: the sentence in Russian is constructed by combination of words and word-phrases, this is done on the constructive syntactic stratum, then for needs of communication words may be rearranged, receive special intonation and stress, and this is done on the communicative syntactic stratum. It is clear, however, that in the process of speech generation the speaker first decides what is known, thus thinking about the communicative structure, and then constructs an item (sentence or expression). It is unlikely that generation of an expression is more difficult than that of a sentence. So we can draw a line between well-formed grammatically constructions and non-well-formed ones by its recurrence: those constructions that occur frequently are included into syntactic paradigm, they usually lose the special connotational meaning; those that appear occasionally are linked syntagmatically with words, included into the sentence, and a more wide context. The outstanding Russian linguist Lev Shcherba presented once a similar idea pointing out that every regular linguistic process, notwithstanding its lexical or syntactic nature, should be put into the grammar section of the linguistic description, and irregular phenomena are to be dealt with in the dictionary section. For example, in Russian nouns, denoting containers, used regularly in the sense the content of the container : GZeblv\h^u\qZcgbd±<kdbiylblvqZcgbd Pour a water into the pot Boil the pot = Boil the water in the pot This phenomenon is accounted for sometimes in the dictionary entries, but it should be inserted into syntactic and semantic modules of the grammar. Having in view the relation between lexical and grammatical, we tried to use some aspects of the lexical meaning in order to describe grammatical features of words in syntactic constructions. We may base upon conventionally distinguished lexical groupings inside the Russian parts of speech, and those that are represented in a thesaurus composed after the model of English WordNet. Let s consider in what manner we can match lexical groupings to grammatical subcatagories of the AGFL grammar [3] for Russian. The first problem is connected with the question whether there is such category for nouns as Animate/Inanimate in Russian. Some suppose that this is not a grammatical category but a lexical grouping, though causing the change in the grammatical paradigm: for animate nouns the accusative coincides with the genitive in plural, while for inanimate nouns with the nominative. Moreover, the division of nouns into animate and inanimate nouns has somewhat illogical character: the plants are inanimate, dmdez (a doll) is animate, f_jl\_p ihdhcgbd (dead man) is animate, and ljmi (body, corpse) is inanimate, the names for collectives of 2
people (crowd) is inanimate; there are alternatives of interpretation for such words as \bjmk (virus), which may be either animate, or inanimate. We may mention absence of the special inflexion for these cases (its referential character) as some additional reason for its noncategorical nature. Moreover, in some verbal constructions such as ihclb \ khe^zlu (to become a soldier literally «to go to soldiers»), the names of professions and positions, which are usually animate nouns, behave as inanimate: instead of a genitive-like form for the accusative we see a nominative-like form. The variation of nominative/genitive forms may have another sense connotation. Some verbs in Russian may be transitive or intransitive. For example, gzj\zlvp\_lup\_lh\ (to pick up flowers), the genitive form having the quantitative colouring «several». In another construction `^ZlvibkvfhibkvfZ (to wait for the/a letter) the nominative-like form of the accusative connotes the definiteness of the object while the genitive-like form indefiniteness, this meaning being rather optional for Russian grammar. In both cases only inanimate nouns may have such variants of the form. All these examples might be interpreted as showing the virtual character of the accusative in Russian substantives if it were not for feminine gender nouns, for which there exists a special inflexion in the singular. We have two variants of analysis. 1. If we consider the Animate/Inanimate to be a full grammatical category, we should insert it beside categories of case, number, and gender for nouns. (1) CASE :: nom; gen; dat; acc; abl; loc. NUMBER :: sg; pl. GENDER :: masc; fem; neut. ANIM :: anim; inanim. The rule for noun form generation includes two source elements the noun stem and the noun inflexion: (2) NounForm (CASE, NUMBER, GENDER, ANIM) : NounStem (GENDER, ANIM, DEC_TYPE), NounInflexion (CASE, NUMBER, GENDER, ANIM, DEC_TYPE). In this case we may list the accusative inflexions among the rest without even paying attention to its coincidence between accusative and nominative or genitive forms. In the construction ihclb\khe^zluthe verbal phrase should be described with 2 alternative types of government for inanimate and animate nouns in the plural: 3
(3) VerbPhrase ( MVM) : VerbForm«090VSDFH \ªVSDFH1RXQ)RUP (nom, pl, masc fem, anim) 1 ; VerbForm«090VSDFH \ªVSDFH1RunForm (acc, pl, GENDER, inanim). We see that a nominative form appears after the preposition that is highly improbable for such inflexional languages as Russian. 2. Another variant of analysis is to split the accusative form into 2 subforms coinciding with the nominative and the genitive. Thus we should insert other categories instead of (1): (4) CASE :: OCASE; ACC. OCASE :: nom; gen; dat; abl; loc. ACC :: acc_n; acc_g. NUMBER :: sg; pl. GENDER :: masc; fem; neut. The construction of case noun forms except the accusative are done according the same rule as (2) without an extra category ANIM: (5) NounForm (OCASE, NUMBER, GENDER) : NounStem (GENDER, DEC_TYPE), NounInflexion (CASE, NUMBER, GENDER, DEC_TYPE). When defining the accusative forms, we may use ready forms for the nominative and the genitive except the nominative-like form for the feminine singular of the regular declension type: (6) NounForm (acc_n, pl, GENDER) : NounForm(nom, pl, GENDER); NounForm (acc_n, sg, masc neut) : NounForm(nom, sg, masc neut); NounForm (acc_g, NUMBER, GENDER) : NounForm(gen, NUMBER, GENDER); It is the accusative singular for the regular feminine declension that requires a special inflexion, therefore, an additional rule is necessary: (7) NounForm (acc_n, sg, fem) : NounStem (fem, reg), NounInflexion (acc, sg, fem, reg). Thus splitting the accusative form, we receive several advantages in morphology description. For example, in pronominal declension we see the coincidence of both subforms (acc_n and acc_g) with the genitive form. These forms are used in noun phrases notwithstanding the inanimateness of the dominating noun: <b^_e Bo [hevrb_ ^hfz Verb (trans PronounForm(acc_g acc_n AdectiveForm(acc_n NounForm(acc_n 1 We skip those verbal categories which we didn t use. 4
We should, however, somehow define which subform (acc_n or acc_g) to choose in combination with transitive verbs: in case of an animate object it should be acc_g, otherwise acc_n. One of the possible answers is to insert something like selection rules from another level of a linguistic description, referred to an Animate/Inanimate feature of nouns as a lexical grouping characteristic. This way of splitting the case form is applicable not only to the accusative but to other cases as well. For example, the local case is sometimes called prepositional due to its obligatory combination with a preposition. Some prepositions in this combination are linked with a specific ending, but such prepositions as \ (in) or gz (on) may have an ending coinciding with the dative form. In this case the form is local proper. The case variant is defined only for singular forms of the regular masculine declension, having two additional selectional restrictions: (1) such words usually have a simple stem, and in most cases a monosyllabic one; (2) these nouns belong to several lexical groups: locality names fhkldjzc[_j_], parts of human body ghk jhl eh[, substances e_^ imo f_^, people groups ihed kljhc, abstract notions hlimkd, or they appear in some idiomatic collocations. We can split the locative case into 2 subforms: (8) CASE :: OCASE; ACC, LOC. OCASE :: nom; gen; dat; abl. ACC :: acc_n; acc_g. LOC :: loc_p; loc_d. For the locative case it is impossible to unite rather different types of nouns under one heading of some category. These lexical groupings should be referred to as a selectional rule. Moreover, the mentioned lexical groups don t act in the same manner concerning the usage of the local subform. Sometimes, Russian speakers may be uncertain about the reality of the dative-like subform for a particular noun. There is the similar partition for the genitive case though this division is now disappearing in the modern Russian. The variant subform of the genitive has the quantitative meaning («small amount of something») and refers to the dative form. Usually this case subform appears for the singular of the masculine regular declension, selection rules including partially such lexical groups as substance names and abstract nouns. In this manner we may describe variant forms for Russian cases. Being morphological phenomena, they are to be dealt with in the morphological part of the grammar description, but there should be some device for connecting particular forms with lexical groupings described by features, that is, selectional rules in syntactic module of grammar description. In some 5
cases we d better use a list of nouns, it being some sort of the dictionary (lexicographic) definition. Another problem is in what manner we should represent semantic features that are extracted through the morphological and syntactical analyses. REFERENCES 1. Baker M.C. Incorporation: A theory of grammatical function changing. Chicago, 1988. 2. Kiparsky P. Agreement and linking theory. Stanford, 1989. 3. Koster C.H.A. Affix Grammars for natural languages // Attribute Grammars, Applications and Systems, International Summer School SAGA, Prague, 1991. 4. Raspopov I.P. Syntactic Theory Essays. Voronezh, 1973. 5. Russian Grammar: Syntax. M., 1980. V. 2. 6