Проблеми масштабування семантичних інформаційних ресурсів зі складною структурою

Abstract

We analyze scaling problems arising in modern intelligent information systems (IISs) and classify main reasons for their occurrence in their practical solutions. IISs integrate various elements of artificial intelligence (AI) for acquisition of knowledge relevant to actual user tasks. Important properties of these IISs are use of data with complex structure and orientation on semantic information resources (IRs). Therefore we analyze main features of the Data-Centric AI and opportunities for acquiring domain knowledge in various representations from Big Data. Knowledge organization systems (KOS) provide models and methods for effective store, retrieval and use of information processed by the Web-oriented IISs, and we consider existing approaches for their software platforms.We analyse the specifics of the scaling for systems focused on the semantic information processing and its differences from traditional data and Big Data scaling. This specifics is caused by complexity of data structure, number of various semantic relations between information objects into IR and complexity of semantic queries executed by KOS. On example of e-VUE – the Wiki-portal of the Great Ukrainian Encyclopedia – we analyze various situations that arise in process of practical development of semantic information resources with large volume and complex structure. Various ways of semantic retrieval into this information resource that use possibilities of the Semantic MediaWiki plugin are considered from the point of view of scaling aspects (such as increase of information objects, their relations and complication of their structure and characteristics). On base of this analysis we generate a set of recommendations aimed at ensuring more efficient development of such resources and their efficient functioning for practical use.