Multi-language person social relation extraction model based on distant supervision

Relation extraction refers to a method of efficiently identifying entities from the text and extracting semantic relations between entities. The person social relation extraction is one of the most important fields in relation extraction. A large number of techniques have been proposed on relation extraction thus far, and supervised machine learning methods are the most widely used. However, the disadvantages of supervised machine learning methods are that manually annotating training data set is costly and time-consuming, which block the improvement of the supervised relation extraction model. Aiming at the limitation, we propose a novel person social relation extraction model on both Chinese and English corpus with distant supervision. Distant supervision method can make full use of the information in the knowledge base and provide training data without manual effort. In particular, it is an effective method in the very large corpora which contains thousands of relations. In this model, we use distant supervision method to get the weak-labeled data set. Then, a supervised method is used to train a classifier, which is expect to distinguish the relation between the person entities in the input sentence. Experiment results on real-world datasets show that, our model can take advantage of all informative sentences in knowledge base and outperforms several competitive analogous methods, what's more, it does not need any human-labeled training data.