Digital Content and Media Sciences Research Division

Digital Content and Media Sciences Research Division, Associate Professor
Degrees: Doctor of Engineering, University of Tokyo, 2002
Research Fields: Text and Language Media


Developing log infrastructure for better services

Who used what, when, and where?

National Institute of Informatics (NII) offers various content services. One of these is the Japanese article search service CiNii (Scholarly and Academic Information Navigator), which has the largest database of its kind in Japan and sees the greatest number of access attempts. A worldwide trend has emerged in recent years for "open science," which aims to make information on scientific research available to the public whenever possible. In response, NII is seeking to greatly expand the service content of CiNii. Meanwhile, we are supporting the growth of CiNii utilizing logs. In the case of CiNii, logs record details such as who looked at what, when, and where. "What does the use of logs enable us to do differently to before?"-this is the question that I am suggesting and implementing answers to.

Improvement and stability in service operation

There are two ways of using logs. One is to find out about increases and decreases in users from the logs, and use the results to create attractive content and devise ways to get more people to use the service; I call this "deep log data utilization for service improvement." Logs can also be used to find out the operation status of the system; for example, checking whether the system has become unstable. Focusing on this allows "log-based homeostasis" aimed at operating the existing content service appropriately.
It is particularly necessary to devise strategies regarding what kind of information should be collected in logs, as well as what other information when analyzed in combination with the log information will produce meaningful results in service improvement. Therefore, many discussions with companies and internal researchers are being held as we try to create an overall picture of the kind of service that CiNii will be developed into. Figure 1 shows access status at one-minute intervals based on current CiNii log information displayed on a world map. This was created to assist the discussions, but it alone makes the circumstances of CiNii very clear. I am looking forward to seeing what will be revealed when we start exploiting logs fully.

Implementing search technologies in services

NII is full of talented researchers in the field of informatics, and I have been carrying out research for creating search systems since before the arrival of Google. Even after starting a business in 2002, I created services based on search technologies. With this background, my role is to ensure that the research results at NII do not end up simply as research but are implemented in services such as CiNii. Also, I am currently concentrating on improving the accuracy of name identification in my specialism of search technologies. Name identification determines whether similar items are one and the same, which is essential when dealing with large amounts of information. I want us to be proactive in adding this kind of technology to the services provided by NII.