Research And Application Of Key Technologies Of Knowledge Graph Of Trading Chain On Dark Net

Knowledge Graph has broad application scenarios which can extract required information from vast amounts of texts and images.Domain Knowledge Graph is one of the Knowledge Graph.The correlational research of Domain Knowledge Graph could hardly be found,because the data of it is difficult to access and process.As a mysterious field,the dark net is different from the general network.Most of the trades on the dark net are illegal,and they could even threaten national security.Therefore,it is urgent to build the Knowledge Graph of the dark net.Aiming at those problems,the main research in this thesis is research and application of the key technologies of Knowledge Graph of trading chain of dark net.The main contents include:(1)this thesis designs and implements a Chinese named entity recognition model for the dark net.First of all,in view of the lack of clear entity classification in the dark net,through the research of the dark net,the entities that can be identified are divided into 9categories.Then,preprocess the data set and annotate the data of the named entity recognition.After that,the word vector of the dark net Chinese transaction data combined with the location information is obtained by using the Transformer bidirectional encoder model,and the word vector is input into the Bi-directional Long Short-Term Memory to obtain the word vector with context semantics.Then,the word vector is used to add weight to the local feature of the sentence through the attention mechanism.Finally,the above results are input into the classifier to obtain the relational category of the sentence.The above training is used to identify the entities in the dark net domain.Compared with other traditional methods,the proposed model has a significant effect on the Chinese named entity recognition task on the dark net,and its accuracy,recall and F1 values have achieved 74.9%,76.2% and 75.5%.(2)this thesis designs and implements a Chinese relation extraction model for the dark net.First of all,the abstractable relationships of the trading chain of dark net are divided into 5 categories through investigation.Then,the data set is annotated with relation extraction data to obtain the experimental corpus.Then,Word2 vec is used in this thesis to train word vectors,and word vectors combined with location information are put into the network of bidirectional gating units as input to obtain word vectors with context semantics.Finally,the word vector is used to obtain the importance degree of the word in the sentence through the attention mechanism,and different weights are given according to whether the word is a target relationship or not.Through the above training,the purpose of identifying the relationships in the dark net domain is achieved.Through comparative experiments,the model can effectively extract the relationships between dark net entities,and the accuracy,recall and F1 values of the model are 65.1%,62.3% and 63.7%.(3)this thesis designs and implements a Knowledge Graph system of trading chain of dark net.Firstly,the Knowledge Graph of trading chain of dark net is preliminarily constructed,and the triple knowledge obtained above,combined with entity supplement and relationship supplement,is stored in the Neo4 j database.Then,based on the knowledge query of the Knowledge Graph,the application of the dark net trading chain Knowledge Graph is realized and visualized.Finally,the purpose of presenting the dark net trading chain and tracking the key targets in the dark net is achieved,and each function of the system is ensured to complete the response within 2 seconds.