计算机科学
仿射变换
稳健性(进化)
旅游
联想(心理学)
数据挖掘
信息抽取
人工智能
情报检索
模式识别(心理学)
自然语言处理
地理
数学
生物化学
化学
哲学
考古
认识论
纯数学
基因
标识
DOI:10.1109/iceib57887.2023.10170172
摘要
For the nested use of entity objects and overlapping associations in extracting tourism English text, a joint extraction model BAMRel is proposed using the dual affine attention mechanism for obtaining entity object-related data information. This model uses the dual affine attention mechanism with common encoding layer parameters to form an information matrix of the entity object identification and the association extraction. At the same time, association extraction is integrated into the entity-type information system, enhancing the role of association extraction and the interaction between the two types of tasks. Due to the use of remote monitoring and manual verification to extract TFRED from the tourism English text information relationship dataset, the F-1 value of BAMRel mode in this dataset exceeds 91.8%, effectively overcoming association nesting and data duplication. To demonstrate the robustness of the pattern, comparative experiments are conducted on the DuIE dataset and mainstream joint extraction patterns, and the BAMRel pattern achieves the highest F-1 value of 8.2%.
科研通智能强力驱动
Strongly Powered by AbleSci AI