[To Research Bulletin of the NACSIS page]  [To NII publication page]  [To NIIJournal page](Japanese)

Research Bulletin of the NACSIS, No.9 (1997.3)

[last updated: 1997.12.12]

1. Takamitsu SAWA
Political and Economic System in the Twenty-first Century (Japanese)

2. Toshiharu AOKI
The Future of Multimedia Network (Japanese)

3. Hisao YAMADA
Globalization of Information Society and Japanese Language (Japanese)

ABSTRACT Because of the arrival of global information society, the Japanese language and its writing system now face new challenges. In spite of the fact that the Japanese language belongs to the structurally simplest category in the world, it has the most complex writing system, and, therefore, it is not readily embraced in the world theater, while it is also cumbersome, inefficient and stressful to render into machine readable form for processing. In this note, we shall examine some possible means to make the writing system more globally acceptable as a long-term prospect, and also some possibilities to streamline its input process as a short-term task.

4. Nobuya AIHARA, Eisuke NAITO
Adaptation of Universal Standardized Subject Heading System in the CJK (Chinese / Japanese / Korean) Environment (Japanese)

ABSTRACT The paper is a Japanese version of the presentation made at the One Day Workshop held on Thursday, August 29, 1996 in Beijing, China by the IFLA Standing Committee: Library Services to Multicultural Populations and sponsored by IFLA Standing Committees: Cataloguing; Classification / Indexing during the 52nd IFLA General Conference. This Japanese version is enhanced by adding notes and references. The paper explores possibilities of adaptation of a Universal Standardized Subject Heading System to multi-script language environment such as Chinese, Korean and Japanese language, of which data string is a set of description, reading and Romanized character string in many cases. Subject heading systems in Japan are reviewed with some quantitative aspects to set a background to the universal standardization of subject heading system. Essential difference between MARC and subject heading data is that the former is the container and the latter is the content. Factors to be solved for achieving the Universal Subject Heading Systems are discussed in terms of human resource factors, technology factors, language factors, and administrative factors.

5. Keizo OYAMA, Atsuhiro TAKASU, Hiroshi TSURUOKA
Configuration of an IR System linked to a Union Catalog On-line Database (Japanese)

ABSTRACT This paper describes a method to link NACSIS-IR, an information retrieval system, with NACSIS-CAT, a catalog and location information system. The authors studied the configuration method of a system which automatically builds index of bibliographic and holding records stored in databases updated on-line, and which provides information retrieval services to the end users. Several methods were evaluated from viewpoints of load to the database server, response time of retrieval, file capacity and so on in order of the basic design of an actual system. As the result, the optimal configurations corresponding to structure and characteristics of databases are obtained.

Information Seeking in Terms of the Visual Data in an Office (Japanese)

ABSTRACT There are a few research work on information seeking and/or information organization in offices. This article discusses information seeking behavior and personal file organizations in offices. Spatial information, colors, forms of materials and usage time provide useful index to recall their contents and find a certain intended materials.

7. Yuan SUN
A study on IRT models for multidimensional data (Japanese)

ABSTRACT >Determining a correct response to many test items frequently requires more than one ability. Multidimensional IRT models which define m-dimensional latent space and locate persons as ability vectors are very useful to deal with multidimensional response data, especially in the case of cognitive diagostic testing. However, sometimes it is hard or even impossible to define and explain some dimensions as characteristics of items because dimensionality is an interaction of persons with items. GIRT model is a modified unidimensional IRT model which takes out one target ability dimension and consider the rest nuisance ability dimensions as person's ability distribution. In this paper a compensatory MIRT model and GIRT model are discussed, the mapping of the statistics in the two models is derived from an analytical aspect.

A study on lexcal items in Yakut language (2): Terms of direction in Saxa epic poem olonxo (Japanese)

ABSTRACT This paper is concerned with lexical items in Yakut language, which belongs to the Turkic language family. The main part of the Yakut language community is consited with Yakut or Saxa people. Their language traces well the process of their immigration to Siberia. In this paper, the author presents the data of the set of lexical items "directions" from the olonxo "Kiis Dabilija", one of the most famous epics of the Saxa people and suggests that such approach to the study of lexicon is effective to uncover the complicated situation of languages of Siberia. This paper is the second serial of the study on the lexicon of Yakut language, which we published in the preceeding number of the bulletin.

9. Hironobu GOTODA, Kinji ONO
Compression of Synthetic Movies Using Simplified Geometric Models (English)

ABSTRACT The MPEG standard is not necessarily suitable for compressing computer-generated synthetic movies. When transferring such a movie over a network, decomposing it into an outline and details, and compressing them separately, can lead to an improved compression ratio than compressing the whole movie using MPEG. Here, the outline is the portion of the movie that can be reproduced using simplified models, while the details are the sequence of difference images between the outline and the complete movie. We will elaborate this idea further by introducing mechanisms to exploit the interframe coherence. Our method adopts simplified geometric models both to reproduce the outline and to approximate the details. More specifically, an approximation of the optical flow is computed using the geometric models, which is then applied for estimating the coherence in the difference images. This leads to further reduction of the amount of details, and therefore less traffic over the network.

10. Frederic ANDRES, Christopher CLAUDE, Kinji ONO
AODBMS: Phasme and its Image Retrieval Plug-ins Implementing shape-based retrieval processing (English)

ABSTRACT We describe Phasme, an Application-Oriented database system, designed both to provide high performance and to meet the needs and the requirements of today's multimedia system as an opened and customisable DBMS. Phasme is an opened database system using the Extended Binary Graph(EBG) data structure and employing both inter operation and intra-operation parallelism. Phasme is fully customisable from the data type to the execution model. It provides a dynamic plug-ins mechanism and a set of many-sorted algebra to efficiently support different types of data structures. We discuss how these features impacted the design and the implementation of shape-based image retrieval applications. We show how Phasme features provide both challenges and opportunities in building a high-performance image retrieval system on top of it. In particular, we show how a novative shape-based image retrieval method can be mapped into Phasme's Extended Binary Graph. Capitalizing from previous works as archetype transformation, we propose a solution to support similarity image retrieval processing.

11. Kazuyuki IKEDA, Atsuhiro TAKASU, Jun ADACHI
A Method of Document Retrieval using Dependency Relationships between Words (Japanese)

ABSTRACT Although keyword-based retrieval occupies the major position in document retrieval methods, it has obviously limitation in retrieval effectiveness because it is difficult to select documents that coincide with user's requests using only query words and its statistical characteristics. Thus we propose a method to enhance retrieval effectiveness using language information such as dependency relationships between words in retrieving documents' title and keywords. In our method , we make structured index represented with binary tree through modification analysis by using case rules and compound nouns analysis which employes word bigram. Queries are issued by psuedo natural languages such as expressions found in documents' title, so users can reflect their requests on query without being concerned about query languages. In this paper, we describe these methods and inspect an effectiveness of our methods through retrieval experiments on our prototype system.

12. Manabu OHTA, Atsuhiro TAKASU, Jun ADACHI
Graduate School of Engineering, The University of Tokyo (Japanese)

ABSTRACT This paper presents three probabilistic text retrieval methods designed to carry out a full-text search of Japanese documents containing OCR errors. By searching for any query term on the premise that there are errors in the recognized text, the methods presented can tolerate such errors, and therefore manual post-editing is not required after OCR recognition. In the applied approach, confusion matrices are used to store characters which are likely to be interchanged when a particular character is missrecognized, and the respective probability of each occurrence. Moreover, a 2-gram matrix is used to store probabilities of character connection based on 2-gram statistics, i.e., which letter is likely to come after another. Multiple search terms are generated for an input query term by making reference to confusion matrices, after which a full-text search is run for each search term. The validity of retrieved terms is determined based on error-occurrence and character-connection probabilities. Those with a validity value greater than a given threshold are judged to satisfy the input query. In addition, the performance of these methods is experimentally evaluated by determining retrieval effectiveness, i.e., by calculating recall and precision rates. Results indicate marked improvement in comparison with exact matching.

13. Weiping ZHAO, Suwan RUNGGERATIGUL, Shoichiro ASANO
A Study on ARQ Mechanism in ATM Networks (English)

ABSTRACT ATM networks are designed to support a wide variety of services with different requirements of QoS. Some applications requiring very high QoS, and the progresses in IC technology and processing capacity make it possible to introduce ARQ mechanism in high-speed ATM networks economically and efficiently. In this paper, an ARQ scheme based on selective repeat strategy is proposed. Its performance in terms of throughput, traffic increment and required buffer size is evaluated by both theoretical analysis and simulation. The analytical and simulation results show that the proposed scheme has a very good error detection capability while maintaining high throughput with an acceptable requirement of buffer size.

14. Suwan RUNGGERATIGUL, Weiping ZHAO, Shoichiro ASANO
Analysis of Cell Loss Performance in ATM Networks (English)

ABSTRACT Asynchronous Transfer Mode (ATM) is known as a target solution for implementing broadband high-speed communication networks. To study the performance of an ATM network, it is necessary to consider an analysis model that can take general traffic model into account since the network can provide service to various types of sources.In this paper, cell loss performance in ATM networks for general traffic model is analyzed by studying the upper bounds of cell loss probability in GI/D/1 queue.It can be shown that generating function of traffic arrival probability is an important factor for the bound analysis but derivation of this kind of function is usually complex, especially for bursty traffic models. We then propose an approximation derivation method for the arrival probability generating function of bursty traffic source models. The idea of the proposed method is based on the fact that bursty traffic model often consists of multiple states each of which is a simple stochastic process including Poisson process, deterministic process. By some numerical results, it is shown that the proposed method gives very accurate results comparing to those obtained from exact calculation. Moreover, the computation amount of the method is very small. These features lead to the conclusion that the proposed derivation method is very effective to the control and design problem of ATM networks.

15. Yusheng JI
The Characteristics of ATM Traffic Constrained by Different Shaping Mechanisms (Japanese)

ABSTRACT GCRA has been made as the standard Usage Parameter Control Algorithm for ATM networks. It is equivalent with the so called Leaky Bucket algorithm. Besides, other algorithms, such as Jumping Window or Moving Window, have also been proposed. On the other hand, algorithms used in traffic shaping are usually as same as those used in UPC. Nevertheless, performance of these algorithms used in traffic shaping is not quite clear as them used in UPC. In this paper, we focus on cell delay and loss both at the shaper and the network switching nodes, and give a comparison of the characteristics of traffic shaped by Leaky Bucket and Moving Window. It is shown that the Leaky Bucket algorithm has a more disirable feature than the other one.

16. Akiko AIZAWA
Constructing Open Network Platform for Video Services (Japanese)

ABSTRACT An overview of current technologies for video services over wide-area communication networks is given. The importance of application level video gateway for heterogeneous internetworking is also discussed.

17. Jihad BOULOS, Yann VIEMONT, Kinji ONO
Neural Networks Approach for Query Cost Evaluation of DBMS (English)

ABSTRACT This paper presents a new approach for cost evaluation of query execution plans that may help or replaces the known analytical approach. Our proposed approach is based on neural networks and the connectionist concept. A neural network is trained to learn the execution cost of the implementation algorithm for a logical algebra operation (e.g., a merge-join implementation of the join operation) under some database conditions and in a certain environment; after that, this network is used to estimate the cost of this operation with other entries. The approach is based on a curve fitting like since neural networks have been proven to be "universal approximators." The main advantage of this approach is its applicability to user defined methods where the user does not need to estimate the cost of its method since the system may apply this method several times, collects measurements, and capture its behavior with its curve fitting capacity. Another advantage of this approach is its implicit covering of environment depending parameters.

18. Hiroaki MORINO, Akiko AIZAWA
QOS guarantee for real-time applications in the Internet (Japanese)

ABSTRACT Application in the Internet is currently being expanded form simple data transmission to video and audio communications. As the existing protocols such as TCP or UDP are not intended to guarantee QOS(Quality Of Service) for real-time applications like video or audio transmission, new protocols are required. In this article, we present some protocols which can guarantee the QOS for the real-time applications in the Internet. Main functions of these protocols are dynamic rate control corresponding to network load and media scaling. We discuss the rate control algorithm for each protocol and its effectiveness in improving picture quality. Finally, we show a brief result about performance comparison and make a report on the research which will guarantee the QOS in the area of future network infrastructure.

19. Shigeaki YAMAZAKI
Activities of research funding and the role of research databases for supporting the evaluation of scientific performance in the Wellcome Trust (Japanese)

ABSTRACT The Wellcome Trust was founded under the will of Sir Henry Wellcome, who died in 1936. The objectives of the Trust are to support research in the biomedical sciences and in the history of medicine. Now, the Wellcome Trust that is the most richest medical charity in the world, spent £242 million for the grant expenditure of 1994. PRISM, the Unit for Policy Research in Science and Medicine, has been developing ways of helping the Trust choose between areas of research and between different applicants. It has also been developing experimental approaches to priority setting and creating an exciting new tool for evaluation of outcomes, the Research Outputs Database. This unique database is develpped with the bibliographics detailes of UK biomedical research linked to their funding sources. Using the Research Outputs Database, UK funding organizations have been able to track the output of the research they have supported. The Wellcome Trust have also developed other useful databases. WISDOM consists of three databases on funding sources for biomedical research, science policy and the catalogue of the library. Evaluation of scientific performance would become more important in oredr to establish the effective research funding. The Wellcome Trust appears to present us a model for the system.

20. Shigeaki YAMAZAKI
Bibliometric analysis of the papers on clinical trials and randomized controlled trials in the Medline database, with reference to the Cochrane Collaboration (Japanese)

ABSTRACT The purpose of this survey is to study the growth of number of papers and the distribution of language and country of publication in the field of clinical trials (CTs) and randomized controlled trials (RCTs). Based on the Medline database between 1966 and 1995, the number of papers on CTs and RCTs were obtained using the publication type that was introduced into the Medline in 1991. Though Japan has a large share of world publications in science, its share in the field of CTs and RCTs is smaller than that of the other developed countries. CTs and RCTs are now increasingly being used to evaluate the effectiveness of health care intervention. The Cochrane Collaboration is working with the National Library of Medicine to identify the published papers on CTs and RCTs. As the consequence, practicing physicians and allied health personnel are able to find the best evidence from clinical trials in the Medline database. In order to improve the use of Japanese databases in medicine, JMEDICINE and Igaku Chuo Zasshi (Japana Centra Revuo Medicina) should be suggested to index adequate information on CTs and RCTs.