Research at Kaneko Lab.

Today’s information society is supported by the large-scale information space represented by the World Wide Web. However, direct human involvement, such as keyword searches, is essential, and access to truly valuable information has not always been easy. Kaneko Lab. aims to create new information value by “socializing information resources” through a radical redesign of existing information and communication services from the viewpoint of network architecture.

The current Internet technology has developed by interconnecting computers that existed in disparate locations, allowing each computational resource to be utilized in a highly functional manner. However, focusing on social services such as YouTube, Instagram, and X and the utilization of digital objects owned by companies and individuals, many modern information resources are only used in a limited context, like pre-Internet computational resources confined to LANs . We believe that we need to reach a “utilization phase of another dimension” in which users can freely utilize these information resources beyond the boundaries of organizations and service providers.

To make this vision a reality, Kaneko Laboratory is working on the following four goals.

1. Realization of a personalized information space

We will build an environment where information that is truly valuable to each and every user is promptly provided when needed.

2. Establishment of a sustainable information processing infrastructure

Provide a sustainable information distribution and processing infrastructure that is scalable to an ever-increasing number of information resources and users.

3. Compatibility of information security and asset

Compatibility of information asset for information resource holders and privacy protection for users, and verification of authenticity and trustworthiness.

4. Creation of a competitive information service environment

We aim to create an open environment where anyone can be an information provider, an information user, and a curator of high quality information.

The most originality of our research lies in the fact that we try to solve these problems from the viewpoint of network architecture. We take the following technical approaches based on the idea of “networking” information resources.

・ “Graphification” of information resources

Abstraction of various information resources and their relationships in the form of nodes and edges is captured as a graph structure. This enables information distribution that protects the asset nature of information resources by hiding their concrete contents, and realizes an information processing infrastructure that is not bound by the contents of information resources. In addition, graph data is much lighter than the original data, making it scalable.

・ Utilization of information resources across service boundaries through interconnection of graphs

By dynamically interconnecting autonomously distributed managed graphs according to the services used by the users, a wide area information network is constructed. This enables free distribution and utilization of information resources beyond the boundaries of organizations and service providers.

・ Computer-centered personalized information search

Discover groups of information resources with potential value by having computers autonomously perform random walks on the constructed graph. This eliminates the processing bottleneck of existing large-scale search engines and LLMs, which assume human knowledge and understanding. In addition, the user can freely choose how to search, thus enabling user-centered, personalized information search with security considerations.

・ User-friendly information presentation by integrating with existing AI technologies

Information resources discovered by graph analysis can be used for content summarization, interpretation, and various application services by integrating with existing technologies such as AI and large-scale language models. This will provide information that is truly valuable to users in a form that is easier for them to understand, and explore new possibilities for information utilization.

Research Keywords

graph analysis distributed system personalization
inter-networking security