Research on the Extraction of Entity Relationships From Fusion Syntactic Information

Information extraction is an important branch in the field of natural language processing (NLP), and relationship extraction is particularly important as its basic task. At present, most approaches in this domain represent relationship extraction with a single word vector or combine different features to process relationship extraction. However, different methods have different advantages and disadvantages in the classification of different relationship types. Therefore, this paper proposes a method to separately use the syntactic structure information in syntactic information and the shortest dependency path as an input to the experiment. Then combined with the advantages of each method, the relationship types with outstanding classification results are extracted to increase the weight as an important basis for the subsequent relationship classification and the algorithm is optimized. Finally the relationship type is determined through the softmax activation function. At the same time, in order to capture the local key information of the sentence, the convolutional neural network (CNN) is incorporated to improve the feasibility of the experiment. This method not only extracts more sufficient physical relationships, but also enriches selectivity; the experimental results show that the F-score increases by 5.85%, which proves that this experimental method is feasible.