Joseph Phillips

Ontology-guided knowledge discovery in databases

We present work in progress on a new methodology for leveraging the semantic content of ontologies to guide knowledge discovery in databases. Our system scans new databases to obtain type and constraint information, which users verify. Our system then uses this information in the context of a shared ontology to intelligently guide the potentially combinatorial process of feature construction. Further, our system learns each time it is applied, easing the users verification task on subsequent runs..

Towards a Method of Searching a Diverse Theory Space for Scientific Discovery

Scientists need customizable tools to help them with discovery. We present an adjustable heuristic function for scientific discovery. This function may be considered in either a Minimum Message Length (MML) or a Bayesian Net manner. The function is approximate because the default method of specifying theory prior probabilities is a gross estimate and because there is more to theory choice than maximizing probability. We do, however, effectively capture some user preferences with our technique. We show this for the qualitatively different domains of geophysics and sociology.

A Proposed Semantics for the Sampled Values and Metadata of Scientific Values

We propose semantics for the manipulation of sample values and metadata for scientific data. Our approach uses domain knowledge from an ontology; computational knowledge about dimensions, units, coordinate systems, etc; and cultural and linguistic norms. It recombines annotated values with specialized operators. This semantics does several types of dynamic error checking and results in “selfdocumented computation”.

Scilog: A Language for Scientific Processes and Scales

We present Scilog, an experimental knowledge base to facilitate scientific discovery and reasoning. Scilog extends Prolog by supporting (1) dedicated predicates for specifying and querying knowledge about scientific processes, (2) the different scales at which processes may be manifested, and (3) the domains to which values belong. Scilog is meant to invoke more specialized algorithms and to be called by high-level discovery routines. We test Scilog’s ability to support such routines with a simple search through the space of geophysical models.

Toward a Computational Model of Hypothesis Formation and Model Building in Science

We present our two-part approach to computational scientific discovery. The first part is a heuristic search through the space of hypotheses to find good hytheses. The second stage is a heuristic search through the space of sets of hypotheses to find coherent models. Both of these heuristic searches strive for four criteria: empirical support, simplicity, novelty and utility. Four types of domain knowledge inform both searches: transforms, taxonomies, sequences and default models. We mention several experimental results that suggest the viability of our approach.

Normothermic Ex Vivo heart perfusion: Effects of live animal blood and plasma cross circulation

Prolonged normothermic ex vivo heart perfusion could transform cardiac transplantation. To help identify perfusate components that might enable long-term perfusion, we evaluated the effects of cross-circulated whole blood and cross-circulated plasma from a live paracorporeal animal on donor porcine hearts preserved via normothermic ex vivo heart perfusion. Standard perfusion (SP; n = 40) utilized red blood cell/plasma perfusate and Langendorff technique for a goal of 12 hours. Cross-circulation groups used a similar circuit with the addition of cross-circulated venous whole blood (XC-blood; n = 6) or cross-circulated filtered plasma (XC-plasma; n = 7) between a live paracorporeal pig under anesthesia and the perfusate reservoir. Data included oxygen metabolism, vascular resistance, lactate production, left ventricular function, myocardial electrical impedance, and histopathologic injury score. All cross-circulation hearts were successfully perfused for 12 hours, compared with 22 of 40 SP hearts (55%; p = 0.002). Both cross-circulation groups demonstrated higher oxygen consumption and vascular resistance than standard hearts from hours 3–12. No significant differences were seen between XC-blood and XC-plasma hearts in any variable, including left ventricular dP/dT after 12 hours (1478 ± 700 mm Hg/s vs. 872 ± 500; p = 0.17). We conclude that cross circulation of whole blood or plasma from a live animal improves preservation of function of perfused hearts, and cross-circulated plasma performs similarly to cross-circulated whole blood.

Checking scientific assumptions by modeling

We describe extensions to the science querying system Scilog to enable it to make efficient simulators. Given a scientific models details and assumptions Scilog can create a C++ simulator to check assumption consistency. We used extended Scilog to test claims that Intelligent Design makes about Evolution and found them to be at odds with basic Biology.