Research And Application Of Multisource Data Intergration Based On Ontogy

With the rapid development of information technology,the form of information representation has become increasingly rich,and there exists a huge amount of data and valuable information in the Internet.However,different data and information can only be exchanged internally.There are many difficulties in information exchange.In addition,a single data source cannot reflect data feature information in many aspects and carries incomplete information attributes.The integration of multi-source data can provide a better representation of data information,which is of great significance to information exchange and reuse.Therefore,this paper conducts an intensive study on the task of multi-source data fusion from three aspects: ontology modeling of multi-source data,ontology similarity calculation,and semantic query of integrated database,and the specific work is as follows.First,this paper adopts ontology as data representation form and constructs hybrid ontology model to carry out data integration work,which solves the problems of traditional methods with poor knowledge acquisition capability and unclear semantic expression.Based on the existing ontology construction methods combined with domain knowledge features,the construction process of domain-specific ontologies is designed,and the matching rules from database data to ontology data are formulated to complete the mapping of different data to ontology data.Secondly,this paper proposes an ontology similarity calculation method based on graph convolutional network to solve the similarity calculation problem in the ontology mapping process.Ontology mapping is a key step in the data integration process,and the calculation of similarity is the core of ontology mapping,and the graph convolutional network effectively uses different node information and has superior ability to preserve relational information.The experimental results show that the graph convolutional network outperforms other models in the ontology similarity calculation task,and achieves the largest Pearson coefficient value,and the error maximum,error average,and error standard deviation are the smallest,which proves that the model can efficiently solve the ontology similarity calculation problem.Thirdly,this paper proposes an interrogative entity recognition method based on ERNIE-Bi LSTM-CRF to solve the entity recognition problem in the process of semantic query application.The method captures contextual information through Bi LSTM model,models strong dependencies that are difficult to represent using CRF model,and performs word vector semantic enhancement representation of input interrogative sentences using ERNIE model,and the entity recognition capability of this model is better than the baseline model Bi LSTM.The experimental results show that the model has good recognition ability for movie entities and people entities,and the method has better performance in accuracy,recall and F-value compared with existing methods,which proves that incorporating the knowledge-enhanced semantic representation model effectively improves the entity recognition ability.Finally,an intelligent question and answer system for fused data is designed and implemented based on the above research.The user inputs the question to be queried,the system parses the question for intention recognition and entity recognition,and finally queries the knowledge base of the fused data to return the query results.The system can fuse multiple sources of heterogeneous data,establish a unified knowledge base of fused data,and use the fused database to perform semantic query work,which effectively brings into play the value of data and realizes data sharing and information exchange.