CAREER: iProvenance: Integrating Data Provenance with Software Traceability

University Of Washington, Seattle WA

Investigators

Abstract

Scientific research is increasingly data-driven -- hypotheses, simulations, and models, which shape our understanding of the world, are formulated based on observed data. In order to make valid inferences from these observations, researchers are finding that they also need to understand data provenance, which is the origin and context of formation of the data. To tackle the increasing amounts of data, scientists rely on many different software programs (both homegrown and commercial) to collect, process, and analyze data. While the eScience community has made strides toward supporting data provenance, understanding the software that manipulates the data is a key research area that has largely been unexplored. Meanwhile, in software engineering, a concept related to provenance known as software traceability plays a key role in understanding source code. Traceability enables engineers to link source code to various pieces of documentation to answer questions such as "Why is this code used?" and "What tests were executed on the code?" Software traceability also facilitates other development tasks, such as debugging, determining software quality, and ascertaining the impact of a change. This CAREER Award project focuses on the crucial interplay between data and software in eScience by using a conceptual framework, iProvenance, that integrates data provenance with software traceability and is grounded in a holistic examination of provenance challenges. iProvenance provides the ability to model and capture software-centric and data-centric information in tandem, create provenance methods that holistically address challenges similar to those in software traceability, and develop powerful yet accessible automated provenance software for scientists in various domains. The approach is beneficial to every scientific field that engages in eScience and data provenance. The results of the research also have the potential to transform the management of provenance in industries with large amounts of digital records, including electronic health records and clinical trials. Finally, the educational activities interwoven with research activities will not only equip future computer science professionals with knowledge in data provenance techniques and methods, but these activities will also provide curriculum and educational materials that can be used at other institutions.