Current trends in SNePS - Semantic network processing system. 1st annual workshop, Buffalo, NY, November, 1989. Proceedings (Q1188696)

[The articles of this volume will not be indexed individually.] These proceedings of the First Annual SNePS Workshop, held in Buffalo, N.Y., November 1989, contain 12 papers which report progress on the SNePS 2.1 version of a knowledge representation and formal reasoning system, which is based on the model of semantic networks. One major topic is the extension of the theory behind SNePS and implementations of proposed language extensions. The first paper sets the scene for SNePS 2.1, since it describes the features that distinguish this release from previous versions of SNePS. These include the integration of contexts, belief spaces and belief revision, hypothetical reasoning, a knowledge debugger and a graphics package a description of path-based inference and a logic-based inference to SNePS (\textit{Shapiro} and \textit{Martins}). Path-based inference, i.e., inference that relies upon the existence of a path between nodes, is studied in detail in another contribution (\textit{Cravo} and \textit{Martings}) as an efficient and more reliable alternative to performing belief revisions based on node- based inferences within the SNePS 2.1 environment. An expansion of SNePS capabilities is described with the resource-limited approach underlying the specification of LORE, a four-valued logic that, in particular, distinguishes two ``unknown'' values in terms of knowing that nothing is known, and not knowing --- with the available resources --- whether it is known. After a brief introduction to the LORE language specification framework, its potential is demonstrated with respect to hypothetical and abductive reasoning, learning with reasoning, and reasoning about its own knowledge (\textit{Mamede} and \textit{Martins}). Based on the assumption that three difference senses of order dependence of declarative knowledge (spatial, temporal, and conceptual) can be distinguished, it is argued (taking examples from well-known part --- of and is --- a relations) that any cognitively valid system dealing with changing knowledge base should maintain the conceptual order implied by the chronological order of the concepts is acquiring and that this notion of order dependence can be incorporated into SNePS without any changes to the theory of interpreter of the system (\textit{Geller}). Another paper presents an integrated model of rational cognitive agents capable of acting, reasoning, and planning about a world based on their beliefs. The notion of a transformer is introduced, a representation that specifies a belief/act transformation under the influence of forward and backward transforming procedures (under this scheme a reasoning rule in SNePS is a belief/belief transformer). In addition, a scheme is outlined for modifying the existing inference machine of SNePS to do forward/backward transformations (\textit{Kumar}). Based on recent and classical work in philosophical action theory, an attempt is made to characterize the notions of agent, agency, and action, and also make it fit into formal knowledge representation systems, e.g., the theory of representing intensional entities in SNePS (\textit{Dipert}). Various natural language processing applications have been worked out under the SNePS framework. The combination of linguistic and pictorial information is considered in a computer system that extracts visual information from parsing the text of a caption of a newpaper photograph in order to generate predictions which help to identify human faces in the photo. The resulting relational graph roughly describes the structure of the photograph and can be used subsequently by a computer vision system to further guide the interpretation of the picture (\textit{Srihari} and \textit{Rapaport}). A new unified representation for lexical information integrating morphological, syntactic, and semantic as well as encyclopedic information in terms of knowledge-based relational lexicons is proposed and illustrated with the representation of lexical relation hierarchies in SNePS networks (\textit{Nutter}). Another paper discusses issues in the representation of fictional entities and the representation of propositions from fiction, using SNePS. It briefly surveys four philosophical ontological theories of fiction and sketches an epistemological theory of fiction (still to be implemented in SNePS) using a story operator and rules for allowing propositions to ``migrate'' into and out of story ``spaces'', the latter being formally equivalent to a belief space (\textit{Rapaport}). Next, two main kinds of logic-linguistic phenomena of referential opacity are distinguished --- according to whether or not sentences are understood as referring to intensional entities --- and distinct representations of their logical form in the SNePS framework are explicated (\textit{Wyatt}). The design and SNePS implementation of a system that parses and processes transcripts of actual doctor/patient conversations is described in the following paper. The system implements an emotion profiler, i.e. it produces a profile of the patient's emotional state which is constituted by a list of emotions together with the emotions' intensity (\textit{Rapp}, \textit{Evens} and \textit{Garfield}). After discussing the psychological research on natural concept taxonomies, the use of natural categorizations in multiple sentence generation systems is described. They allow the selection of appropriate category names, provide mechanisms to help determine salience, and provide for the shallow modeling of audience expertise (\textit{Cline} and \textit{Nutter}).