Computing meaning. Vol. 2 (Q5915484)

This is a book containing 15 selected, outstanding papers within a new constituted research area of computational linguistics: Computational Semantics (CoSem). An interesting discovery in CoSem in the last decade (and probably the most important topic in the book) is the use of underspecification within all four important levels of meaning representation: (1) lexical (item meaning); (2) grammatical (combination of words into phrases and clauses); (3) discourse (other concrete linguistic objects); (4) context level (linguistic, discourse, nonlinguistic information, and word knowledge). The contributions in this book cluster around four major themes in CoSem: (a) lexical semantics; (b) inference; (c) underspecified semantic representation; (d) context and contextual meaning. The first introductory and guiding paper is written by \textit{Harry Bunt} (the first editor): From Lexical Item to Discourse Meaning: Computational and Representational Tools. In few words about the contents of the other 14 papers (chapters) of the book. (1) \textit{P. Blackburn} \textit{J. Bos}, \textit{M. Kohlhase}, \textit{Hans de Nivelle} (Inference and CoSem) argue that the current methods in first-order theorem proving and model generation are of direct relevance to inference for Natural Language (NL) processing. Implementation of a presupposition projection algorithm in Discourse Representation Theory (DRT) is shown to demand sustained use of powerful inference mechanisms. (2) \textit{F. Busa}, \textit{N. Calzolari}, \textit{A. Lenci}, \textit{J. Pustejowsky} (Building a Semantic Lexicon: Structuring and Generating Concepts) present a general model, based on recent extensions of Pustejowsky's Generative Lexicon theory, for the development of a set of large-scale lexical resource that are necessary in the context of the SIMPLE project. (3) \textit{F. Gayral}, \textit{D. Kayser}, \textit{N. Pernelle} (In Search of the Semantic Value of an Occurrence: An Example and a Framework) develop a framework of interpretation based on inference rather than on reference, the discussion centering on the main features of the use of French word ``examen'' taken in a school context. (4) \textit{H. Bunt}, \textit{L. Kievit} (Agent-dependent Metonymy in a Context-change Model of Communication) outline a strongly context-dependent approach to metonymy that is based on the explicit representation of contexts, viewed as the mental states of communicating agents. For a metonymic utterance, this often involves using context information in order to resolve a type mismatch between a predicate and an argument. (5) \textit{Alan Ramsay} (Weak Lexical Semantics and Multiple Views) is concerned with the ``merciless precision'' of traditional representational formalisms, arguing that much subtler and much more parsimonious accounts can be obtained if proper attention is paid to the role of inference and background knowledge in the construction of interpretations. (6) \textit{Patrick Saint-Dizier} (Lexical Conceptual Structure for Sense Variations) shows how Lexical Conceptual Structures (LCSs), paired with elements from the Generative Lexicon, can be used to deal with the meaning variations introduced by arguments with respect to the basic sense of a predicate. Different forms of underspecification, instantiations, a procedure and an implementation of the system based on constraint resolution are outlined. (7) \textit{A. Willis}, \textit{S. Manandhar} (The Availability of Partial Scopings in an Underspecified Semantic Representation) focus their investigation on the underspecified representation of quantifier scopings, facing the problem that partial scopes which can be expressed do not correspond to well-formed final meaning representations. An efficient deductive framework for determining a partial scoping is also provided. (8) \textit{A. Joshi}, \textit{K. Vijay-Shanker} (Compositional Semantics with Lexicalized Tree Adjoining Grammar (LTAG)) exploit the nature of elementary trees and derivation tree (distinct from derived tree) in LTAG. An extension of LTAG, weakly but not strongly, equivalent to LTAG, is shown to be applicable to scopal semantics without resorting to underspecification. (9) \textit{D. Crouch}, \textit{A. Frank}, \textit{J. van Genabith} (Glue, Underspecification, and Translation) discuss how one can construct Underspecified Discourse Representation Structures (UDRSs), via glue semantics. In most cases, UDRSs can be constricted in linear time, and this encoding can be used in ambiguity preserving, transfer-based machine translation, etc. (10) \textit{Mark Hepple} (A Functional Interpretation Scheme for D-Tree Grammars) suggests a new method for interpreting D-Tree Grammar (DTG) derivations, based on derived trees rather then derivation trees, and allowing the constraints on the derivation process to be eliminated. (11) \textit{D. Duchier}, \textit{C. Gardent} (Tree Descriptions, Constraints and Incrementality) present a constraint-based method for enumerating the models satisfying a given tree description, apply their method to the underspecified semantic representation of discourse, and indicate how to extend the proposed approach to support discourse level incremental processing. (12) \textit{B. Webber}, \textit{A. Knot}, \textit{A. Joshi} (Multiple Discourse Connectives in a Lexicalized Grammar for Discourse) consider the situation when the computed `discourse relation' that holds between adjacent units of text (for clauses containing more than one discourse connective) seems to allow (such) a clause to relate in more than one way to more than one other textual unit. The authors show that an approach based on lexicalized grammar and a lexical semantics for discourse connectives cannot solve the above problem. The result is an argument for syntax not stopping at the sentence boundary and for discourse not starting there. (13) \textit{J. Ginzburg} (Fragmenting Meaning: Clarification Ellipsis and Nominal Anaphora) describe how the process of utterance clarification licences a form of ellipsis which requires meanings to be stored in a highly structured fashion and to encode presuppositions concerning the structure and sound of previously occurring utterances. This approach offer a view on nominal anaphora resolution that circumvents to several semantic theories originally designed to process text/monologue. (14) \textit{Laurence Danlos} (Event Coreference between Two Sentences) discusses temporal relations between two eventualities \(e_1\) and \(e_2\) where \(e_1= e_2\) (event coreference), and furthermore, coreference between existentially quantified elements. The involved phenomena lead to the introduction and definition of new discourse relations considered in the framework of SDRT. The book addresses a large spectrum of possibly interested people, from (computational) linguists and logicians to computer scientists and NL software engineers.