openspg

First, create the schema of the project, which defines the ontology of the business domain of the knowledge graph. For example, in our RiskMining project, its ontology includes App, Cert, Person, Company, and their relations, etc.

knext schema commit

Then, we will build the knowledge graph based on the schema. The data source of the knowledge graph is a local CSV file located in the ./builder/job/data directory. During the process of the knowledge graph construction, it may be necessary to preprocess the raw data. In such cases, we need to invoke construction operators. Therefore, before constructing the knowledge graph, we need to publish the construction operators.

knext operator publish CertLinkerOperator

Next, we can start building the knowledge graph based on the created schema and published operators. We can import entities such as App, Cert, Person, Company, along with their attributes and relations. After completing these operations, we will obtain a RiskMining knowledge graph.

knext builder submit TaxOfRiskUser,TaxOfRiskApp,Cert,Company,CompanyHasCert
knext builder submit App,Device,Person,PersonFundTrans,PersonHasDevice,PersonHoldShare

Next, we can also explore some more exciting possibilities. We define some logical rules, which define what kind of App is a gambling App and what kind of users are developers of gambling Apps". These logic rules are calculated dynamically when necessary.

knext schema reg_concept_rule --file ./schema/concept.rule

Finally, we execute GQL/KGDSL to perform queries and reasoning, inferring hidden knowledge based on existing factual data and logical rules within the knowledge graph. In the following example, we can output the phone number of each person, generate risk applications based on previously defined rules, and also output the developers and owners of these risk applications. It can be observed that this knowledge does not exist in the original factual data; rather, it is derived through inference based on the original factual data and the defined expert rules. Now, Let's run it!

knext reasoner query --file ./reasoner/gambling_app.dsl
knext reasoner query --dsl "MATCH (phone:STD.ChinaMobile)<-[:hasPhone]-(u:RiskMining.Person) RETURN u.name,phone.id"
knext reasoner query --dsl "MATCH (s:\`RiskMining.TaxOfRiskApp\`/\`赌博应用\`) RETURN s.id"
knext reasoner query --dsl "MATCH (s:\`RiskMining.TaxOfRiskUser\`/\`赌博App开发者\`) RETURN s.id,s.name"