ONTOLOGICAL GEOSCIENCES

Since the end of last year, the author has led a team of graduates and senior undergraduates at China University of Geosciences (CUG) to experiment using the geoscience concept models to represent theme-oriented knowledge in geoscience literatures. One of the objectives of this project is to integrate geoscience data using a subject-oriented concept model. Another objective is to explicitly characterise, within the concept model, the knowledge that is related to or implied in the theme-oriented knowledge, but is extracted from published literatures other than those based on which the concept model was developed, in order to gain further insight of the mechanisms a concept model should possess. This project is still ongoing. However, there are two conclusions we seem to be able to draw from the progress made so far:

1.The knowledge related to and implied in any geoscience concept, conclusion or proposition is unboundedly vast, as it relates to almost every other concept in geosciences via some relations;

2.Geosciences contain various recognition patterns, including those that can be used to seek or infer new geo-scientific knowledge.

Therefore, we consider that it may not be possible that a conventional automated computer system could be developed such that it can correctly comprehend and operate the geoscience concept model that is capable of figuring out new geo-scientific knowledge patterns. Furthermore, it does not seem feasible that one can just use a limited number of geoscience concept models known to us to consistently provide adequate representations of such complex and ever evolving knowledge. For the geoscientists, the knowledge so represented may be especially hard to accept. For these reasons, we propose to develop ontological geosciences.

We refer ontological geosciences as a kind of geosciences that is based on geoscience data and encompasses all the contents of the traditional geosciences. The essential characteristic of the ontological geosciences is that it is composed of discrete geoscience elements formally. Hence, it possesses better correlation, integrity, and consistency than the traditional geosciences. Similar to the traditional geosciences, the ontological geosciences must first be recognizable and operable to people who want to use it, especially the geoscientists; second it must also be programmable, i.e. recognizable and operable insofar as the computers are concerned, which is an important characteristic of the ontological geosciences. However, it must be said that the ontological geosciences should be neither considered nor developed as an ordinary computer system; it is ultimately a programmable knowledge system.

The traditional geosciences is a study mainly concerned with the knowledge of the structures, constitutes and evolution of the Earth. Whilst the ontological geosciences focuses on the study of the underlying structure of the geoscience knowledge, or the way geosciences is represented so that it can be used to represent geoscience knowledge on the computers. The main challenges faced with the ontological geosciences are:

1.How to extract and standardise geoscience knowledge patterns from the traditional geosciences, and use the knowledge patterns so derived to represent the partial geoscience knowledge that has not been conclusively proven and generally recognised, and

2.To determine the extent to which one can surmise, generalize, and unify these knowledge patterns.

The subjects of geosciences are the Earth and its evolution processes, both of which are beyond human being's control. Our knowledge of the Earth has not been and will never be complete; all the geoscience knowledge we acquired through a gradual process of studying the nature is in theory partial. Having studied new natural phenomena they have observed, the geoscientists often provide explanation by proposing theories or hypotheses, which are often personal and subjective. The verification process of these theories and hypotheses may take a long time, because the Earth is an extraordinarily complex entity and the geo-scientific survey is very expensive. On the other hand, in order to enable the computers to recognize and operate geosciences, we must assume that we have known the underlying structures of geosciences and that we can use them to “teach” computers to understand geosciences. We hope we can also use the general or commonly recognised knowledge patterns to represent geoscience knowledge so as to develop just one computer system to support the whole geosciences.

Our solution to the aforementioned challenges is to design a user-computer collaborative learning hierarchical ontological geoscience system. Its architecture is illustrated in Figure 1. Firstly, the ontological geoscience is a hierarchical system of geoscience knowledge. The hierarchy of the architecture is not limited within the geoscience concept model; it extends to the geoscience knowledge levels, including the levels of the knowledge base that can be used to design the model or applications. Secondly, it is a system that is self-improving through the collaborative efforts between system users and computers. The computer network facilities provide a convenient mechanism for communication and sharing knowledge among the system users. The users gain knowledge from the system and improve it according to their knowledge of geosciences. One of the important learning and improving process takes place between the hierarchical levels. The users learn the knowledge patterns from the upper layer and transform them into the lower level model or applications. Lastly, the middle layer of the architecture can be regarded as “dictionary based geosciences”, i.e. the geosciences that is only represented by the terminologies in the dictionary. The dictionary provides the meaning of the geoscience data in the databases as well. Therefore, the ontological geosciences can be built up on the basis of geoscience data via the dictionary.

In summary, ontological geosciences is a structured geosciences that is geoscience data based, and is recognizable and operable on the computers. In view of the current status of research and development, we regard it as being essentially a human being centred geosciences. It is the people who learn and generalize the knowledge patterns that are deemed to be appropriate for the computers to represent geoscience knowledge, and teach the computers to understand the ontological geosciences and to support the geo-scientific research.