基于BART神经网络的爆破文献摘要生成与优化  

Generation and Optimization of Explosive Literature Abstracts Based on BART Neural Network

作　　者：洪珂 李琳娜[1] 

机构地区：[1]武汉科技大学理学院,湖北 武汉

出　　处：《人工智能与机器人研究》2024年第4期814-821,共8页Artificial Intelligence and Robotics Research

摘　　要：针对爆破行业的从业人员对爆破英文文献阅读难度大,而又希望快速获取文献核心信息的需求。本文基于BART神经网络进行优化,对于爆破英文文献进行针对性提取出生成式摘要,并与传统的抽取式摘要进行对比,结果表明基于BART神经网络的摘要结果在流畅度和信息完整性上更加优秀,对于爆破英文文献的信息内容提取有很大的帮助。Aiming at the needs of blasting industry practitioners who find it difficult to read English literature on blasting but want to quickly obtain the core information of the literature, this paper optimizes the BART neural network and extracts generative summaries for English literature on blasting. The summaries are compared with traditional extractive summaries. The results show that the summary results based on the BART neural network are more excellent in fluency and information completeness, which is of great help in extracting information content from English literature on blasting.

关 键 词：爆破英文文献 自然语言处理(NLP) 神经网络 

分 类 号：TD2[矿业工程—矿井建设]