Entity Relationship Extraction of Chinese Enterprises on Web Data

Entity Relationship Extraction of Chinese Enterprises on Web Data

Enterprise entity relationship extraction is an important part of entity relationship extraction. Extracting corporate relationships from open data is of great significance in market analysis and selection of business partners. Due to the complexity of grammar and flexible expression in Chinese language, the traditional method for extracting Chinese enterprise entity relationship has a very poor effect. We propose an algorithm based on the integration of dependency grammar analysis of self-adaptive attention mechanism and long short-term memory network (DEP_ATT_LSTM) by vectorizing the text on which word segmentation is performed and inputting it into the LSTM network to obtain the text feature representation of sentences, then adopting self-adaptive attention mechanism based on dependency parser to calculate the weight of the text feature, and sending the obtained feature vectors into a classifier for entity relationship extraction. Experiments prove that the algorithm performs well. The accuracy, recall rate and F1 value reach 83.23%, 89.55% and 86.81%, respectively.