A study on relation extraction of historical figures based on bibliographic description

Figure relation extraction is an important and hard field in information extraction. In this paper, aiming to improve the performance for relation extraction of historical figures, we propose a novel method based on bibliographic description. In the proposed method, by analyzing the species and co-occurrence relation of responsibility in a bibliographic record, we combine diverse person responsibility, person name and time as features, whose values are the quantity of the species clustering concerned, to build a Decision Tree model. Accordingly, relation extraction of historical figures is performed through the model. It is experimentally shown that on average, 83.3% and 83.0% in precision and recall rate are achieved respectively without more linguistic knowledge and complex classifiers.