Content profiling in digital preservation is a crucial step that enables controlled management of content over time. However, large-scale profiling is facing a set of challenges. As data grows and gets more diverse, the only option to control it is to combine outputs of multiple characterization tools to cover the varieties of formats and extract features of interest. This cooperation of tools introduces conflicting measures and poses challenges on data quality. Sparsity and labeling conflicts make it difficult or impossible to partition, sample and analyze large metadata sets of a content profile. Without this, however, it is virtually impossible to manage heterogeneous collections reliably over time.
In this paper, we present the content profiling tool C3PO, which includes rule-based techniques and heuristics designed for conflict reduction. We conduct a set of experiments in which we assess the effect of creating such a mechanisms and rule set on the quality and effectiveness of content profiling. The results show the potential of simple conflict reduction rules to strongly improve data quality of content profiling for analysis and decision support.