Knowledge Extraction via Learning Processes and Data Models with Imprecision

点击翻译按钮获取中文摘要

The web represents an immense repository of information. A number of sources of structured and even more often unstructured data is growing every day. There is no doubt that our dependency on data increases continuously. However, the increased amount of data - although recognized as positive and beneficial fact - creates an even bigger issue related to our ability to fully utilize that data. That situation increases the pressure to develop automatic and more efficient approaches suitable for advanced processing of data leading to creating logic structures that ‘lifting' raw data into a knowledge-like level. ***Fortunately, we are at the onset of significant and far-reaching changes in the way data are represented and stored on the web. The concept of knowledge graphs becomes a new way of expressing pieces of data and relations between them. Resource Description Framework - one of the most fundamental aspects of the Semantic Web - is recognized as the most suitable data format for representing knowledge graphs. ***The proposed research project puts a special emphasis on processes of constructing, updating and utilizing knowledge graphs built based on data and information obtained on the web and extracted from documents. A key innovation of this project is a fusion of web technologies, fuzzy-based techniques, and concepts of category theory and topos to fully explore data taking advantage of Resource Description Framework's intrinsic interconnectivity, and setting up a basis for knowledge synthesis processes.***These activities will lead to establishing coherent rudiments of knowledge creation processes. In a nutshell, the proposed methodology focuses on forming knowledge-rich structures following the steps: 1) extracting information from documents and representing it in a form of so-called information knowledge graphs that contain specific pieces of information; 2) clustering and generalization of those graphs leading to construction of conceptual knowledge graphs; 3) maintaining both types of graphs via incremental updates using aggregation and data assimilation techniques taking into account imprecision and confidence levels in different pieces of data; and 4) constructing logic structures in a form of internal logic of topos based on conceptual knowledge graphs and linking those structures with information graphs for validation and cognitive purposes.***It is expected that the project will lead to significant contributions in methodologies aiming at building a new generation of systems that support the users in their activities related to collecting data from the web, and processing it towards creation of knowledge. This will lead to development of knowledge systems capable of validating correctness of information extracted from data, and synthesizing new concepts based on it.***Overall, the project encompasses state-of-the-art research and HQP training in an important for Canada IT area.