Auto-Adaptive Systems for Monitoring and Updating Dynamic Knowledge Graphs

点击翻译按钮获取中文摘要

Given the extraordinary volumes of fast moving data on the Internet, users need cognitive services to help them make complex decisions based on the best available knowledge. In addition, these services should continually acquire new knowledge from the data fed into the system by mining data for information. The cognitive ability of a electronic commerce system relies on its global dynamic ontology which can be represented by dynamic knowledge graphs. In this research project, we propose a multi-agent system for effectively monitoring and updating large dynamic knowledge graphs of a semantic eCommerce search engine "daneel", which has been developed by Azzimov Inc. The proposed system will be used as an algorithmic tool to categorize eCommerce servers of Azzimov's interest on the Internet and optimize the graph for efficient search. One of the challenges in building such a tool is to calculate the chromatic number of a large scale, dynamic graph in a near real time manner. While calculating chromatic number of a graph is a combinatorial optimization problem that has been well studied in graph theory, however, existing approaches do not satisfy Azzimov's requirement of computing chromatic numbers of very large graphs with near real time responsiveness in a distributed and dynamic environment, such as the Internet. With the proposed multi-agent system for computing chromatic numbers, Azzimov's cognitive service system can monitor and updates its dynamic knowledge graphs with better speed and accuracy which enables better search results. The multi-agent system is designed based on the Auto-Adaptive Systems (AAS) architecture. The design of decision making modules of the agents and the interaction protocol between them are based on a distributed graph colouring algorithm which will also be developed in this project.