Author

Date of Award

10-14-2017

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Computer Science

First Advisor

John Talburt

Abstract

Entity Resolution (ER) can be characterized as the process by which various records (references) become linked to the entities (identities) with which they related. The ER process becomes exponentially more complex and time consuming as datasets approach “Big Data” volumes. Datasets that are beyond the ER capabilities of even the most powerful computers are no longer unusual. The overarching objective is to significantly reduce the amount of elapsed time required to achieve all aspects of the ER life cycle when contrasted with conventional implementations and algorithms. Reductions approaching an order of magnitude are conceivable. Reductions in Big Data ER from weeks to days, days to hours, and similar can be attained given ER life cycle and ER via parallel deployments. The research effort implements distributed ER, via many computers working together in parallel. Specifically, Hadoop Map/Reduce distributed processing solutions addressing three of the challenges characteristic of large-scale (Big Data) entity identity information life cycle management: (1) Strategies to perform ER in the Identity Capture and Identity Update life cycle phases in the case where the ER match rules call for multiple indexing (blocking), (2) Strategies to maintain persistent identifiers for the same entity in the Identity Update Phase, and (3) Quantity ER timing reductions attainable via parallel/distributed ER methods.

Share

COinS