Entity network extraction based on association finding and relation extraction

Abstract

One of the core aims of semantic search is to directly present users with information instead of lists of documents. Various entity-oriented tasks have been or are being considered, including entity search and related entity finding. In the context of digital libraries for computational humanities, we consider another task, network extraction: given an input entity and a document collection, extract related entities from the collection and present them as a network. We develop a combined approach for entity network extraction that consists of a co-occurrence-based approach to association finding and a machine learning-based approach to relation extraction. We evaluate our approach by comparing the results on a ground truth obtained using a pooling method. © 2013 Springer-Verlag.