Core Discovery and Relation Extraction in Organization Profiling

Abstract

A comprehensive organization profile can be used for targeted collaboration and information analysis, as well as offering freshman a more objective resource to know an organization such as a lab or a party well. User profile has been studied for many years and there has been lots of applications based on it. Considering many differences between user profiling and organization profiling such as dynamic attributes, we propose the concept of organization profiling which had not been investigated before and present two aspects of our work: core discovery and community detection, relation extraction. We come up with a Double BGLL method based on Core discovery(DBGLL_C) for community detection which visually displays the core relationships and communities in a graph; We also improve an Unsupervised Chinese Open Entity Relation Extraction (UCOERE) approach, results of which show improvement on precision and aligned results on recall and F1 value. Extracted relations can be used to classify different cores, both of which will then also be used together for interest discovery for the whole organization in our future work.