Site: Nara Institute of Science and Technology (NAIST)
Ikoma, Nara
Japan
http://nara.aist-nara.ac.jp/

Date Visited: 26 March 1998

WTEC Attendess: L. Goldberg (report author), T. Ager, M. Shamos, R.D. Shelton

Hosts:

• Kaoru Tsuda, Deputy Head of Cooperative Research Division
• Hikaru Tsuji, Cooperative Research Division
• Muneo Matsumura, Deputy Head of Information and Digital Library Services Division
• Shunsuke Uemura, Professor, Graduate School of Information Science
• Kosuke Sato, Associate Professor, Graduate School of Information Science
• Masatoshi Yoshikawa, Associate Professor, Graduate School of Information Science
• Shinji Kimura, Associate Professor, Graduate School of Information Science

BACKGROUND

The Nara Institute of Science and Technology (NAIST) was established in 1991 as one of two national universities in Japan devoted exclusively to research and graduate education. The uniqueness of these institutes, in addition to their exclusive focus on graduate education, lies in at least two areas. First, their policies and broad research directions are overseen by external boards of distinguished scientists from academe and industry. Thus, while their faculty enjoys somewhat less autonomy in selecting and pursuing their research topics, they are subject to fewer of the regulations imposed on other national universities by the Ministry of Education, Science, Sports, and Culture. A second distinction lies in the limited number of disciplines encompassed. NAIST currently has only three faculties: the Graduate School of Information Science (established in 1991), the Graduate School of Biological Sciences (1992), and the Graduate School of Materials Sciences (1996). A fourth faculty under discussion is the Graduate School for Library Science. If realized, this will be one of the first faculties to educate students as digital librarians.

The NAIST library opened two years ago as a true digital library intended for educational and research usage, the first such national library in Japan. The library provides five services:

1. real-time service for users, in which its main goal is to digitize all newly received materials within three days
2. network access from a university room or anywhere
3. media center, including CD-ROM, video, etc.
4. full text retrieval, using OCR on documents-270,000 pages at the time of this visit
5. alert function, in which users can register key words for alerts on new document arrivals

The Web homepage is provided in both Japanese and English. Statistics on document retrieval show about 3,800 pages accessed per day. Copyright licenses have been negotiated at reduced subscription costs with the publisher, Elsevier, so that all students, staff and faculty are covered, although this is not a significant issue given the university's small size of only 1,000 students. Video search is provided for approximately 800 titles, running about 100 hours of content in MPEG2 (180 GB). The entire digital library content requires approximately 6 TB storage, on magnetic tape and magneto-optic disk media. The library does not plan any new cataloging approaches different from MARC format for its digital content. The WTEC panel was shown NAIST's extensive library technical facility where new journals were photocopied, scanned and converted by OCR.

LAB DEMONSTRATIONS

Speech Processing Technology3/4Speech and Acoustics Laboratory

The major research projects in speech processing include speech recognition synthesis, enhancement, coding, and multi-modal dialog interface. NAIST researchers are conducting interesting new work on synthesizing lip movement based on mapping from input speech using the method of the Hidden Markov Model (HMM). In the acoustic signal processing they are working on active noise control and sound field reproduction. They demonstrated experiments on three-dimensional sound generation and additive cancellation from secondary sources measured by two microphones. The experimental facility includes a sound proof chamber with an array of 112 microphones independently input into a multichannel analyzer.

Electronic Design Automation

This well-equipped laboratory is used for LSI layout and related research and education projects. It includes equipment for printed circuit board design and production, although chip fabrication is conducted at a commercial foundry. One of the projects discussed was a method of binary digital diagrams (BDD), which provides a concise expression of logic functions for small layout areas, while requiring storage of only 2n entries for n variables on many practically used logic functions.

ParaDocs Project3/4A Database System for Documents with Object Links

The goal of this effort is the design and implementation of a database system, which can manage structured SGML documents and other data in an integrated manner. The approach involves the application of database facilities in document processing and the management of document components using database types. The parallel text model, or ParaText, has an appearance layer (= character string) and a reference layer (= database objects or values represented by substrings in the appearance layer). The retrieved information can be linked to external databases.

REFERENCES

Inoue, M., Satoshi Nakamura, T. Yamada, and K. Shikano. 1997. Microphone array design measures for hands-free speech recognition. Eurospeech.

Kimura, S., M. Yukishita, Y. Itou, A. Nagoya, M. Hirao, and K. Watanabe. 1997. A hardware/software co-design method for a general purpose reconfigurable co-processor. In Proc. 5th Intern. Workshop on Hardware/Software Codesign.

NAIST. 1997. Guide to graduate programs and facilities.

NAIST. 1996/1997. Digital library of NAIST.

NAIST. n.d.. A database system for documents with object link. ParaDocs Project, Grad. School of Information Science.

Nakamura, S. and K. Shikano. 1997. Room acoustics and reverberation: impact on hands-free recognition. EuroSpeech.

Nakamura, K., S. Kimura, K. Takagi, and K. Watanabe. 1998. Timing verification of sequential logic circuits based on controlled multi-clock path analysis. IEICE Trans. Fundamentals. 12 Dec.

Shikano, K., S. Nakamura, S. Ise, and J. Lu. Speech and acoustics laboratory. NAIST

Yamada, T., S. Nakamura, and K. Shikano. 1997. Speech recognition of a moving talker based on 3-d viterbi search using a microphone array. CASA Workshop, IJCAI.

Yamamoto, E., S. Nakamura, and K. Shikano. 1997. Speech to lip movement synthesis by HMM. In Proc. AVSP'97 Workshop. Rhodes, Greece.