Abstracts of Magazine FUJITSU 1998-11 (VOL.49, NO.6)

Special Issue : Digital Library

• Digital Library System : iLisminds

Significant advances in information technology will be made in the next decade, along with the digital transformation of library systems. In January 1998 Fujitsu delivered its new digital library system （iLisminds） to Kyoto University. This paper describes the relationship between the support functions of iLisminds and library service functions, and explains other standard iLisminds functions.
Since a digital library requires functions to process bibliographical information, iLisminds is compatible with the new iLiswave library system. iLisminds also provides such new characteristic functions as retrieval support for full text, reading support with book metaphors, information processing to convert printed material in books into electronic code and hypertext data, practical application support for librarians, and user identification.

• Browsing Search Functions for Digital Library Systems

Due to the development of new network and computer technologies, there is a growing demand for digital library systems that can retrieve many kinds of distributed information resources. An effective retrieval system for all media except text has yet to be developed, however. Moreover, existing text retrieval systems cannot effectively retrieve from among a variety of data items on the Internet.
To resolve such problems, we have developed a new 3-D browsing space system. In this system, characteristic values are selected to represent data items in a retrieved set, along with the creation of information distribution and 3-D browsing space. This paper describes how our new system satisfies various user requirements for the retrieval of multimedia data.

• Document Recognition Method for Digital Libraries

This paper describes our newly developed document recognition method by which printed materials such as books, magazines, and newspapers are accurately converted into electronic code data to create the contents of a digital library. After the domains of sentences, tables and figures are extracted from document images using document layout analysis processing, then sentence and table text are recognized using character recognition processing. Document layout analysis processing allows multiple columns to be precisely extracted using the hypothesis test method, even in a document containing horizontal and vertical lines. Therefore, characters are recognized precisely at the recognition accuracy at which characters are accurately recognized by using only a detail classification method. We have also developed a binary conversion method by which uniform color text images are precisely converted into binary images. This method will be used in the future to convert color books into electronic code data. Fujitsu is now developing digital library systems using the methods above.

• Interactive Search System Based on Language Information

We recently developed three types of interactive search interfaces for a new searching system that can be used by anyone to easily and efficiently retrieve desired information. These interfaces support search operation based on such linguistic information as the relationship between words, specific contents and important information extracted from text to be searched. The “narrow support search” interface is used to display candidate key words and enable complex queries to narrow down retrieval results. The “template search” interface is used to extract specified contents and information in printed material for evaluating trends in products and industry, for example. The “heuristic retrieval” interface is used to reveal new factual information by displaying the multiple appearances of key words. This paper describes these three interfaces and the mechanisms by which they operate.

• Document Processing Utilities to Support Document Utilization Cycle Based on Natural Language Processing Technology

This paper introduces a new document list viewer and browser, which are convenient tools that apply natural language processing to support document utilization. The document utilization process involves a cycle in which required documents are first retrieved, then analyzed by content. The user can then formulate ideas and create another document. Both tools help the user perform essential steps in this cycle. The document list viewer provides a well-organized list of documents by clustering documents based on content similarities, and by attaching precise summaries generated automatically, the user can quickly find a required document even from a long list of retrieved documents. With the automatic text summarization function, the document browser allows users to skim through a long document on a computer display, then interactively creates customized summaries. This paper describes the features of these new tools and discusses the requirements for document selection and browsing.

• New Information Extraction System

A major objective of digital library development is to provide users with flexible opportunities to access and use information. The accumulating volume of information, however, poses an obstacle to the effective access of desired information. Conversely, text can be structured by extracting event structures to improve the quality of information retrieval and the arrangement of information. We are developing an extraction system that generates event structures based on top-down pattern analysis, then extracts named entities according to restrictions imposed by these structures. After pattern analysis, the system correlates extracted entities with actual world entities to enable the effective acquisition of required information. This paper describes this new information extraction system.

• Content-based Retrieval System for Video Data

Digital libraries of the future will be able to store video data as well as text data. For this reason, retrieval technology is essential for effectively handling video data. We have developed a new video data retrieval system that uses color distribution and the moving direction of detected object regions as the featured information. Motion vectors and DC coefficients in MPEG-2 data are used to generate featured information to be stored in a database. GUI is used to enter the features of a desired object so that a search can be made for similarities contained in the database. Then several items of highly similar data are displayed as the retrieval result. Our prototype system achieved retrieval precision of 0.83 using color information. Typical retrieval time varied from 9 to 95 seconds on a 166 MHz Pentium PC, according to the number of similar candidates. This paper describes our new content-based retrieval system for video data.

• Product Development for INTERSTAGE

This paper outlines the future direction of network computing technology and introduces a new Fujitsu software product called “INTERSTAGE” designed as a platform for network computing. INTERSTAGE is a distributed object environment developed for the Internet and intranets. Key components include ObjectDirector, which is based on OMG CORBA 2.0 specifications, and TransactionDirector, which is compatible with OMG transaction service. ObjectDirector, the first Object Request Broker （ORB） product released in Japan, can be linked between PCs and the Internet. This particular product is already being used in the Vehicle CALS project for the automotive industry, the SUBARU Telescope System for the National Astronomical Observatory of Japan, and GLOVIA （BPR）. We believe that such infrastructural support software products for realizing distributed environments are essential in today's market, and will serve a vital role in industry.

• Fujitsu's Response to Trends in SGML and XML Technology

This paper describes the basics of Standard Generalized Markup Language （SGML） and eXtensible Markup Language （XML）, which represent the next generation of technology for describing different types of documents. The development background, growth, and targeted applications of SGML are first discussed, followed by a comparison of the characteristics of SGML and Hyper Text Markup Language （HTML） with those of XML, and the advantages of using XML. This paper also describes language specifications applicable to XML, specifications related to XML, an application in which XML and an XML browser are used, and the potential use of XML technology as a core technology of future network environments and various applications.

• New High-performance, Low-power DSP Core : Hi-Perion Series

We have developed a new high-performance, low-power, 16-bit fixed point DSP core （Hi-Perion） for such digital signal processing as speech CODEC used in the digital base band of cellular phones. The maximum operating speed of Hi-Perion is 100 MHz. This device can also operate using a power supply of only 1.8 V, and consumes 70 percent less power than our existing products. Consequently, Hi-Perion enables the complex high-speed speech CODEC processing required in PDC and CDMA （which are essential technologies for the next generation of mobile communications） at significantly reduced power consumption.
This paper describes the architecture of the 16-bit fixed point DSP core and the environment required to develop systems that use a DSP core.