From Knowledge Engineering to Knowledge Science

This paper presents a brief analysis of modern sources of knowledge for artificial intelligence (AI) applications and discusses prospects of their development. It is shown that methods, algorithms, and technologies for knowledge extraction based on machine learning, as well as various knowledge extraction techniques that use the digital twin (DT) technology, are currently quite mature and in great demand. As for expert knowledge, it is rarely used in practice, even though it is essential in a number of critical and often unique classes of next-gen applications for which experts are the only available source of knowledge. To solve the problem of efficient access to expert knowledge, intensive research and development in the field of knowledge engineering is required, capable of elevating it to the level of knowledge science, which will be able to solve knowledge processing problems of the same scale and complexity that are currently solved in data science using big data. An analysis of potential areas of research and development in a hypothetical knowledge science is presented, and some methods, models, algorithms, and technologies for processing large volumes of raw fragments of expert knowledge to serve the next generation of intelligent applications are considered.