Jun Toyoura TRC Information Integration Laboratory

Profile

Affiliation:	Information Integration Laboratory Theory and Novel Functions Department
Date of birth:	May 29, 1962
Birthplace:	Tokyo
Background:
1986	Graduated from the Department of Physics, Faculty of Science, University of Tokyo
1989	Left in mid-course of the Doctorate Course, Postgraduate School of Physics, University of Tokyo
1989	Joined NEC
1991	Joined Mitsubishi Electric Corporation
1994	On loan to RWCP Tsukuba Research Center

My relation with "Tsukuba" may go back to when I was a high school student. When I entered high school, its name was changed to "Affiliated High School of University of Tsukuba" from "Affiliated High School of University of Education." My honest impression was "why it should be named after such a rural place" because the high school was located in Tokyo, not in Tsukuba.

Several years later, I joined the High Energy Institute in Tsukuba* to do experiments on physics while I was on a postgraduate course. At that time, highway bus services were not available. So, I traveled by train named Joban-line from Tokyo to Tsuchiura, the nearest station to Tsukuba, and took local bus to the city center. I made this little journey many times. Living in the city for half a week forced me to learn "how to live in Tsukuba" with reluctance.

And in 1994, I was again appointed to a post in Tsukuba. Frankly, I anticipated that the city might have been changed a great deal after ten years have passed. I found, however, that Tsukuba was still a strange combination of "urban" and "suburban" landscape. When you are stationed in such an environment, you feel like adopting a more healthy lifestyle. Once you allow yourself to have such a lifestyle, you can never return to the hassle and bustle of modern cities like Tokyo. So, I'm determined not to acquire a "Tsukubaniac" lifestyle and spending my days in a very unhealthy manner.

My job? Oh, yes. I'm now engaged in research on information bases at the Information Integration Laboratory of Tsukuba Research Center. Recent years have witnessed a remarkable development of information infrastructure. In particular, a word "Internet" became one of the most popular words in Japan last year and we can now access virtually any kind of information we want. It can be described as a new "Era of Exploration in the Information World." Our mission is to develop navigation technology on the sea of information.

Suppose I have a CD-ROM containing newspaper articles for one year (about 100,000 articles). This will be the base for our discussion of information navigation.

First, we search the database using the keyword "Tsukuba" written in kanji characters. We find nine articles all of which are about "University of Tsukuba" or "Mt. Tsukuba" (both are usually written in kanji). Then, we will make another search with the keyword "Tsukuba" in hiragana letters. We hit 40 articles. Most of them are about the city of "Tsukuba" (the city name is always written in hiragana). The original nine articles are not selected.

This means that, from a computer's point of view, "Tsukuba" in kanji and in hiragana are quite different things. So, the computer cannot find articles about "Tsukuba" in kanji with the keyword "Tsukuba" in hiragana and vice versa. To make the search possible, the computer must know that the two "Tsukubas" in different scripts are synonyms. A set of synonyms, antonyms, and related words is called a thesaurus. The thesaurus is an effective compass for information navigation. Since the sea of information is constantly changing, our thesaurus should be continually updated. For example, "Morihiro Hosokawa-Prime Minister" should be changed to "Tsutomu Hata-Prime Minister," "Tomiichi Murayama-Prime Minister" and finally "Ryutaro Hashimoto-Prime Minister." One of the primary objectives of our research is to establish the automatic creation of such a thesaurus on a computer.

Another method of navigation is possible by utilizing overall similarity of articles themselves rather than linguistic relations. For example, you would select a certain article (subject) and "look for any related articles in all data from the past year." To identify this similarity among articles, multivariate analysis and other methods would be needed to set up an appropriate space for expressing significance. This is the difficulty we are faced with.

We now plan to have a demonstration at the RWC Symposium to be held in 1997.

* When I joined the High Energy Institute, the name "Tsukuba" in hiragana was not yet in use. See if you can identify which "Tsukuba" is used for the following names: the kanji or hiragana version? (1) Tsukuba Mitsui Building (2) Tsukuba Bank (3) Tsukuba Ham (4) Minami Tsukuba Golf Club