Advances in artificial intelligence. Natural language and knowledge-based systems

From MaRDI portal





[The articles of this volume will not be indexed individually.] The 14 papers contained in this volume were originally presented at various conferences on artificial intelligence held in Israel from 1987 to 1989. Major topics covered are methodological as well as applicational aspects of knowledge-based systems and natural language processing. Papers related to natural language processing include a preference-based parsing algorithm for identifying the unexpected meaning of linguistic jokes (\textit{Ephratt}), a formal approach to Hebrew morphology with focus on inflexion and derivation, which is based on the notion of root pattern arrays (\textit{Ornan}), and an outline of the methodology for deciphering a corpus of the ancient Etruscan language (\textit{Nissan}). Based on a fundamental critique of feature-based interpretative and model-theoretic approaches to natural language semantics, an alternative is proposed that exploits the sublanguage orientation of natural language processing systems and the ability to predefine the granularity of the required meaning analysis by combining the ontological and a semantic network based approach (\textit{Raskin}). Within the text generation task, a procedure is introduced which anticipates the effect of a given message on a model of listener's conjectured beliefs, and proposes rhetorical devices to preclude possible adverse effects resulting from a mismatch between the communicative goal of the text and the listener's belief (\textit{Zuckerman}). Methodological as well as applicational topics in relation to knowledge- based systems are dealt with from different perspectives. A prototype- based design shell is presented which provides complementary types of design knowledge and uses them for prototype selection and prototype refinement (\textit{Oxman}). The design principles of a knowledge-based consulting and tutoring system that communicates with a finite element software package and offers facilities for effectively training users in the applications of the system's resources are considered, with emphasis on its natural language understanding and planning components (\textit{Bykat}). A computationally feasible application of Bayesian inference to the problem of estimating from past data the possibilities that patients with certain symptoms might have certain diseases is described in an expert system application, without assuming independence of the symptoms (\textit{Gammerman} and \textit{Thatcher}). Two papers address the problem of formulating practical semantics for constrained-based systems as an improvement over standard production system techniques. Central to this problem is the issue of partial solutions which give rise to partially ordered structures of representational states. These are conceptually rather close to states of incomplete and uncertain information which are likely to occur in the reasoning of knowledge-based systems, given realistic application domains (\textit{Battle}; \textit{Morgan} and \textit{Gammack}). An incremental, heuristic problem-solving approach to scheduling, considering software project planning and scheduling more as a schedule revision problem than a schedule generation problem, is proposed with focus on the design of a set of schedule revision operators and a control strategy for applying these operators during the scheduling process (\textit{Safavi} and \textit{Smith}). Finally, emerging principles for knowledge base design and models for knowledge base architectures are discussed (\textit{Golumbic} and \textit{Grinberg}). An appendix contains the edited transcript of a panel discussion on benefits and problems that are expected from the transition from databases to knowledge bases, given at the `3rd International Conference on data and knowledge bases', held in Jerusalem, Israel, June 1988. Papers dealing with machine learning focus on a theoretical model for a system that learns from its own experience in playing board game (\textit{Ben-Porat} and \textit{Golumbic}), and an incremental conceptual clustering algorithm that reduces input-ordering bias together with a formal analysis of its computational complexity; the procedure incrementally updates the hierarchy of concepts as new items are added, thus correcting the bias due to the insertion order (\textit{Maarek}).











This page was built for publication: Advances in artificial intelligence. Natural language and knowledge-based systems

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3999089)