The Information Technology Research Institute (ITRI) of the National Institute of Advanced Industrial Science and Technology (AIST), an independent administrative institution, will start from April 18, a service trial of the Researchers Network Search Engine (POLYPHONET) for robotics category, developed by the AIST in collaboration with a non-profit organization, R&D type NPO Promotion Organization (ISERD). The trial will be operated at the Robot Laboratory (Robot Lab) of the Business Innovation Center, Osaka, established by the Economy Bureau of Osaka City and managed by the Urban Industry Promotion Center, Osaka.

In these years, needs are steadily growing for making collaboration among researchers in some interdisciplinary fields such as robotics and nanotechnology, as well as those working at the academic and research organizations and the government and industrial sectors. Up to now, however, the collaboration has been made mostly on the basis of personal acquaintance or encounter by chance, far away from efficient match making. For finding effective collaboration, it seems that the use of search service based on information technology (IT) provides a helpful support. There have been a number of researcher search systems using publication database and membership registrations at various academic institutes. However, since the publication data become available only after research works have been finished, they may fail to be latest ones. While for the effective support of collaboration it is essential to search researchers in consideration of their detailed specialty and position within a particular specialty area, the currently available researcher search systems lack the function to meet these requirements.

The POLYPHONET, developed at the ITRI-AIST under the leadership of Dr. Yutaka Matsuo, a researcher of ITRI, makes it possible to display a researcher network and to search individual researchers, through the collection of research works and publication data posted on the entire webs and the extraction of collaborative relationship among researchers. In the POLYPHONET, the linkage between researchers is assessed through the analysis of frequency and pattern of appearance of particular researcher's name based on the research works presented on the web by use of a new technology designated "Web Mining". The system extracts various aspects of researchers' relationship in connection to research activities, such as affiliation with same research group or section, participation in a joint project, frequent co-attendance in scientific conference, and so on. Moreover, it is possible to search researcher's name based on the interrelationship between researchers. In this way, it becomes possible to search for researchers in different disciplinary areas and those required by the industrial and administrative sectors, to promote the support for effective collaboration.

The ITRI-AIST and the ISERD have embarked on a service trial for members of the Robot Lab through the experimental use of POLYPHONET (Robotics Version). The Robot Lab was founded in November 2004 by the Osaka Municipal Government for the purpose of creating robot technology (RT) industries. In parallel with this trial, it is attempted to verify whether the use of POLYPHONET contributes to the making of effective collaborating system. This is the first of the full-scale experimental operation in Japan of the system for finding researchers' relationship out of research activity data disclosed on the web.

The Osaka Municipal Government is trying the integration of next generation robot industries around the Robot Lab as one of the priority industrial areas. The Robot Lab is going to dig out research and technology seeds on robotics from among robot researchers not only in the Kansai District but also all over Japan, and to extend them to robot business.

The ITRI-AIST is going to apply the POLYPHONET to the area of robotics as the first target, for the purpose of system development and service trial, covering published research data for 3,600 Japanese researchers and reflecting the trial results to the system upgrading. It is expected that the application area will be extended to other interdisciplinary fields such as bioinformatics, nanotechnology and others, in collaboration with ISERD which is motivated by the construction of researchers network.

Recently, it has become an accepted practice to disclose researcher's data on his/her research works on the web. For instance, researcher's personal home page, publication list, programs of scientific conferences and meetings, grant-aid information and other data on research activity are readily accessible in the web. On the other hand, R&D works in interdisciplinary studies such as robotics, nanotechnology, bioinformatics, disaster science, and so on, where collaboration among researchers of different specialties is essential, are getting increased weight in these days. Furthermore, the industrial-academic-governmental collaboration scheme involving enterprises, regional communities and administration has been bearing greater significance for bringing research outcomes to the social contribution.

Under such a circumstance, it will become necessary to organize a collaborative team involving researchers of various specialty areas, and to offer cooperative aids to researchers from the side of enterprises, regional communities and administration. In case of robotics development, for instance, intimate collaboration is needed among specialists in device technology, software, speech processing, communications, and so on. For disaster science, cooperative works among regional communities, administration and researchers are essential. Up to now, the collaboration among researchers of different areas and tripartite cooperation have been promoted on the basis of personal acquaintance or intended arrangement, and none of systematic and efficient collaborative schemes have been available.

Now that R&D works in science and technology are finely fragmented and diversified, and the technological advancement is so rapid as hardly to follow up, it has become indispensable for creating effective collaborative organization to seek support from the information technology (IT) in the form such as search tools and information services. For providing effective IT support, it is essential to acquire not only data on researcher and his/her research works, but also those on the nature of research group, and its position in the particular research area, the name and career of the leading researcher, researchers and groups in competitive position, and other personal relation in research works. If resorting to personal acquaintance only, it may be overlooked who is the leading researcher, and whether or not any competitive research group exists.

While search systems have been available to look up researchers by using the publication database and registration info in scientific meetings, the publication of R&D work usually lags behind the actual activity by a few years, and new technological areas may not be covered by existing categories. As the conventional search systems for researchers are not able to cope with new requirements, IT supports are needed which can follow dynamic changes in researchers and industries, search researchers in specified areas of detailed specialty, and take into consideration of position and role of the researcher concerned in particular specialty area.

The ITRI-AIST has been engaged in the study on IT support in the ubiquitous computing environment based on user's position. As a part of this study, a project for providing info support service to participants of a scientific meeting at the conference hall, "event space info support project" has been started since 2003. In the 2003 and 2004 Meetings of the Japanese Society of Artificial Intelligence (JSAI), sensors were installed at the hall, and info service was offered to locate a specified participant. In parallel with this trial, another service of presenting researchers network was provided to extract a network of researchers based on their activities disclosed on the web. The service allowed a participant looking up networks among participants at kiosk terminals (or info terminals on street) or in web pages. This service is going to be offered in the 2005 Meeting of the JSAI. The POLYPHONET is a revised version of the system for searching researchers based on the technology of researcher network extraction, having been developed since July 2004. "Polyphonet" is a coined word consisting of "polyphony" or music with two or more independent melodic parts sounded together, and "network". It means linking researchers with human resources in enterprises and local communities through a network, so as to make up a better music while taking advantage of respective merits. In the course of studying POLYPHONET development, technologies have been developed for automatically extracting research keywords out of researchers' data, for automatically identifying his/her research category, and for extracting a network involving a lot of researchers.

The ITRI-AIST has been carrying out a joint research with the ISERD since December 2004 on the operation and upgrading of researcher network search engine, for the purpose of promoting its use for the tripartite collaboration and research exchange for interdisciplinary areas. On the other hand, the Robot Lab opened in November 2004 has been adopting the advancement of exchange between researchers and the industrial sector as one of major targets. Through the collaboration among the three organizations, it has been attempted to experimentally introduce the researcher network search engine, POLYPHONET to the Robot Lab for providing a service trial for promoting the collaboration among researchers, enterprises and local communities.

The POLUPHONET collects researchers' data on their research activities as disclosed in every web, to create and store a database for researchers' activities. The system will make it possible to display a researchers network by taking researchers' interrelationship out of the database (Fig. 1), and to find out relevant research partners in different areas of specialty. The POLYPHONET is characterized by (1) looking up researchers in need through the "researcher search" function by entering name, affiliation, research keywords and categories as the search conditions (Fig. 2); (2) Automatically extracting keywords and categories out of the web on the basis of analysis on what terms and words are associated with researcher's name on the web; (3) Finding out specified researcher's partners, co-authors and colleagues in the same section; and (4) Taking a bird's-eye view of researchers' interrelationship and group make-up in the overall field concerned (Fig. 3).

Fig. 1   A display of researcher's network retrieved by the researcher network search engine POLYPHONET

Fig. 2   A display of researcher's detailed data retrieved by the researcher search function

Fig. 3   A list of researchers closely linked to a retrieved researcher

With the "linkage search" function, it is possible to find out paths from a particular researcher to other related ones. For instance, it is possible to identify the possible paths from researcher A to researcher B (Fig. 4). If multiple goals and starting points are set, the system can be used for various aspects of search, such as "who is closest to you among researchers in the Osaka University" or "who is the most suitable partner for you among researchers listed in the automatic summary". In the event of none of wanted researchers included in the database, additional registration of new researchers in the database is also provided.

Fig. 4   A display of linkage pathways from a particular researcher to his/her possible partners as revealed by the "linkage search" function.

The POLYPHONET is based on the technology, named "web mining", to find out useful data not explicitly written in the database through the pattern extraction and analysis out of web-related information such as web page text, link structure and access pattern. First of all, disclosed information on the web for activities of a particular researcher is extracted by using researchers' list in the category concerned, and the strength and description of researchers' interrelationship are identified through the search engine. For instance, if entering "Yutaka Matsuo AND Ichiro Tanaka" and "Yutaka Matsuo AND Jiro Sato" yields 150 hits and 15 hits, respectively, Matsuo- Tanaka linkage may be regarded stronger than Matsuo-Sato linkage. In this way, the interrelationship between two persons can be assessed through the statistical analysis for number of hits on entering names of two persons in the research engine. Furthermore, the linkage between the two persons can be identified through the language processing for the contents of the hit page, discriminating four categories: sharing same affiliation, co-authorship, participating in same project, and making presentations at same scientific meeting. The relationship can be discriminated in an automatic manner by extracting necessary attributes out of hit page, and by applying a discrimination rule acquired by "machine learning", that is, exemplary pages and correct answers are presented previously, and the classification is automatically acquired.

The keywords and research areas of particular researchers are identified on the basis of the information on what technical terms are associated with the researcher's name on the web. For instance, it is highly probable that a researcher associated with a term "image processing" is engaged in the study on image processing. In this way, the research area of specialty and research keywords can be extracted through the statistical analysis of data for frequent association on the web of a researcher's name with a technical term.

As the POLYPHONET makes it easily possible for an outside person to identify key persons or to list up researchers being engaged in research on a particular theme, it may be regarded as a useful tool for promoting the study in the interdisciplinary areas, and advancing effective collaboration among industrial, academic and government sectors.

The ITRII-AIST and the ISERD introduce experimentally a robotics version of the POLYPHONET at the Robot Lab set up by Osaka City for the purpose of creating robot technology (RT) industry in November 2004 and start a service trial for members of the Robot Lab. It will be attempted to verify whether the use of POLYPHONET provides an opportunity for building effective collaboration scheme and whether the researcher's privacy is affected in view of personal data protection, through a hearing survey. This is the first full-scale service trial in Japan for a system to extract researchers' interrelationship based on the information on researcher's activity disclosed on the web.