A novel clustering algorithm for Unsupervised Relation Extraction
Jing Wang, Yang Jing, Yue Teng, Qingling Li · 2012
Clustering of entity pairs is the core content of the unsupervised relation extraction method. However, most of the clustering algorithm in the previous unsupervised relation extraction does not take into account the influence of the duality between entity pairs and the relationship characteristics on clustering results. In order to overcome this defect, this paper proposed a novel clustering algorithm for unsupervised relation extraction. It introduces co-clustering theory on the basis of the k-means clustering, not only clustering the entity pairs but also clustering the relationship characteristics to make full use of the duality of the clustering dataset. The final experimental results demonstrate that our clustering algorithm get more higher accuracy rate than k-means clustering algorithm in unsupervised relation extraction.