Generalization of Named Entities - Using Linked Data

Abstract

In the climate sciences, it is not feasible for a scientist to read more than a small fraction of all the papers currently being published. Our group is attempting to help climate scientists by making a tool for discovering knowledge in scientific literature. Our approach involves the extraction of causally related events involving either increase or decrease. To facilitate reasoning and search, the relevant parts of the extracted variables need to be identified and generalized. This is done by a combination of pruning the variables, Named Entity Recognition, and using background knowledge about recognized entities. In the semantic web, an ever increasing amount of machine readable information is becoming freely available. With growing support for Linked Data standards and open licensing, there is now an abundance of structured data waiting to be explored and utilized. This study seeks to find and utilize knowledge bases in order to generalize specific types of named entities, and finds that Linked Data resources are well suited for this task.