Entity relation joint extraction method and system based on poultry BERT model and multi-label pointer network

• Main Authors: GU SHUSHU, JIANG GUOXIN, WU KUN, ZHANG HAOQIAN, SU QUAN, DONG YONGYI, CHEN XIANGXIANG, KUANG WEI, YEO JONG-DUK, LIU YONG, HUANG JINHU
• Format: Patent
• Language: Chinese; English
• Published: 01.11.2024
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; PHYSICS

The invention relates to an entity relation joint extraction method and system based on a poultry BERT model and a multi-label pointer network. The method comprises the following steps: step 1, inputting a text into a text coding layer for coding to form a text code; 2, precisely positioning and marking main body elements in text codes based on a multi-label pointer network; 3, systematically exploring all potential relationships and corresponding objects by adopting a subject attention-oriented normalization network strategy, and assembling a complete subject-relationship-object triple structure; and 4, fusing the probability balance strategy into loss calculation, and carrying out model training. According to the method, not only is the comprehensive efficiency of entity relation extraction optimized, but also the adaptability and accuracy of the model in a complex text environment are remarkably enhanced, and a powerful tool is provided for the fields of knowledge graph construction, information extraction