面向国防科技前沿的开源信息主题识别
Subject identification of open-source information for defense science and technology frontiers
刘任烨 1于凯1
作者信息
- 1. 成都飞机设计研究所,四川 成都 610091
- 折叠
摘要
面向国防科技前沿细颗粒领域主题出现的不平衡数据分类问题,提出一种基于融合特征的国防科技前沿主题识别与分类模型.该模型在基于Sentence-BERT预训练模型提取的语义特征基础上,融合基于命名实体识别技术提取的实体特征,实现面向国防科技前沿领域特定专题的追踪、监测能力构建.实验结果表明,融合特征主题识别与分类模型有较好的模型指标.该模型已在具体实践中取得一定成效.
Abstract
A feature fusion subject identification and classification model based on deep learning to address the imbalanced data classification problem affecting fine-grained subjects in the field of defense science and technology frontiers is proposed.Based on the semantic features extracted using the Sentence-BERT pre-trained model,the proposed model fuses the entity features extracted with the help of the named entity recognition technology and establishes tracking and monitoring capabilities for specific subjects in the field of defense science and technology frontiers.The experimental results show superior model performance.The model has been practically applied and has achieved noticeable results.
关键词
国防科技前沿/主题识别/融合特征/语义相似度/命名实体识别Key words
defense science and technology frontier/subject identification/fusion feature/semantic similarity/named entity recognition引用本文复制引用
出版年
2024
NETL
NSTL
万方数据