Process Ontology Languages
From Semantic Portal Wiki
Process studies of natural systems and their associated features and phenomena have a long and valuable history in science endeavors. For example many studies move back and forth between postulating possible processes and analyzing data to support or refute the postulates – see for example the study of the El-Nino Southern Oscillation (ENSO; Walker 1923) and the corresponding La-Nina, and the study of the role of gravity waves in coupling of the middle and upper terrestrial atmosphere (Fritts 1995a, b). Scientists are used to writing computer programs in procedural and compiled software languages developed by computer scientists. However no work has aimed specifically at declarative representations for exploring the synergies between scientific data and natural science processes to yield a computational environment that is geared toward supporting interdisciplinary teams of natural and computer scientists. These teams currently attempt to model scientific observations with the goal of sharing, reusing, and integrating models, data and observations. The teams are often globally distributed, highly interdisciplinary and evolving. The interplay among data, knowledge and understanding of natural processes is sequential and often inefficient, ineffective or disconnected. Our goal is to develop both the scientific principles and the technological toolkits needed to significantly decrease the effort that scientists must go through to extract scientific knowledge from the very large data sets being generated in this area.
In humans, the process of going from data to knowledge is part cognition, part algorithmic (often performed using a computational aid) and part intuition. The resulting knowledge, or indirect pointers/ reminders to that knowledge (such as the authors of relevant papers) are stored in the brain. For tools to properly help scientists with their tasks, however, the purely algorithmic is not enough. Computers must be able to help scientists to formulate and access encodings of not only the scientific data but also some or all parts of the processes (and their inputs and outputs) involved. The tools of the scientist must thus be extended to allow scientific data to be transformed into knowledge via a process that takes in observations, formulates representations, and integrates those representations into a knowledge base for storage and later retrieval. Rather than just saving the data, the knowledge itself must be kept in a computer manipulable form.
However, despite decades of work on knowledge representation, the work has primarily resulted in good options for languages for representing objects and data e.g., KRL (Bobrow and Winograd, 1977), KIF (Genesereth and Fikes, 1992), and more recently OWL (McGuinness and van Harmelen, 2004). In the past few years, motivated by the World Wide Web Consortium’s Semantic Web Activity, commercial interest has grown and more tools and support are emerging for applications using knowledge encoded in standard languages such as XML, RDF, RDFS and OWL. The term “ontology” has entered the lexicon of the scientist, using these languages to describe the terminologies of the objects and data in their areas. However, research efforts on process representation and manipulation have not been as prolific, and ontologies that relate to processes lag significantly behind those of objects. In this area, some foundational work exists (e.g., PSL, now an ISO standard, BPEL) for representing processes aimed at facilitating capture and exchange of process information for manufacturing (Bock and Gruninger, 2005) and PSL’s ability to represent more general processes has been recognized and work has proceeded in connecting it with today’s ontology languages with the resulting SWSL effort (Battle et al, 2005). However, PSL is not adequate for the needs of representing the complex processes of large-scale scientific systems such as occur in the heliosphere, does not provide standard mechanisms for tracking provenance, and reasoners that can use process models to interpret data are not yet available. These latter needs are what we address in this project. our goal is to be in a position five years from now where the fundamental conduct of science,the discovery, and understanding of natural phenomena and processes, is integrated with state-of-the-art computational thinking, enabled by declarative process-oriented knowledge representation and reasoning tools.
We have conceived this project to provide the fundamental and exploratory work necessary for the new natural process and mathematic algorithm ontologies (NPO, MAO) that we believe is critical to sharing and integrating scientific data across and within disciplines. A primary goal is to generate declarative representations of processes for integration and reuse, we must build modular representations with explicit evolution and provenance information. Based on our previous efforts, we have learned that this is the distinguishing factor that supports reusability and longevity of the ontologies. We believe the key to future integration is for ontologies to be broadly maintained and reused by communities rather than continuing the current practice of a number of individual efforts. The potential range of natural processes that are studied across science is immense: from diffusion at molecular scales, to shock waves within our interplanetary environment to formation and evolution of the large-scale structure of the universe, as well as processes that span scales, such as turbulent cascade in homogenous turbulence. As an increasing number of sciences are being challenged with making scientific discoveries with exponentially increasing amounts of data, communities are finding that new discoveries are being made at the edges of their discipline, or more often, between disciplines.
When we want knowledge to be computer manipulable form what items do we consider manipulable - objects involved, numerical equations, assumptions? Why should machine understand the mechanism behind a process, what can it do with that? We don't know everything, and there are new discoveries in sciences on a regular basis. Can we provide tools to make that process easier so that a scientist can try different mechanisms he thinks possible quickly to test out his hypotheses. QR may not be used for very sensitive mechanisms (i.e. every change of 1 or 2 degrees brings drastic changes in the mechanism.)
Semantic Web advantages without a new process language- start tagging words/ vocabularies (including those referring to objects) with URIs or by replacing words with dereference-able URIs. Assign every description with URI and annotate them with information about authors, related models. These descriptions can be then annotated with assumptions, etc. A standards ontology for describing assumptions, defining transitions from one set of assumptions to others?
Different levels/ natures of specifications - Flow chart of broad processes to express enforced temporal relationships, and each process (system) has description of its dynamics in QDE and numerical model.
