Natural language and logic. International scientific symposium, Hamburg, FRG, 9-11 May 1989. Proceedings (Q1188904)

[The articles of this volume will not be indexed individually.] The papers contains in this volume were presented at the international symposium on ``Natural language and logic'', held in Hamburg, Germany, May 1989. The book covers different topics in the field of theoretical foundations of natural language understanding with emphasis on logic formalisms for knowledge representation and natural language semantics. Several contributions treat problems of semantic description from the angle of theoretical linguistics. The issue of context in semantic interpretation is raised in to papers, one discussing the necessity of a formal reconstruction of context in terms of a theory of contextual inference as illustrated by the use of indexical expression (\textit{Bosch}), the other considering context-dependent changes of the arity of formal expressions of semantic representation languages by a process termed (de-)contextualization (\textit{Dahl}). The semantic analysis of adjective constructions based on a formal notion of degree information (including scales and dimensions) for the treatment of comparatives demonstrates the usefulness of first-order predicate logic (\textit{Pinkal}). Finally, model-theoretic foundations of computational semantics for natural language processing are dealt with, first, considering current theories of linguistic form (lexical-functional grammar, head- driven phrase structure grammar) and their coupling to semantic theories, such as various kinds of situation calculi and a logic of partial information, and then turning to an extension of standard first-order logic covering the interpretation of the so-called collective readings of plural noun phrases (\textit{Fenstad} and \textit{Lønning}). A second major group of papers considers problems of semantic description and knowledge representation issues from the angle of computational linguistics. Various of these papers communicate results from the LILOG project, whose emphasis was on a proper logic foundation of text understanding and question answering processes. First, an overview of the knowledge processing, i.e. inference machinery underlying two LILOG prototypes is given (\textit{Bollinger}, \textit{Bläsius} and \textit{Hedtstück}). Two variants of \(L_{LILOG}\), the underlying order-sorted logic knowledge representation language, are discussed together with appropriate inference strategies (forward vs. backward chaining) and inference rules to implement various inference strategies (ordinary resolution extended by procedures for order-sorted unification as well as special inference rules capable of handling equality relations, sets, etc.). The violation of consistency criteria which is likely to arise in knowledge representation languages offering rich type systems is discussed by comparing \(L_{LILOG}\) and related approaches to sorted knowledge representation languages and by considering sources for inconsistencies in sort signatures, sort hierarchies, and signatures (\textit{Pletat}). Finally, a general solution is presented for the problem of integrating full first-order predicate logic in standard procedural programming languages, thus making available sophisticated knowledge representation concepts (as implemented by corresponding inference engines) in the framework of conventional programming environments (\textit{Schönefeld}). Linguistic aspects of knowledge representation and formal reasoning techniques are incorporated in the remaining papers of this volume. Issue of a dual representation of spatial knowledge, propositional and depictorial (non-propositional), are discussed with emphasis on the topological properties of depictorial representations. The usefulness of the `path' concept for the analysis of spatial natural language assertions is considered and illustrated by the in-depth analysis of the preposition `between' and the motion verb `follow' (\textit{Habel}). This part also includes the treatment of some anaphoric expressions (noun phrases) in certain referentially opaque contexts within the discourse understanding system LODUS, building upon a theory of discourse semantics that integrates concepts from discourse representation theory and situation semantics (\textit{Bernth}). LODUS' parsing machinery is described in more detail in a companion paper which propotes the use of lexically oriented, highly modularized slot grammars as a practical tool to construct broad-coverage natural language engines. Their underlying architectural considerations, parsing principles, and particular grammatical specifications for the analysis of extraposition, coordination, and implicit subjects are introduced (\textit{McCord}). Logic formalisms as a proper mean for grammar specifications are also proposed for a transfer approach to unification-based machine translation using the particular framework of lexical functional grammar (\textit{Rohrer}). As far as the issues of knowledge representation proper is concerned, one article considers the role of multiple-valued logic --- illustrated by a system of three-valued logic --- as opposed to classical two-valued logic (\textit{Schmitt}), while the final one considers rationales that lead to the determination of and selection among various new abductive reasoning schemes. Viewing particular instances of discourse understanding as an abductive inference process of generating the best explanation as to why a particular sentence is true, given what is already known, the assignment of numeric costs to axioms and assumable literal permits the specification of preferences on different abductive explanations and allows the determination of minimum-cost explanations for these abductive reasoning methods (\textit{Stickel}).