Extracting Online Recruitment Information Based on BiLSTM-Dropout-CRF Model

To improve the feature learning based on different job requirements, BiLSTM-Dropout-CRF (BLDC) model was proposed. Firstly, the original sentence sequences are imported into the embedding layer to obtain the word vectors. Then, BiLSTM and the dropout layer are used to learn the contextual information and key features. Finally, the CRF layer is used to accomplish the optimal sequence labeling and complete the training. For evaluating the model performance, precision, recall rate and F1 score are used to assess the extraction accuracy. The result shows that compared with traditional models, the F1 score of BLDC model severally increases 4.4% and 1.5% averagely based on two datasets about different industries and positions. It adequately illustrates that the effectiveness of online recruitment information extraction has been promoted.