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

Research Bulletin of the NACSIS, No.12 (2000.3)

Kinji ONO
Preface (Japanese)

1. Eisuke NAITO
Reading, Writing and Abacus in the Age of Digital Contents(English)

Mission is discussed together with emergence of digital contents, Japanese definition of information literacy for pursuing the universal elements of library and information services. Traditional skill such as "Reading, Writing and Calculus (Abacus)" is the basis of intellectual comprehension and assimilation in the age of digital contents. Social role or function of librarianship, in the light of the Global Information Society, is discussed to be still the same as the traditional function defined by S. R. Ranganathan.

Building Japanese Compound Words Analyzer Based on Grammatical Constraint (Japanese)

In the previous research of analyzing compound words, the technique using statistics of word co-occurrence is well known. However, since some grammatical and semantic restrictions must control the construction of compound words, only the statistical technique is not enough to analyze compound words. In this paper, we analyze technical terms that consist of two nouns whose head is verbal noun by means of the rich lexical knowledge that are a level ordering assumption, an argument structure and a lexical conceptual structure used on morphology or syntax. We will show the following three results: 1) there are only two kinds of relationship that is a modifier or a argument for a verbal noun in the relationship between two words in a compound word, 2) the analysis of compound words depends on argument structure of verbal noun, and 3) the structure of some compound words are changed according to the context. Finally we discuss constructing compound words analyzer from the results.

3. Kazuko KURIYAMA,Noriko KANDO,Toshihiko NOZUE,Koji EGUCHI
Pooling for a Large-Scale Test Collection : Analysis of the Search Results for the NTCIR-1 Workshop (Japanese)

The purposes of this study are (1) to verify the effectiveness of the pooling method for constructing a large-scale test collection, (2) to examine whether the number of documents in a pool affects the relative evaluation of the IR systems, and (3) to verify the reliability of a test collection by investigating the effects of variations in relevance assessments on the evaluation of search effectiveness, since individual relevance assessments are known to vary widely. For (1) and (2), we carried out an experiment using the qrel's (queries and their relevance assessments) and the search results submitted for the pre-test and test of the NTCIR-1 workshop for the pooling method. Our results verified the effectiveness of the pooling, the exhaustiveness of the relevance assessments, and the fairness of the evaluation using the test

4. Koji EGUCHI
Query Expansion using Context-sensitive Local Information of Clusters (Japanese)

Cluster-based browsing is one of the effective methods for accessing needed information out of a large amount of search results. We have investigated the framework which can adapt to dynamic information needs of users by searching against database using the reformulated queries, in which users' interests are captured at every interaction of the system with users while cluster-based browsing. This paper proposes a new query expansion method using context-sensitive local information extracted from the clusters judged as being relevant by users. We found that the proposed method prospectively has good effectiveness through basic experiments.

5. Shoichiro ASANO,Hisao UOSE, Shigeo URUSHIDANI
Trend in Next-Generation IP Router Technologies (Japanese)

Next generation IP routers have been researched and developed toward next-generation Internet era. To construct a scalable network for IP traffic growth, next generation IP routers are required to accommodate ultra-high-speed circuit interfaces, have very large switching capacity and high-scalability, have high-reliability as a core network system, satisfy diverse quality of service (QoS) demand, and so on. This paper surveys the recent IP router technologies to satisfy these requirements and considers the prospect of next-coming technologies.

6. Katsumi MARUYAMA
A Java library gCape hfor building distributed active object systems (Japanese)

Most distributed systems are based on either the C/S (Client/Server) model or the P-to-P (Peer to Peer) model. In C/S based distributed systems, a client invokes a server and waits for the server reply. Because of this sequential nature, C/S based distributed systems can be implemented by the RPC (Remote Procedure Call) scheme. Most tools for developing distributed Java objects are based on the RPC scheme. Whereas, in P-to-P based distributed systems such as process control systems, telecommunication systems and distributed agent systems, each distributed objects work concurrently, by exchanging asynchronous messages, without waiting for the receiver's action. To implement these P-to-P distributed systems, the RPC scheme is not powerful enough, and the active object model and asynchronous messages are suitable. This paper explains the pure Java library which supports active objects and very flexible asynchronous messages.

7. Frederic ANDRES, Nicolas DESSAIGNE, Vanessa MILONE, Kinji ONO
The MISE Metadata Engine for Multimedia Information Retrieval (English)

With more and more information on the web, users, and moreover novice users, are quickly lost when they need to find relevant information. To face this problem, search engines include new technologies from different fields (e.g. artificial intelligence, natural language processing, data mining, user profiling.) But while textual information is well covered, a huge lack still exists when dealing with visual information. The MediaSys Image Search Engine (MISE) is developed to tackle these issues with images. It combines an image search-by-content engine, namely \n Find\n IM \n age \n, with metadata capabilities via a meta-engine. In this paper, we present the metadata engine of MISE. Its main contribution is to use fuzzy logic to address problems of inaccuracy and uncertainty contained in natural language. Both discrete and continuous information are supported using the concept of fuzzy thesaurus.

8. Weiping ZHAO
Dynamic Synchronization of Media in Distributed Environment National Center for Science Information Systems (English)

Media synchronization is a basic problem that happens spontaneously in a packet based network which naturally introduce end-to-end delays varying both within and across data streams. The presence of delay variation (jitter) has an impact on audio-visual media synchronization and its presentation quality as perceived by a human user. It is a necessity to mitigate the impact of network jitter on multimedia applications in a distributed environment. In this paper, we propose a media synchronization method called dynamic temporal playback control (DTPC) algorithm that makes use of network delay information. In the DTPC algorithm, the optimal playback point is determined by using some certain knowledge on transmission delay of packets and based on this playback point the original stream can be faithfully recovered with an acceptable distortion while the packet loss rate is kept under a given bound. The performance of DTPC has been evaluated by simulations and the simulation results have shown that the DTPC algorithm reduces the end-to-end jitter, presents superior performance over other methods, especially in the case of bursty network delay. The algorithm has several parameters which can be used to tune its performance.

9. Hironobu GOTODA
A Multiview Stereo Algorithm for Reconstructing Polygonal Models (English)

TMultiview stereo reconstruction is an attempt to recover 3D structures from multiple collections of view images. It is an extension to the conventional stereo matching method that reconstructs the depth information from pairs of images. Recently, the so-called "large-scale stereo algorithms" have been developed, which enables us to recover geometrically or photometrically very complex shapes from large collections of images. However, most of the multiview stereo algorithms can produce low-level geometric primitives, such as points or voxels, which are not so useful in real applications. To obtain high-level geometric primitives, such as curves and surfaces, the output of those algorithms should be analyzed and processed again using other methods. This paper proposes a new algorithm that extracts polygonal models directly from the multiview images. More specifically, the algorithm finds such sets of polygons on which the multiview images can be projected coincidently. Since the output of the algorithm will consist of polygons, it can be easily converted to a surface representation. Furthermore, the algorithm can efficiently exploit the parallel processing capability of 3D rendering hardwares.

10. Hiroshi HOSOBE

Extending Constraint Hierarchies for Better Visualization (Japanese)

Constraints are useful for modeling and solving various problems such as reasoning, scheduling, design, and visualization. Particularly, nonmonotonic schemes of constraints have been drawing considerable attention since they allow us to specify over-constrained real-world problems. This paper proposes hybrid constraint hierarchies as a novel nonmonotonic scheme constructed by extending the traditional popular nonmonotonic scheme of constraint hierarchies. This new scheme combines different kinds of problems into a single hybrid constraint hierarchy, relating it with a multiobjective optimization problem. After the proposal, this paper provides how to effectively apply hybrid constraint hierarchies to visualization systems. They are powerful especially for the integration of graph layout with constraints. Also, this paper outlines a prototype system that uses a genetic algorithm for constraint satisfaction, and presents its preliminary experimental results.

11. Shunji ABE,Shoichiro ASANO
Accuracy Refinement for Diffusion-Approximated Tail Probabilities of GI/G/1 (Japanese)

In next generation Internet, not only computer data communications but communications such as voice, music, and moving video which severely require real time quality must be handled completely. On this account,it is very important to research and develop technology which can guarantee required quality of each communication. In order to develop such the technology, it is necessary to establish an estimation method of performances such as overflow probability from a queueing buffer, delay of queueing buffers in router systems and switching systems. This paper considers a tail probability P{Q m} ,which is one of important performances, for a queueing buffer system with inputs of burst packet traffic sources. We use a diffusion approximation method as analysis of P{Q m}. The tail probability is given by P{Q m}=e -c2m in the diffusion approximation. In this approximate solution, the error is tendency to increase as offered traffic volume decreases. We can not get enough accuracy to estimate it. To improve the accuracy, we consider a new decay function of e -c1m , and we propose the tail probability is approximated by P{Q m}1e -c1m { 2e -c2m , that is, the complementary distribution of the two-stage hyperexponential distribution. A determination method of parameters of 1, 1, and 2 is described. Also, we show efficiency of our proposed method by comparing approximation results with computer simulation results.

Implementation of an Internet--Based Dictionary Development Environment: "SAIKAM" (Japanese)

In this paper, an ambitous on-line Japanese--Thai dictionary development project called "Saikam" initiated by a group of Thai professionals and students in Japan is discussed. The project provides an on--line collaborative integrated environment to support development of Japaese--Thai dictionary on the Internet. Developers from all over the World can connect to the centralized dictionary database and update the content anytime at their convenience using standard web browsing tools. Some technical efforts have been made to enable trilingual editing environment on the existing WWW tools. Also, indices on large--scale Japanese text corpus were created to assist in selection of frequently--used words, and provide word usage navigating features to students of Japanese language.

13. Keizo OYAMA,Noriko KANDO,Shin'ichi SATOH,Hiroyuki KATO,Soichiro HIDAKA
Development of an Online Journal Editing and Publishing System (Japanese)

The editing and publishing system for online journals which is under development at NACSIS is to realize efficient distribution of research results, by means of supporting for academic institutions to make all the processes of producing academic journals electronic and online. The in-house system which is operated and used by each academic institution is designed to be flexible enough to adapt to various editing organization and document formats. The online system which is operated by NACSIS and is used for online publishing has a powerful searching capability and provides various reading support functions. This article describes the functions and usages of the system from viewpoints of authors, editors, subscribers, etc., and explains possible introduction methods to academic institutions.

14. Masaki NISHIZAWA,Yuan SUN,Masaharu YANO
An Investigation into the Classification of the Research Fields of Information Sciences in Japan (Japanese)

This paper reports a result about the classification of the research fields of information sciences in Japanese universities. We use the data of "1995 Academic Research Activities Survey" (Directory of Research Activity Database) for this classification. Authors refer to the validity of the classification of information sciences such as Computer Science, Intelligent Informatics and Study of Information Systems in the survey. This is an interim report for not only quantitative analysis but a study of relationship between information sciences and other fields.

15. Toshihiko NOZUE
Study on Information Use Instruction in Public Libraries: Toward Considerations of Information Literacy Education (Japanese)

To promote information literacy is getting more important in social education. In this article current status and problems of user education in public libraries are discussed.

16. Masaharu YANO,Akio KAMEOKA
Analysis of the Decisive Factors in Competitiveness of Industrial Technology (Japanese)

We gave a survey of competitiveness of each field and each technical item of industrial technology in order to compare the Japanese situation with that of U.S.A., Europe, and Asia. By the statistical analysis of answers concerning the technical level, marketing competitiveness, and their decisive sources, we found that the sources are classified into three factors. They are power of research and technology, that of product development, and that of management. We also found strength and weakness of these three factors of each field. After this we will improve the questionnaires and collect more data to analyze further.

17. Akitoshi EDAGAWA
The study on the relation between the result and the investment of the research activity in universities (Japanese)

It makes the 21st century imminent and for the science and technology promotion in Japan, our Government will establish the Ministry of Education and Science and The General Science and Technology Conference. For the purpose of the promotion of the research activity, it is necessary to investigate relation between the research investment and the research result, which could be the evidence in the analysis. So far, such a study wasn't much done, because the common index between the researcher, and the administrative person in charge didn't exist about the result, while for which it is possible to measure the investment comparatively. In my study, making use of the Science and Technology Agency investigation and the questionnaire survey to the university researchers which was done in 1998, I show relation between the research investment and the result, i.e. the relation which should be called a research activity production function. Incidentally, I used the number of the papers as the result, but I will referr to the problem of the quality, too, in the future.

18. Mitsuhiro FUKUDA
A Study on Laws Regulating Markets for Information Goods : Toward Construction of "Information Law and Economics" (Japanese)

The Civil Law regulates mainly markets for material goods and the Intellectual Property Law regulates only producer markets for information goods. In other words, consumer markets for information goods are not taken into sufficient consideration under the present legal system. This problem was latent when information was firmly combined with material media on account of undeveloped reproduction technologies. But now, information is untying from material media on account of developed reproduction technologies, so that we are becoming conscious of this problem. As a result, we shall need the unified law regulating markets for material goods and information goods. When we design laws regulating markets for infomation goods,we should consider the risk that strengthening of intellectual property rights causes monopoly and gambling markets.

19. Kouichi DOI
A Research of Japanese Web Pages based on Attitude Expressions at the ends Sentences (Japanese)

It is one of the important problem that the search engine shows the subjective web pages, such as the diary or the experiences when you would like to get the objective information. On the contrary, it is the problem that the search engine does not show the subjective web pages when you would like to get the information about the new products which can be get by the users. The existing search services mainly search using "key words". The auxiliary words are not usually used in these search services. In Japanese the auxiliary words show the writers' attitude. The auxiliary words show whether the sentence gives the objective information or the subjective information. We discriminate between these two types of web pages using attitude expression at the ends sentences. We investivate the actual state of web pages before we research the discrimination.In this paper, we research the use of the sign of end senteces. we extract the attitude expression at the ends sentences. we argue the relation between the attitude expression at the ends sentences and the genre of the web pages. We gather the basic statistics and analyze the web pages by the factor analysis. We measure the propotion of the subjective pages such as the diary.