Knowledge Graph And Accurate Portait Of Scientific And Technological Academic Conferences

In recent years,with the continuous progress of science and technology,the number of scientific research achievements is increasing day by day,as the exchange platform and medium of scientific research achievements,the scientific and technological academic conferences have become more and more abundant.The convening of scientific and technological academic conferences brings large number of academic papers,researchers,research institutions and other data,and the massive data bring difficulties for researchers to obtain valuable information.Therefore,it is of great significance to use deep learning technology to mine the core information in the data of scientific and technological academic conferences,and to realize a knowledge graph and accurate portrait system of scientific and technological academic conferences,so that researchers can obtain scientific research information faster.The main contributions include:(1)The method of data acquisition and entity extraction of scientific and technological academic conferences is proposed.The Scrapy crawler technology is used to obtain the data collection of scientific and technological academic conferences.Owing to the multi-source and heterogeneity of the data,targeting data matching rules are designed to obtain and preprocess the data,and store to the database.Next,with entity extraction of scientific and technological academic conferences,a named entity recognition method combined keyword-character LSTM and attention mechanism is proposed.The pre-trained model is used to obtain keyword vectors,and the character-level vector is extracted in the embedding layer.It is fused with word vectors to obtain potential semantic information,combined with bidirectional LSTM and attention mechanism at the network layer,and considers context information and global text information at the same time,this algorithm achieves better recognition results.(2)The method of semantic similarity calculation and knowledge graph construction in scientific and technological and technological academic conferences is proposed.Since there is no obvious part-whole,upper and lower relationship between academic conferences,the similarity is used as its association relationship.For the data set of scientific and technological academic conferences,the Siamese-BERT semantic similarity calculation algorithm fused with domain feature is proposed(SBFD),which integrates the domain features of the Siamese-BERT semantic similarity calculation algorithm by adopting the Siamese network of fusion domain features,and the fusion of technical entities,keyword entities and other information as vector input.The experimental results show this algorithm has better performance on the dataset of scientific and technological academic conferences.After obtaining entities and associations,we use them to build domain knowledge graphs,study the storage of knowledge graph data in scientific and technological academic conferences,and visualize them.(3)The trend prediction and accurate portrait construction method in the field of scientific and technological academic conferences is proposed(TPMR).Using multiple independent LSTM networks to learn different features,fusing the learned content to encode,using attention mechanism and LSTM network to decode,realizes the development trend and prediction of different fields.The experimental results show the effectiveness of the TPMR algorithm.Combined with Java Api and ECharts component library,building accurate portraits of scientific and technological academic conferences,and visualizing prediction results by field distribution,word clouds,etc.(4)Designing and implementing the knowledge graph and accurate portrait system for scientific academic conferences.The system obtains and extracts entities from scientific and technological academic conferences,analyzes semantic similarity,predicts domain development trends,and uses knowledge graphs and precise portrait construction to display the results,which verifies the feasibility of the algorithm and engineering implementation proposed in this thesis.Using ElasticSearch to build a retrieval system as the main interaction method,it provides functions such as display of popular conference papers,conference retrieval,paper retrieval,conference knowledge graph,and field trend analysis.Finally,the system function and stability are tested.