Pattern-based Aggregation of Named Entity Extractors

Despite significant advances in named entity extraction technologies, state-of-the-art extraction tools achieve insufficient accuracy rates for practical use in many operational settings. However, they are not all prone to the same types of error, suggesting that substantial improvements may be achieved via appropriate combinations of existing tools, provided their behavior can be accurately characterized and quantified. In this paper, we present an inference framework that leverages the joint characteristics of their error processes via a pattern-based representation of extracted entity data. This approach has been shown to produce statistically significant improvements in entity extraction relative to standard performance metrics and to mitigate the weak performance of entity extractors operating under suboptimal conditions. Moreover, this aggregation methodology provides a framework for quantifying uncertainty in extracted entity output, and it can readily adapt to sparse data conditions.