Research On Construction Technology Of Discipline Inspection And Supervision Knowledge Base Based On Knowledge Graph

Discipline inspection and supervision are two functions exercised by the Party’s disciplinary inspection organs and governmental supervision departments,and strengthening discipline inspection and supervision is an inevitable requirement for strengthening the Party’s capacity building.With the further strengthening of the work,the traditional way of storing information gradually becomes difficult to meet the needs of staff handling information.The confidentiality and professionalism of the discipline inspection and supervision field lead to relatively complex and high threshold procedures for handling cases.In response to the problems of large quantity of notification data,low relevance,complex content and difficulty in organizing in this field,this paper proposes the technical process of constructing a knowledge map of discipline inspection and supervision,and fusing legal law information with historical cases through similarity algorithms,and finally using the entities and relationships existing within the knowledge map to construct an intelligent question and answer system.Under the requirements of adhering to the fact-oriented and scientific characterization and measurement of discipline and law enforcement,the important significance of constructing the mapping is to use the built system to discover the hidden relationships between cases and provide knowledge access and reference for actual anticorruption work based on historical cases.In response to the above requirements,firstly,this paper establishes a model layer containing multiple types of entities and relationships through a bottom-up approach according to the business requirements to build a knowledge map for discipline inspection and supervision.The data of illegal cases are collected by using the information on the website of the Discipline Inspection Commission and relevant reports.After pre-processing such as removing deactivated words and removing irrelevant information,the violation case dataset is formed by BIO annotation.Comparing the accuracy of multiple deep learning models on the named entity recognition task of the current dataset,it was determined that the entity extraction was completed using the BERT-Bi LSTM-CRF model.The obtained entities and semantic relations are composed into a triad and deposited in the Neo4 J graph database to realize the visual display of the knowledge graph.Secondly,the bidding data provided by the relevant departments of the Discipline Inspection Commission and other relevant raw data made public are used to create structured data tables after filtering and cleaning and other operations,and the schema layer is designed according to the data tables and the association between the data.Different entities are connected through relationships such as relatives and company positions,and stored in the graph database to realize the function of deducing violations using bidding knowledge mapping.Again,the data of laws and regulations in the disciplinary field are obtained,and similarity algorithms such as cosine distance algorithm and Euclidean distance algorithm based on vectorized representation of text are used to weight the summation,and the legal code is fused with the facts of disciplinary cases within the knowledge graph according to the highest matching value to meet the demand of having law to follow when referring to historical cases for qualitative measurement of discipline.Finally,using the knowledge map as the information source,the AC automaton is used to quickly match the entities in the natural interrogative sentences,and the empirical template is used to match the entities and the intent of the interrogative sentences to create inference question query statements based on the matching optimal solutions,and return the required answers from the knowledge map.The above method realizes the whole process from unstructured data to knowledge graph establishment to intelligent question and answer using the graph,and provides support and assistance to disciplinary staff in scientific decision-making through knowledge graphs combining multiple sources of information such as historical cases,and provides a technical reference for domain-based knowledge graph construction.