FOXEA@VTVAX3.BITNET (05/04/88)
IRList Digest Saturday, 30 April 1988 Volume 4 : Issue 29 Today's Topics: Abstracts - Abstracts of interest from Susanne Humphrey News addresses are Internet or CSNET: fox@vtopus.cs.vt.edu BITNET: foxea@vtvax3.bitnet ---------------------------------------------------------------------- Date: Fri, 29 Apr 88 21:55:41 EST From: "Susanne M. HUMPHREY" <humphrey@MCS.NLM.NIH.GOV> Subject: abstracts of possible interest [Note: The usual copyright and other restrictions go along with this submission -- see earlier similar submissions by Susanne for the actual wording. - Ed.] AN University Microfilms Order Number ADG87-27184. AU ABRAMS, DAVID HOWARD. IN Georgia Institute of Technology D.B. 1987, 220 pages. TI A REFERENCE MODEL FOR HETEROGENEOUS DATA BASE MANIPULATION AND AN EXPERT SYSTEM PROTOTYPE FOR HOMOGENEOUS DATA BASE MANIPULATION. SO DAI V48(09), SecB. DE Computer Science. AB A reference model for heterogeneous database manipulation (HDBM) is presented. The model provides a generalized approach for allowing applications software to perform data manipulation on centralized or distributed heterogeneous database systems. The HDBM model has been developed using layering and techniques similar to those used in the International Standards Organization's reference model for open systems interconnection. A working prototype system, based on the HDBM model and called the Integrated Data Delivery Expert (IDDE), was developed as a series of layers using rapid prototyping methodology and expert systems technology. The IDDE serves as a front-end system for manipulating data in a distributed, relational, database system environment. The IDDE software is developed as four layers: applications interface; data presentation services module; data control services module; and database services module. The IDDE model was tested by retrieval of data from database systems that are accessible to it. The prototype has been shown to process data retrieval requests successfully from the distributed, relational, database system environment. Sample IDDE executions, as well as the prototype source code, are included. The dissertation addresses several future research possibilities and implications. AN University Microfilms Order Number ADG87-26875. AU CHEN, HSING LUNG. IN Illinois Institute of Technology Ph.D 1987, 175 pages. TI OBJECT-ORIENTED ALERTER SYSTEM DESIGN. SO DAI V48(09), SecB. DE Computer Science. AB Database systems are usually 'passive.' Database alerting techniques provide a database system with the capability to take actions by itself. Hence, the database system with alerting techniques can play a more intelligent role. The primary objective of this research effort is the development of a distributed intelligent database system. This design problem encompasses the issues of the development of methodology to decompose the complex alerters into simple alerters, the investigation of algorithm for allocating the simple alerters and design of protocol for ensuring that alerters correctly monitor database. The proposed system is useful in office information systems, decision-support systems and user-friendly specification of 'chip' expert systems. We approach the problem of monitoring database updates by using the object-oriented approach. A methodology is proposed to decompose a complex alerter into several objects. These objects form a tree-structure. Each object can be considered as monitoring a virtual database view. If the database view is updated, the updated message is sent up the tree for further monitoring. The top object can check whether alert condition is met and then invoke the alert action. An alerter can usually be expressed in terms of an associated query on database view. The conventional approach for implementing the complex alerters is that the database view is constructed whenever related elementary relations are updated. The response time of view-construction approach is longer than that of object-oriented approach. However, object-oriented approach needs much more space to store customized alerters. Another approach we proposed is the combination of both approaches. The performance of three approaches is compared. The combination approach can achieve better time-space tradeoff. If the object-oriented approach is applied in the centralized computer systems, the complex alerter can be correctly invoked. But in the distributed systems, the actions which should be invoked may be missing or the extraneous actions are invoked. Some concurrency-control methods are proposed to make sure that the complex alerters monitor the database updates correctly in the distributed systems. Some allocation algorithms are also proposed to allocate the alerting objects so that the response time is minimal under object-number constraints. AN University Microfilms Order Number ADG87-26878. AU HARR, HENRY MAXIM. IN Illinois Institute of Technology Ph.D 1987, 160 pages. TI ABF: AN EXPERT SYSTEM FOR OFFICE AUTOMATION AND AN INTERPRETER FOR LEGAL DOCUMENT CONSTRUCTION. SO DAI V48(09), SecB. DE Computer Science. AB The ABF system creates an environment for document construction that allows people who have never used a computer before to create complex client-specific drafts. Legal experts develop model documents and templates in ABF; then law students and paralegal personnel process these documents in the same system, answering system-generated questions about the client to produce a custom-tailored version. The ABF system protects the user from the operating system in every conceivable way. It provides facilities for copying, storing, formatting, displaying, printing, and deleting documents. It allows the user to convert documents to and from MS-DOS ASCII files and organize them into libraries. It supplies the same editing facilities whenever the user presses a key, whether in answering a question, giving a command, or editing a document. During the interviewing process the system automatically constructs a client data file from answers to questions; that file is searched for relevant information any time that questions are asked as documents are processed for that client. Model documents are ABF programs in disguise; ABF contains all the facilities for conditionals, looping, multiple-values, and subroutines provided by any modern programming language. In addition the ABF user can modify the program while it is running. The user can also switch contexts at any time from editing to processing a document or issuing a command. If the system finds an error, it puts the user in the editor with the cursor pointing to the place where the error occurred. Although the system was originally designed for use by lawyers, paralegal personnel, and law students, it embodies new techniques for automating any office that produces documents of a repetitive nature. ABF is a cross between an expert system shell and an interpreter, embedded in a sheltered microcomputer environment. AN University Microfilms Order Number ADG87-26879. AU HSIEH, CHENG-YUAN IN Illinois Institute of Technology Ph.D 1987, 146 pages. TI OFFICE PROCEDURE LANGUAGE: AN OBJECT-ORIENTED APPROACH. SO DAI V48(09), SecB. DE Computer Science. AB This thesis presents the concept and design of a general-purpose programming language: Office Procedure Language (OPL). OPL is based upon a formal model for the specification of knowledge-based information system, the OPM model. From the database viewpoint, the OPM model utilizes the database alerting techniques to serve the purpose of office activities management. OPL is developed to specify an OPM model with object-oriented approach. An OPL program can be translated to the corresponding OPM model. The OPM model consists of databases, messages, office activities and alerter rules. The general goals of the OPM model are (1) to describe the relationships between these office objects; (2) to provide a perspective view for the coordination and integrating of office activities; and (3) to facilitate the protocol analysis and the verification of office procedures. OPL is the linguistic interface between OPM model and system programmers. It is a high-level language developed to meet the specification requirements of the above goals. In OPL, the mechanism of database-altering is used to specify the knowledge of an Office Information System (OIS). OPL is developed for OIS design, where knowledge is expressed as database alerter rules. However, due to its generality, it is also applicable to the design of general information systems. AN University Microfilms Order Number ADG87-26518. AU MYAENG, SUNG HYON. IN Southern Methodist University Ph.D 1987, 175 pages. TI THE ROLES OF USER PROFILES IN INFORMATION RETRIEVAL. SO DAI V48(09), SecB. DE Computer Science. AB One difficult problem in information retrieval (IR) is the proper interpretation of user queries. It is extremely hard for users to express their information needs in a specific yet exhaustive way. From a different perspective, user variability in information seeking behavior is not well reflected in a query. As an effort to alleviate this problem, two theoretical models have been proposed to utilize user characteristics maintained in a form of a user profile. Although the idea of integrating user profiles into an IR system is intuitively appealing, and the models seem viable, no research to date has established a foundation on the roles of user profiles in such a system. Aiming at the investigation of the roles of user profiles, therefore, this study first identifies and extends various query/profile interaction models to provide a ground on which the investigation can be undertaken. From a continuum of models characterized based on interaction types, metrics, and parameters, nearly 400 models are chosen to investigate the "model space." Following the preparatory work, a series of experiments are conducted using an experimental IR system built for this study. In recognition that existing measures are not sufficient for the evaluation of so many models, new measures are developed based on the notion of user satisfaction/frustration. In addition, three different criteria are introduced to guide users in making judgements on the quality of retrieved items. A number of interesting results are produced through the analysis of the data obtained from the experiments. It is first shown that, regardless of a criterion or a metric used, there are always some query/profile interaction models that outperform the query alone model. In addition, preferable characteristics for different criteria are identified in terms of interaction types, parameters, and metrics. To ensure the significance of the results, three statistical tests are used for different purposes. AN This item is not available from University Microfilms International ADG05-61333. AU CHEN, JASON S. J. IN University of Southern California Ph.D 1987. TI DISTRIBUTED QUERY OPTIMIZATION IN FRAGMENTED DATABASE SYSTEMS. SO DAI V48(09), SecB. DE Engineering, Electronics and Electrical. AB Join is the most critical operation in distributed query optimization. In this thesis, the problem of optimizing multiple joins in fragmented database systems on both broadcast and nonbroadcast type computer networks is analyzed. Semantic information associated with fragments are used to eliminate unnecessary processing. Data redundancy is considered. Furthermore, we allow more than one physical copy of a fragment to be used in a strategy to achieve more parallelism. In our proposed approach, the problem of optimizing multiple joins is decomposed into two subproblems: the problem of finding a good join sequence and the problem of optimizing each two-way join in the sequence. A dynamic programming algorithm is developed for determining a join sequence. During intermediate steps of the join sequence, we have the join results remaining fragmented to achieve more parallelism and allow more local executions. All the partial results are assembled at the last two-way join. If the network has broadcast capability, graph models are introduced to represent two-way joins. The two-way join optimization problems are mapped into equivalent graph minimum-weight vertex cover problems. An algorithm based on network flow is developed for optimizing two-way joins with results fragmented. The two-way join optimization problem with results assembled is proved to be NP-hard. For nonbroadcast network environments, the problem of optimizing two-way joins either with results fragmented or with results assembled is also proved to be NP-hard. For those NP-hard optimization problems, properties are identified to reduce the solution search space. Efficient heuristic procedures based on the identified properties are developed for suboptimal solutions. Theoretical bounds are provided to ensure the heuristic solutions are within a certain range from the optimal solutions. Semijoins are also included in our approach. A new operation called domain-specific semijoin is introduced which can be performed in a fragment-to-fragment manner as opposed to a relation-to-relation or relation-to-fragment manner as in the application of regular semijoins. For a given query, there is always a strategy, using both domain-specific semijoins and semijoins, which is at least as good as the best strategy with only semijoins. (Copies available exclusively from Micrographics Department, Doheny Library, USC, Los Angeles, CA 90089-0182.). AN University Microfilms Order Number ADG87-27824. AU CHEN, ZI-TAN. IN University of California, Santa Barbara Ph.D 1987, 156 pages. TI QUADTREES AND QUADTREE SPATIAL SPECTRA IN LARGE SCALE GEOGRAPHIC INFORMATION SYSTEMS--THE HIERARCHICAL HANDLING OF SPATIAL DATA. SO DAI V48(09), SecB, pp2768. DE Geotechnology. AB The demand to manipulate large volumes of geographic data is growing. Besides conventional maps and statistics, huge volumes of geographic data are produced by remote sensing, conventional mapping, and auto-cartographic processes. Such data need to be manipulated efficiently in very large scale geographic information systems. However, current geographic information systems exhibit major shortcomings in the efficient handling of spatial data. The purpose of this dissertation is to explore the use of quadtrees and quadtree spatial spectra to improve spatial data handling efficiency. Artificial intelligence has a potential to eliminate some of the disadvantages of present spatial data handling methods. However, the gap between the theory of AI and its practical application in spatial data handling is still very wide. Many previous efforts in this area, including contextual and syntactic analysis in digital image processing, have shown interesting functions, but also have serious limitations for applications involving very large spatial data files. This study is an effort to shorten this gap by presenting several related studies at different levels. (1) At a high control level, a spatial heuristic search module is proposed. It achieves a significant gain in CPU time efficiency by using spatial knowledge at an early stage to find a short cut strategy. A heuristic search substitutes for the blind search of most current geographic information systems. (2) At the data structure level, quadtrees are used to represent very large volumes of geographic data, including both binary and gray tone images. Finally, (3) A global quadtree coordinate system is proposed for uniform data representation among different geographic information systems that have various sizes and locations. An approximate spatial distribution knowledge--quadtree spatial spectra (QTSS)--is proposed. It provides the necessary spatial knowledge for a spatial heuristic search module. Quadtree spatial spectra contain rich spatial distribution information over a wide spatial wavelength domain. The speed of generating QTSS is two orders of magnitude faster than the generating Fast Fourier Transform (FFT), and its storage form is very compact. Combining these studies, a practical geographic information system has been designed. Real geographic data sets are tested. Their results are analyzed, and compared with current systems. ------------------------------ END OF IRList Digest ********************