摘要
在大数据时代,医药专利数据的有效收集、整理和挖掘分析对医药行业发展愈发重要。当前文本分类神经网络对医药专利标签的分类准确率不够高,为了有效提升专利标签的分类效果,设计了一种基于注意力机制的双向长短时记忆神经网络分类模型。该模型避免了传统循环神经网络的长期依赖问题,并充分利用全局信息,以实现文本信息的权重分布。
In the era of big data, the effective collecting, collating, mining and analysis of medical patent data is becoming more and more important for the development of pharmaceutical industry. The current text classification neural network is not accurate enough for the classification of medical patent labels. In order to improve the effect of patent label classification, a bidirectional long short-time memory neural network classification model based on attention mechanism is designed. The model avoids the longterm dependence of traditional cyclic neural network and makes full use of global information to realize the weight distribution of text information.
作者
李绪夫
Li Xufu(School of computer science and technology,Chongqing University of Posts and Telecommunications,Chongqing 400065,China)
出处
《计算机时代》
2020年第5期50-53,58,共5页
Computer Era
基金
重庆市技术创新与应用示范(产业类重大主题专项)“医药专利大数据智能分析决策系统与应用示范”(cstc2018jszx-cyztzxX0025)。
关键词
深度学习
专利分类
长短期记忆网络
注意力机制
depth learning
patent classification
long short-term memory
attention mechanism
