In order to solve the problem of fast and accurate identification of food safety events from large-scale Web texts and the extraction of entity relations, which is limited by the complex grammatical characteristics of Chinese, a method of entity relation extraction based on dependency parsing for news texts of food safety events FSE_ERE (entity relation extraction of food safety events) was proposed. This method combined the dependency parsing results of sentences with the entity relation extraction model to conduct unsupervised entity relation extraction for unstructured Chinese texts, and also introduced a semi-supervised classification method combining text similarity with positive and unlabeled learning (PU learning) classification method, which used an improved feature weighting processing method to improve the classification accuracy. That can make the FSE_ERE method to complete the entity relation extraction work in the high-quality news text of food safety events. The experimental results showed that the FSE_ERE method achieved advanced performance in entity relation extraction on food safety event news text dataset and multi-type hybrid news text dataset, and the F-measure achieved 71.21% and 67.42% respectively, which proved the effectiveness and portability of the FSE_ERE method.