LINGUIST List 9.1813

Sun Dec 20 1998

Review: Cook: Case Grammar Applied

Editor for this issue: Andrew Carnie <carnielinguistlist.org>




What follows is another discussion note contributed to our Book Discussion Forum. We expect these discussions to be informal and interactive; and the author of the book discussed is cordially invited to join in. If you are interested in leading a book discussion, look for books announced on LINGUIST as "available for discussion." (This means that the publisher has sent us a review copy.) Then contact Andrew Carnie at carnielinguistlist.org

Directory

  • Tania Avgustinova, Review: Cook: Case Grammar Applied

    Message 1: Review: Cook: Case Grammar Applied

    Date: Fri, 18 Dec 1998 09:24:28 +0100
    From: Tania Avgustinova <taniacoli.uni-sb.de>
    Subject: Review: Cook: Case Grammar Applied


    Walter A. Cook. S.J. (1998) "Case Grammar Applied" A Publication of The Summer Institute of Linguistics and The University of Texas at Arlington, 1998, 271 pages

    Reviewed by Tania Avgustinova, University of Saarland.

    SYNOPSIS

    The book is intended as a companion volume to "Case Grammar Theory" (1989) by the same author, and is devoted to the application of the model developed in this earlier publication to English text analysis. This in particular means that the main stress of the present work is on methodological issues, which are presented on an extended textual analysis of Hamingway's "The Old Man and the Sea".

    The actual goal of the discussed monograph is twofold: for the grammar - to present a method for sentence semantics which describes the meaning underlying each simple sentence; for the lexicon - to present a method for further defining each sense of each verb / predicate in terms of semantic classes (case-frame-based typology). Each clause in its underlying structure is reduced to a kernel sentence which is defined as a simple complete active-voice affirmative statement. All texts are reduced to a sequence of kernel sentences

    The model used is the Case Grammar Matrix Model. It presents a clear predicate-argument structure, builds a lexicon which distinguishes various verb senses, and not only describes arguments occurring in the text, but through covert roles, describes implicit arguments. This model is supplemented by the use of conceptual graphs, following Sowa (1984). Each verb type has a generalised conceptual graph to indicate the predicate-argument structure; and each example in the text contains the canonical graph of that sentence using the verb and nouns as concepts, and case labels as the verb-to-noun relations.

    The book is organised in the following way.

    Chapter 1 (pp.1-54) presents the Case Grammar Matrix Model in summary form and introduces conceptual graph representation. The next four chapters treat verbs / predicates in four distinguished semantic domains.

    Chapter 2 (pp.55-90) deals with the Basic domain that includes all State, Process, and Action verbs which use only the Agent and Object cases.

    Chapter 3 (pp.91-126) deals with the Experiential domain, the domain of sensation, emotion, cognition, and communication. It involves State, Process, and Action verbs which include an Experiencer case in their descriptions.

    Chapter 4 (pp.127-42) deals with the Benefactive domain, the domain of possession and transfer of property, and describes State, Process, and Action verbs that include a Benefactive case.

    Chapter 5 (pp.143-82) deals with the Locative domain, the domain of physical location and movement, and describes State, Process, and Action verbs that include a Locative case.

    Chapter 6 (pp.182-220) deals with other elements involved in the logical representation of sentences, including tense, aspect, modals, performatives, and negatives, showing how these elements may be included in the logical structure, and ending with a sample sentence parse using Case Grammar.

    Chapter 7 (pp.221-46) summarises the analysis of more than 500 examples in chapters 2 through 5 and demonstrates the verbal hierarchy expressed by the twelve cells of the Case Grammar Matrix, organised by verbal domain, verb type, and argument structure. Each of the verb types is described together with its conceptual graph, its frequency of occurrence, its subtype, and its defining characteristics.

    The appendix to the text contains an alphabetical lexicon, listing all of the verbs / predicates in the examples together with their case frames (pp.247-52), and a case lexicon, with the verbs sorted by case frame together with references for each verb to the pages where the use of the verb is exemplified (pp.253-60). Finally, there is a list of references (pp.261-6), and an index (pp.267-71).

    CRITICAL EVALUATION

    Case Grammar (CG) works with labelled predicate-argument structure, and thus, in its core, belongs to the Dependency Grammar paradigm. CG develops a semantic valence system that describes the logical form of a sentence in terms of a central predicate (usually and typically a lexical verb, but also a predicate adjective, a predicate noun, or a predicate adverb) and a series of case-labelled arguments (nominal, adverbial) required by the meaning of that predicate. So, the type of case considered in the book is 'governed case', and in fact, nothing is said about the way CG model would treat 'concordial case' - cf., e.g., Blake (1994) for a detailed discussion of these notions. There is no case concord in English and, hence, the challenges posited by 'concordial case' are trivially out of the scope of the presented analysis (which is based exclusively on English data).

    The ambition of the proponents of CG is to develop a semantic interpretation system that is universal across languages, and not tied to the syntax of any particular sentence. On the basis of the presented extensive data analysis it is claimed that the five case labels used in the work (i.e. Agent, Object, Experiencer, Benefactive, and Locative) are necessary and sufficient for the description of all the verbs / predicates in the language (in this case English). The possibility of creating different lists of cases is left open, whereby full translatability is theoretically expected between any consistent list of cases and the one employed in the CG matrix model.

    Both the predicate and its arguments are viewed as concepts (i.e. the conceptual universe is made up of verbs which describe states or events, and nouns which describe things), while the case role labels indicate relations (which arguments bear to their predicates). CG is written in conceptual graph format by placing the concepts (predicates, arguments) in boxes and relations (case roles) in circles. The arrows in the notation point away from the predicate which is the source of the case relations. An advantage of such an approach is the possibility of defining the default position of the lexical predicate in its case frame, i.e. with respect to the arguments it governs.

    In the discussed book, a fairly credible ontology of predicators is achieved by hierarchically sorting them according to domain as Basic, Experiential, Benefactive, Locative, (and possibly) Time, and within each domain as State, Process, Action. Further sub-classification of the predicates is based on the number or the position of the arguments. The resulting hierarchical taxonomy is similar to a thesaurus in which all entities are organised into semantic domains. The organisation of the semantic domains is worked out in detail and with precision.

    The question of case inventory is central to CG, as well as to any theory working with the notion of case. Also methodologically, a clear distinction must be made between essential cases, which are required by the meaning of the predicate, and modal case, which are mainly adverbial adjuncts. Only essential cases can be used for describing predicates. Despite of the lack of universal agreement on the number of case role labels or the way in which they are defined, the author gives a clear step-by step method, accompanied by various concrete tests, how to simplify in practice the case assignment. Thus, the labelling is deferred until three basic questions are answered: How many arguments are required by the verb? What verb type - state, process, or action - is in the structure? To what semantic domain does the verb belong? Then, the naming of the arguments can be simplified by a set of principles, in combination with a set of tactics for the formation of case frames. Within case frames, cases are listed left-to-right according to a subject choice hierarchy (Agent-Experiencer-Benefactive-Object-Locative) which is merely a generalisation covering the unmarked choice. Marked choices which violate the subject choice hierarchy are indicated in the lexicon by changing the order of cases in the case frame. Possible variations involve "equatives" (in copular constructions), regarded in this model as Double-Object frames in the Basic domain, and rank shifts in the subject choice hierarchy.

    In the lexicon, predicators are classified according to case frames. The case frame is understood as a configuration of one to three cases that are required by the meaning of the verb (or more generally, the predicate). Let us remind, however, that, in lexical semantics research, cases are known where the frame would contain up to five slots. A famous example is the verb 'to rent' (Russian: arendovat' - cf. Apresjan (1974) p.134) involving the following arguments: (1) who, (2) what, (3) from whom, (4) for how much, (5) for how long.

    Special attention is devoted to the possibility of lexicon organisation in terms of derivations intended both to represent linguistic generalisations and to simplify the lexicon (sections 1.13 - 1.15). However, the method followed in the work is listing each item separately and supplementing the lexicon with redundancy rules that relate these lexical entries to each other. Also, each sense of each predicate is treated as a separate item with its own case frame.

    In order to set the guidelines for applied CG textual analysis, a concise introduction to the principles of lexical decomposition and to the interpretation of covert case roles is given. Lexical decomposition is understood as the process of analysing predicates as consisting of more basic atomic predicates. With some sentences this is unavoidable for determining the actual predicate-argument structure.

    Covert case roles, which are required by the meaning of the predicate, are sometimes (as in the case of partially covert (deletable) case roles) or always (as in the case of totally covert (coreferential and lexicalised) case roles) missing in the surface structure. Since the CG analysis advocated for in this book maintains the obligatory object hypothesis, covert roles assume greater importance. In all cases where the Object role can be sometimes deleted, or can be coreferential with another role, or can be lexicalised into the predicate, a deeper analysis is needed to find the obligatory object. Certainly, the lexicalisation of the manifestation of propositional, essential roles is of primary interest in the context of revealing the central predicate in a sentence.

    The assumption that a predicate, even with covert (hidden) case role, has to be defined in terms of its full complement of case roles requires a clear distinction between deletable, coreferential and lexicalised roles. The author offers in this respect not only theoretical background but also concrete tests, procedures and instructions to guide and facilitate practical analysis. Two methodological principles are postulated ensuring that all conceptual relations flow from the central verb, and that deletable roles are included in case descriptions.

    A major asset of the book is, with no doubt, the extensive textual analysis performed with precision, consistency and conformity to the postulated principles and theoretical assumptions. The limits of the approach are realistically recognised by the author, and are stated explicitly in the appropriate places throughout the presentation. The reader will find a well-developed and detailed ontology of predicates, which covers not only verbs but also predicative adjectives, predicative nouns and predicative adverbs. The most important linguistic phenomena are considered in a systematic and easy-to-follow way. This makes the book a valuable guide to a practical sentence analysis, as well as a useful reference material for research purposes and computer applications.

    REFERENCES

    Apresjan, Jurij D. 1974: Leksicheskaja semantika. Sinonimicheskie sredstva jazyka. Moskva: Nauka

    Blake, Barry J. 1994: Case. Cambridge Textbooks in Linguistics.

    Cook, Walter A., S.J. 1989: Case grammar theory. Washington, D.C.: Georgetown University Press.

    Sowa, John F. 1984: Conceptual graphs: Information processing in mind and machine. Reading, Mass.: Addison- Wesley.

    =============== Tania Avgustinova, Ph.D. Computational Linguistics, University of Saarland Postfach 151150, 66041 Saarbruecken, Germany taniacoli.uni-sb.de, http://www.coli.uni-sb.de/~tania/ (+49) (681) 302.4504 (phone) (+49) (681) 302 4115 (secretary) (+49) (681) 302.4700 (fax)