基于深度学习的图书馆文献精准检索系统

Library document precise retrieval system based on deep learning

针对传统检索算法与检索模型之间的对应性偏弱而导致的图书馆文献检索系统的精准度不高,且耗时较长等问题,文中研究基于深度学习的图书馆文献精准检索系统。在硬件方面,采用MT48LC4M32BP芯片对SDRAM电路进行扩展设计,选择SD卡的SD模式设计外接SD卡的卡座电路;引入示波器进行降噪处理,利用不易受分布参数影响的集成运放的方式设计阻抗变换电路。在软件设计上,建立一个语义相似度计算模型,并利用深度学习的深层结构和学习能力,设置与之匹配的文献精准检索算法,实现对图书馆文献的精准检索。最后进行仿真测试,测试共分为3个阶段,分别为存储正确性、检索结果正确性、检索效率。通过测试结果可知,此次设计系统的存储正确性、检索结果正确性均符合设计要求,检索效率较高。

In allusion to the weak correspondence between traditional retrieval algorithms and retrieval models,resulting in the low accuracy and long time⁃consumption of the library document retrieval system,a library document accurate retrieval system based on deep learning is researched.In terms of the hardware,the MT48LC4M32BP chip is used to expand the SDRAM circuit,the SD mode of the SD card is selected to design the card socket circuit of the external SD card,the oscilloscope is introduced for the noise reduction processing,and the impedance conversion circuit is designed by the integrated operational amplifier that is not easily affected by the distributed parameters.In the software design,a semantic similarity calculation model is established,and the deep structure and learning ability of deep learning are used to set up the matching document accurate retrieval algorithm to achieve accurate retrieval of library documents.The simulation test is conducted,which is divided into three stages:storage accuracy,retrieval result accuracy and retrieval efficiency.The testing results show that the storage accuracy and retrieval result accuracy of the design system can meet the design requirements,and the retrieval efficiency is high.