Construction of disease-pathological feature knowledge graph based on entity recognition and relation extraction

The pathological features of the diseases are helpful to the diagnosis of disease. There are no one-to-one correspondences between most of pathological features and diseases. If the detailed relations between all diseases and all pathological features are constructed, they can be used to assist doctors and the machine to reason from the huge knowledge relations. In this paper, an unsupervised entity recognition and relation extraction algorithm based on semantics is proposed, it can extract a large amount of relation data from massive pathological report texts, and with it the corresponding disease-pathological feature knowledge graph has been constructed. By processing 3 million web pages, 11755 effective disease entities, 83511 pathological features and 533 relation words are obtained. The correct rate of extracting relation words is 83.17%, and the recall rate is 73.26%. The experimental results show that the knowledge graph constructed in this paper is relatively complete and reliable.