System Architecture

In training stage, our biologists firstly constructed an entity knowledge base (EKB) consisting of biology vocabularies through online biology databases. Then we used EKB to label word/phrase of NEs/sentences in Revised JNLPBA training set. The labeled NEs were used for generating principles by principle generation. Our pattern matching method used principles to identify BNEs from labeled sentences, and learned model for each principle throught logistic regression model. In test stage, test sentences were also labeled by EKB. Then the pattern matching is used to identify BNEs from labeled sentences. As shown in the following, and please read our manuscript for more detail.