基于深度学习的智慧图书馆文献快速检索

Fast retrieval of documents in smart library based on deep learning

图书馆数字化信息量呈指数式增加,导致文献检索过程中运算复杂度高,检索速度下降。为此,提出基于深度学习的智慧图书馆文献快速检索方法。以文本类文献为检索目标,构建文本语义矩阵,采用标记判定陈述句、求解信任事实可信度及文本整体可信度;针对纯文本形式与向量形式输入输出语料,基于单隐藏层反向传播神经网络与自组织映射空间结构,构建多层深度学习模型,实现不同形式文献快速检索。从某智慧图书馆选取六种学科文献组成实验数据,采取查全率、查准率以及检索效率评估检索方法性能,验证所提方法检索的有效率较好。

The amount of digital information in library increases exponentially,which leads to high computational complexity and low retrieval speed in the process of documents retrieval.Therefore,this paper puts forward a fast literature retrieval method of smart library based on deep learning.Firstly,the text literature is treated as the retrieval target to construct the text semantic matrix,and the declarative sentence is determined by marking,the credibility of trust facts and the overall credibility of the text are solved.Regarding the input and output corpus in the form of pure text and vector,the multi-layer deep learning model is constructed to realize the rapid retrieval of different forms of literature based on the single hidden layer back propagation neural network and self-organizing mapping spatial structure.Six kinds of subject literature are selected from a smart library to form experimental data,besides,the recall,precision and retrieval efficiency are used to evaluate the performance of the retrieval method.