Research On Knowledge Service Component Of Cross Modal Big Data Of Science And Technology

With the vigorous development of science and technology,scientific and technological data has gradually become an indispensable part of people's life.Users usually want to have a more comprehensive understanding of the retrieved scientific and technological content when they query the corresponding scientific and technological content on the scientific and technological thesis platform and the scientific and technological information website.Therefore,it is of great significance to extract the deep features of scientific and technological cross modal big data,and realize the information fusion and complementarity of cross modal scientific and technological data through multi-modal semantic space learning.At the same time,it is of great significance for mining the association between scientific and technological data entities and analyzing the evolution law of scientific and technological entities.The mission for this thesis mainly includes the four following aspects:(1)For the text data of science and technology demand,this thesis proposes a text semantic extension algorithm based on the additional features of science and technology demand,and uses the depth feature dimension reduction network to extract the depth features of the text;for the image data of science and technology demand,this thesis proposes an algorithm using the spatial pyramid pooling layer to calculate the residual network.The experimental results show that the F-measure,recall and accuracy of the proposed depth feature extraction algorithm are improved in the comparative experiments.(2)This thesis proposes a unified semantic space learning method for cross modal technology big data.Based on the proposed depth mapping network and the network of spatial pyramid pooling layer,a cross modal technology big data depth mapping algorithm is proposed.DMCT algorithm mainly maps the depth features of text data and image data to the continuous and unified semantic space,and further maps to the discrete hash semantic space to improve the efficiency of accurate retrieval.Through the verification of the results of comparative experiments,the retrieval accuracy of cross modal big data of science and technology needs by using the unified semantic representation of cross modal science and technology data obtained by DMCT algorithm is higher than that of other comparative experiments.(3)This thesis proposes a algorithm that is multi-channel attention based entity relation extraction for science and technology demand big data.MCMA algorithm will extract the semantic association relationship of scientific and technological demand data in each time period,and store the data according to the technological demand text data semantic similarity and time duration coefficient of scientific and technological entities.Finally,the evolution law of scientific and technological entities is analyzed by graph group detection algorithm.The experimental results show that MCMA algorithm can effectively extract the association relationship according to the cross modal technology demand big data,and the calculated semantic relationship accuracy and other indicators are better than the comparison algorithm.(4)The knowledge service component system of cross modal technology big data is designed and implemented.The system mainly includes four modules:accurate retrieval of cross modal science and technology demand data,query of cross modal science and technology demand association relationship,analysis of science and technology entity evolution law and system display.It can effectively verify the algorithm proposed in this thesis,and has a certain practical value.