Date: Fri, 1 Jul 2005 16:48:08 +0200 From: Milena Slavcheva Subject: Cognitive Modeling and Verbal Semantics: A Representational Framework Based on UML
AUTHOR: Schalley, Andrea C. TITLE: Cognitive Modeling and Verbal Semantics SUBTITLE: A Representational Framework Based on UML SERIES: Trends in Linguistics. Studies and Monographs 154 PUBLISHER: Mouton de Gruyter YEAR: 2004
Milena Slavcheva, Linguistic Modeling Department, Institute for Parallel Information Processing, Bulgarian Academy of Sciences, Sofia
The book is a monograph that introduces a representational framework for verbal semantics, the Unified Eventity Representation (UER). The approach is rather innovative, since it introduces object- orientation as a new paradigm to linguistic semantics, and proposes graphical representation based on the Unified Modeling Language (UML) -- "the current lingua franca for the design of object-oriented systems in computer science" (p.1). The approach is a cognitive one and the semantic representation is elaborated from a theoretical perspective. The aim is to develop a framework for the modeling and description of verbal semantics, which is intuitive and formal at the same time.
The monograph consists of ten chapters, an appendix (containing the notational elements of the UER), extensive notes, a list of references, an index of names, an UER index, and a subject index. A preface, a list of figures and a list of tables are also provided at the beginning of the book.
Chapter 1 is a short introduction where Andrea Schalley points out the main objectives of the proposed representational framework and summarizes the content of the book chapters.
In Chapter 2, "Survey of research positions", A. Schalley overviews the most prominent linguistic research positions relevant for the UER and presents the UER's positioning with regard to the issues discussed. She considers approaches to the lexical semantics of verbs, thus elaborating on and comparing key concepts and representational devices like semantic features, semantic markers, stereotypes, prototypes, semantic roles and protoroles. Aktionsart and its relation to verb classification are also discussed. The structural relations within the lexicon are represented by their notable manifestations like sense relation, polisemy, and lexical fields. A. Schalley discusses five different approaches to decompositional semantics, starting with the older and rather influential Generative Semantics, and continuing with the prominent Jackendoff's Conceptual Semantics, Fillmore's Frame Semantics, Wunderlich's Lexical Decomposition Grammar and Pustejovsky's Generative Lexicon. Another topic relevant to the UER and included in Chapter 2 is that of semantic primitives or semantic primes and the Natural Semantic Metalanguage originated by Wierzbicka.
Chapter 3, "Introducing the UER", outlines the foundation of UER as a graphical formalism, which adopts basic constructs from the Unified Modeling Language (UML), but also makes adaptations and extensions to UML to ensure the linguistic applicability of UER. The UER architecture is defined as a hierarchy of abstraction layers.
Chapter 4, "Basic concepts of the UER" introduces general-purpose mechanisms that provide the extensibility devices for the UER. The basic UER concepts are represented. From one side, there are the static structure elements that typically model ineventities (entities which typically occur as participants in an eventity, that is, generally speaking, an event). On the other side, there are the dynamic structure elements that model the dynamic, behavioral aspects of eventities (i.e., events).
Chapter 5 presents advanced concepts of the UER, which meet the requirements of the linguistic application domain. The syntax and semantics of each modeling element are provided together with an example for illustration. The fundamental UER concept of EVENTITY FRAME is introduced, as well as elements modeling semantic relations among the members of the eventity concept.
Chapter 6 provides an interpretation of UER concepts, presenting motivations and reasons for certain design decisions. The fairly permissive character of UER is pointed out which means that "the different concepts' member sets are not definitely specified" (p.181) and it is a matter of further typological research to establish these sets in order to handle different linguistic phenomena. A. Schalley makes proposals for handling key phenomena discussing EVENTITY FRAMES as central cognitive units, ATTRIBUTES as semantic features of participants, CLASSES as descriptors for sets of OBJECTS, that is "sets or collections of entities with similar structure, behavior, and relationships" (p.193), PARTICIPANT TYPES as references to ontological categories, GENERALIZATIONS as inheritance relationship promoters and taxonomy builders, and so forth.
Chapters 1-6 (summarized up to now) provide the formal specification of the UER, while Chapters 7-9 make account of its application.
In Chapter 7 general issues are discussed concerning the modeling capacity of UER. Directions for further research are also indicated. Along the lines of the UER decompositional approach to verbal semantics, the significant issue of the definition and utilization of semantic primitives is considered, as well as the devices used for modeling granularity and constraining the representation of verbal semantics. A very important point in this chapter is the vision for extending UER in order to account for compositional semantics. The author is aware of the great amount of work to be done in this direction, but nevertheless she marks the milestones of the potential development.
Chapter 8 and Chapter 9 demonstrate two different applications of the UER, in theoretical and in descriptive linguistics respectively. The first application, deployed in Chapter 8, is "a preliminary development of a linguistically relevant eventity classification based on the UER" (p.251). The classification systematizes cognitively primary classes, reflects fundamental Aktionsart distinctions, and takes advantage of the graphical representation of abstracted semantic content. Chapter 9, entitled "Application II: The polysemy of German "setzen" is an attempt to demonstrate the "broad applicability and potential of the UER as a valuable tool for semantic analysis and semantic representation" (p.325). First the eventity concept encoded in the English verb "put" is represented by the respective UER diagrams. Then the polysemous readings of the German verb 'setzen' are described employing the modeling devices of the UER formalism.
Chapter 10 is an epilog where A. Schalley recapitulates the characteristics of UER and stresses upon the main qualities that the representational framework strives to achieve: expressiveness, universality, cognitive adequacy and, where possible, practicability.
A. Schalley's monograph is distinguished for its originality and innovation. The approach is attractive because it utilizes the thinking and the design of object orientation -- at present the most prominent paradigm in programming languages. The approach to verbal semantics is a cognitive one and thus relies on intuition and world adequate conceptualization. At the same time UER is formally adequate, the syntax and semantics of its modeling elements are explicitly specified, a metalanguage guarantees the well-formedness of the building blocks of the representation. The formal soundness is further guaranteed by the utilization of representational devices from the Unified Modeling Language (UML) as formulated in the UML specifications, to which the UER linguistically suitable adaptations and extensions also conform. In such a way, although the complete formal semantics of UER has not been fully elaborated (as is also the case with the still developing UML standard), the UER conforms to the requirement for formal correctness and reliability.
UER is a graphical formalism, "a mixture of graphical, two-dimensional elements as well as elements expressed in either natural or logical languages and thus linear textual constructs" (p.80). I agree with A. Schalley's statement that such formalism provides for linguists an intuitive and at the same conceptually rigorous modeling device. The object- oriented modeling together with the application of UML, the nearly established (i.e., officially published in the near future) standard for design, visual representation and documentation of object-oriented systems in computer science, bring the UER models near to the most up-to-date implementation, although the investigation presented in the monograph does not deal directly with implementation.
The approach, as presented in the monograph, is purely theoretical and purely semantic. Being a practitioner, I feel a bit uneasy about the amount of efforts necessary to put the UER into practice. I am also concerned about the interface to syntax, that is, to the anchors of real life language productions. But I am well aware of the fact that this is the case with all emerging linguistic theories and formalisms. After all the present investigation is the necessary theoretical foundation of a newly developed representational framework of semantics. A. Schalley herself conceives the UER as a first step towards a broader treatment of language data and elaboration of applications.
A special merit of the UER is the possibility to capture variable granularity of the semantic description. The UER metamodel with its multiple layers of abstraction guarantees the deployment of a type hierarchy and makes use of the distinction between type and instance, which is vital to building effective linguistic models allocated to applications in computer systems. The generalization mechanism is a fundamental one and allows the user to adjust the granularity of the linguistic modeling depending on the application requirements.
The last two sentences in the last chapter of A. Schalley's monograph say: "However, many topics were still beyond the scope of this investigation, and it was not possible to test the UER for several different languages. We hope that investigations that use the UER will be carried out, thereby closing the gaps and stimulating further development of the UER." (p.328)
I think that it is worth trying to apply this promising representational framework.
ABOUT THE REVIEWER:
ABOUT THE REVIEWER
The reviewer, Milena Slavcheva, is a specialist in building formal models of language for software applications and in producing large- scale language resources (lexicons and corpora). She is interested in the study and formal representation of verb-centered structures at different levels of linguistic analysis, as well as in the development of real linguistic components utilizable in the automatic processing of language.