CN112925917A

CN112925917A - Method and device for constructing product knowledge graph, terminal and storage medium

Info

Publication number: CN112925917A
Application number: CN202110210076.7A
Authority: CN
Inventors: 邹辉
Original assignee: Ping An Life Insurance Company of China Ltd
Current assignee: Ping An Life Insurance Company of China Ltd
Priority date: 2021-02-25
Filing date: 2021-02-25
Publication date: 2021-06-08

Abstract

The application is applicable to the technical field of computers, and provides a method, a device, a terminal and a storage medium for constructing a product knowledge graph, wherein the method comprises the following steps: acquiring a plurality of risk points corresponding to sample users; acquiring a plurality of product information; and storing each product category, each product name and each risk point in a preset graph database based on the determined node attributes and storage positions to obtain a product knowledge graph. According to the mode, the multiple risk points for expressing multiple risks of the sample user are obtained, and the product knowledge graph is constructed and generated from bottom to top on the basis of the multiple risk points, the product information and the like. The method for constructing the product knowledge graph is fully considered from the perspective of the user, covers risks of various types and degrees which may occur to the user, and enables the product knowledge coverage rate in the constructed and generated product knowledge graph to be wide, the types of the risks which may occur to the user to be complete and the risk information to be rich.

Description

Method and device for constructing product knowledge graph, terminal and storage medium

Technical Field

The application belongs to the technical field of computers, and particularly relates to a method, a device, a terminal and a storage medium for constructing a product knowledge graph.

Background

In the traditional product knowledge graph construction, the product knowledge graph is constructed from top to bottom based on different product information. For example, when building an insurance knowledge graph, it is common to build from top to bottom starting from an insurance product based on different insurance information. The inventor realizes that the insurance knowledge coverage rate in the insurance knowledge map obtained by the construction method is low, the risk types are not comprehensive, the risk information is thin, and the risk cannot be considered from the perspective of users well.

Disclosure of Invention

In view of this, embodiments of the present application provide a method, an apparatus, a terminal and a storage medium for constructing a product knowledge graph, so as to solve the problems of low product knowledge coverage, incomplete risk types and thin risk information in a product knowledge graph constructed by a conventional method.

A first aspect of an embodiment of the present application provides a method for constructing a product knowledge graph, including:

acquiring a plurality of risk points corresponding to a sample user, wherein the risk points are used for representing a plurality of risks corresponding to the sample user, and one risk point corresponds to one risk;

the method comprises the steps of obtaining a plurality of product information, wherein each product information comprises a product type and a plurality of product names corresponding to the product type;

determining a plurality of product categories, a plurality of product names and a top-bottom relationship among the plurality of risk points, and determining node attributes and storage positions based on the top-bottom relationship, wherein the node attributes comprise a node attribute corresponding to each product category, a node attribute corresponding to each product name and a node attribute corresponding to each risk point, and the storage positions comprise a storage position corresponding to each product category, a storage position corresponding to each product name and a storage position corresponding to each risk point;

and storing each product category, each product name and each risk point in a preset graph database based on the node attributes and the storage positions to obtain a product knowledge graph.

A second aspect of an embodiment of the present application provides an apparatus for constructing a product knowledge graph, including:

the system comprises a first obtaining unit, a second obtaining unit and a third obtaining unit, wherein the first obtaining unit is used for obtaining a plurality of risk points corresponding to a sample user, the plurality of risk points are used for representing a plurality of risks corresponding to the sample user, and one risk point corresponds to one risk;

the second acquisition unit is used for acquiring a plurality of product information, and each product information comprises a product category and a plurality of product names corresponding to the product category;

the determining unit is used for determining a plurality of product categories, a plurality of product names and a top-bottom relationship among the plurality of risk points, and determining node attributes and storage positions based on the top-bottom relationship, wherein the node attributes comprise a node attribute corresponding to each product category, a node attribute corresponding to each product name and a node attribute corresponding to each risk point, and the storage positions comprise a storage position corresponding to each product category, a storage position corresponding to each product name and a storage position corresponding to each risk point;

and the storage unit is used for storing each product category, each product name and each risk point in a preset graph database based on the node attributes and the storage positions to obtain a product knowledge graph.

A third aspect of the embodiments of the present application provides a terminal for building a product knowledge graph, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the steps of the method according to the first aspect.

A fourth aspect of embodiments of the present application provides a computer-readable storage medium, having stored thereon a computer program which, when executed by a processor, performs the steps of the method according to the first aspect as described above.

A fifth aspect of embodiments of the present application provides a computer program product, which, when run on a terminal, causes the terminal to perform the steps of the method according to the first aspect.

The construction method, the construction device, the construction terminal and the storage medium of the product knowledge graph provided by the embodiment of the application have the following beneficial effects:

the construction terminal acquires a plurality of risk points for representing various risks of the sample user, and constructs and generates a product knowledge graph from bottom to top based on the plurality of risk points, product categories corresponding to the product information, product names and the like. The method for constructing the product knowledge graph is fully considered from the perspective of the user, covers risks of various types and degrees which may occur to the user, and enables the product knowledge coverage rate in the constructed and generated product knowledge graph to be wide, the types of the risks which may occur to the user to be complete and the risk information to be rich.

Drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the embodiments or the prior art descriptions will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without inventive exercise.

FIG. 1 is a schematic flow chart diagram of a method for building a product knowledge graph according to an embodiment of the present application;

FIG. 2 is a schematic illustration of a partial insurance knowledgegraph provided in accordance with an embodiment of the present application;

FIG. 3 is a schematic flow chart diagram of a method for building a product knowledge graph according to yet another embodiment of the present application;

FIG. 4 is a schematic flow chart diagram of a method for building a product knowledge graph according to yet another embodiment of the present application;

FIG. 5 is a schematic diagram of an apparatus for building a product knowledge-graph according to an embodiment of the present application;

FIG. 6 is a schematic diagram of a terminal for building a product knowledge-graph according to another embodiment of the present application.

Detailed Description

In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.

With the development of science and technology, knowledge maps are widely applied to various industries. For example, the traditional insurance industry is a knowledge-intensive industry, and a great deal of cross-industry knowledge is involved in each business link of pre-sale, in-sale and after-sale of insurance, including insurance products, disease health knowledge, risk and accident knowledge, current economic status of customers, life status of customers and the like. The insurance knowledge map is established, various industry knowledge can be associated, and the insurance knowledge map is applied to scenes such as a question-answering system, an insurance product recommendation system and the like, so that insurance-related problems can be solved for users, and insurance products can be recommended for the users.

In the traditional construction of the insurance knowledge graph, different insurance information is generally constructed from top to bottom starting from insurance products. The inventor realizes that the insurance knowledge coverage rate in the insurance knowledge map obtained by the construction method is low, the risk types are not comprehensive, the risk information is thin, and the risk cannot be considered from the perspective of users well. Further, when the insurance knowledge map is applied to a question-answering system, problems related to insurance cannot be accurately answered for the user, when the insurance knowledge map is applied to an insurance product recommendation system, actual insurance application requirements of the user cannot be grasped, the insurance products recommended to the user are inaccurate, and the sales conversion rate and the sales volume of the insurance products are affected.

In view of this, the present application provides a method for constructing a product knowledge graph, in which a construction terminal acquires a plurality of risk points for representing a plurality of risks of a sample user, and a product knowledge graph is constructed and generated from bottom to top based on the plurality of risk points, product categories corresponding to product information, product names, and the like. The method for constructing the product knowledge graph is fully considered from the perspective of the user, covers risks of various types and degrees which may occur to the user, and enables the product knowledge coverage rate in the constructed and generated product knowledge graph to be wide, the types of the risks which may occur to the user to be complete and the risk information to be rich. Furthermore, if the product knowledge graph is applied to a question answering system, relevant questions can be accurately answered for users, and the question answering accuracy of the question answering system is improved; if the product knowledge graph is applied to a product recommendation system, the actual requirements of the user can be grasped, the most appropriate product can be recommended to the user, and the sales conversion rate and the product sales volume are improved.

Referring to fig. 1, fig. 1 is a schematic flow chart of a method for constructing a product knowledge graph according to an embodiment of the present application. In this embodiment, an execution main body of the method for constructing the product knowledge graph is a construction terminal, a server, and the like, where the construction terminal includes, but is not limited to, a smart phone, a tablet computer, a desktop computer, a Personal Digital Assistant (PDA), a notebook computer, an ultra-mobile Personal computer (UMPC), a netbook, and the like, and the server includes, but is not limited to, an independent server, a distributed server, a server cluster, a cloud server, and the like. The method shown in fig. 1 may include S101 to S104, and the specific implementation principle of each step is as follows.

S101: and acquiring a plurality of risk points corresponding to the sample user, wherein the plurality of risk points are used for representing a plurality of risks corresponding to the sample user, and one risk point corresponds to one risk.

The risk points are used for representing risks corresponding to the sample users, and it can be understood that the risk points are various types and different degrees of risks that may occur to the sample users from the perspective of the sample users themselves. One risk point corresponds to one risk, and a plurality of risk points represent a plurality of risks corresponding to the sample user.

Illustratively, the risk points in this example may be different types of risks that may occur to the sample user, e.g., the risk points may be various diseases that may occur to the sample user (hyperthyroidism, hypertension, coronary heart disease, tumors, cerebral hemorrhage, myocardial infarction, depression, congenital hearing loss, congenital blindness, etc.). The risk points may also be various accidents that may occur to the sample user (loss of business, car accidents, drowning, accidental deafness, accidental blindness, revenue outages, sudden economic consumption, etc.). The risk points may also be the causes of the sample user to develop various diseases (age size, familial genetic disease, unhealthy lifestyle, etc.). The description is given for illustrative purposes only and is not intended to be limiting.

There may be multiple ways to obtain multiple risk points corresponding to a sample user. For example, a plurality of different types of risk information are preset by a sample user, the different types of risk information preset by the sample user are collected, and corresponding risk points are generated according to the risk information. Or extracting risk points from information such as conversation logs, comment information, insurance websites, network magazines and the like. Or, on the basis of risk points generated based on a plurality of different types of risk information preset by a sample user, the risk points are randomly combined, and a proper result is selected from the results of random combination as the risk points, so that the types of the risk points are further enriched.

S102: the method comprises the steps of obtaining a plurality of product information, wherein each product information comprises a product type and a plurality of product names corresponding to the product type.

A plurality of product information can be acquired from a database, each service terminal and each large website by a construction terminal or a server. The product information may include insurance information, insurance product information, business information, and the like, and each product information includes a product category corresponding to the product information and a plurality of product names corresponding to the product category. In this embodiment, product information is taken as insurance information as an example for explanation, and accordingly, the product type is an insurance type and the product name is an insurance product. The description is given for illustrative purposes only and is not intended to be limiting.

Illustratively, each insurance information includes an insurance category (insurance type category) corresponding to the insurance information, and a plurality of insurance products corresponding to the insurance category. For example, the insurance information may be information about property insurance, information about life insurance, and the like.

When the insurance information is property insurance information, the insurance category corresponding to the property insurance information is property insurance; the insurance category of property insurance corresponds to a plurality of insurance products, such as property loss insurance, liability insurance, credit insurance, and the like. Wherein, the property loss insurance refers to various insurance products with tangible property as insurance target, such as enterprise property insurance, family property insurance, transportation tool insurance, cargo transportation insurance, engineering insurance, special risk insurance, agricultural insurance, etc.; the responsibility insurance refers to an insurance product with insurance target for the indemnity responsibility which the insured person should pay for property loss or personal injury of a third party, such as public responsibility insurance, product responsibility insurance, employer responsibility insurance, occupational responsibility insurance and the like. The credit insurance refers to insurance with various credit behaviors as insurance targets, such as commercial credit insurance, export credit insurance, contract guarantee insurance, product guarantee insurance, loyalty guarantee insurance and the like. The description is given for illustrative purposes only and is not intended to be limiting.

When the insurance information is personal insurance information, the insurance type corresponding to the personal insurance information is personal insurance; the insurance category of human insurance corresponds to a plurality of insurance products, such as life insurance, accident insurance, health insurance and the like. Wherein the life insurance may include death insurance, dual insurance, annuity insurance, etc. Unexpected insurance may include unexpected personal insurance, unexpected disability insurance, unexpected subsidy insurance, and the like. Health insurance may include critical illness, medical insurance, and the like.

The insurance information can be acquired by a construction terminal or a server from a database, each service terminal and each large website. The insurance-related data that begins to be collected may include structured data and unstructured data. The structured data refers to data of acquired information which can be directly stored in an insurance knowledge graph, and the unstructured data cannot be directly stored in the insurance knowledge graph. For example, the unstructured data may be spoken data, a large segment of text that is not divided, a picture containing text, foreign language, and so forth. When the collected data is unstructured data, the unstructured data can be converted into structured data. For example, if the data is recognized as a foreign language in the Chinese language, the data can be translated to obtain Chinese; recognizing that the data is spoken, performing word segmentation, word stop and other processing, and converting the processed data into written expressions; if the identified data is a large segment of characters which are not divided, analyzing the segment of characters and dividing; the identified data is a picture containing characters, and the characters in the picture are identified; and different processing is carried out on different unstructured data, and finally a plurality of insurance information are obtained. The description is given for illustrative purposes only and is not intended to be limiting.

S103: determining the upper and lower relations among a plurality of product categories, a plurality of product names and a plurality of risk points, and determining node attributes and storage positions based on the upper and lower relations, wherein the node attributes comprise node attributes corresponding to each product category, node attributes corresponding to each product name and node attributes corresponding to each risk point, and the storage positions comprise storage positions corresponding to each product category, storage positions corresponding to each product name and storage positions corresponding to each risk point.

The node attributes may include a node attribute corresponding to each product category, a node attribute corresponding to each product name, and a node attribute and a root node corresponding to each risk point. The node attribute corresponding to each product category may include a child node, the node attribute corresponding to each product name may include a child node, and the node attribute corresponding to each risk point may include a leaf node (end node).

The storage positions comprise a storage position corresponding to each product category, a storage position corresponding to each product name, a storage position corresponding to each risk point, and a storage position corresponding to data corresponding to the root node.

The root node is a node attribute corresponding to the most central data (core information of the product knowledge graph) of the product knowledge graph. It can be understood that the product knowledge graph is a knowledge graph constructed by taking a product as a core, and then the product can be used as the most central data of the insurance knowledge graph. Namely, the node attribute corresponding to the product is the root node. The data corresponding to the root node may be preset by a user according to a knowledge graph that the user wants to construct, or may be obtained by analyzing a plurality of pieces of product information obtained, for example, extracting a keyword from the obtained plurality of pieces of product information to obtain a keyword "product", and using the "product" as the data corresponding to the root node.

Illustratively, the product knowledge graph may include an insurance knowledge graph, an insurance product knowledge graph, and the like. When the product knowledge graph is the insurance knowledge graph, the root node is the node attribute corresponding to the most central data (core information of the insurance knowledge graph) of the insurance knowledge graph. It can be understood that the insurance knowledge graph is a knowledge graph constructed by taking insurance as a core, and then insurance and insurance products can be used as the most central data of the insurance knowledge graph. Namely, the node attributes corresponding to the insurance product and the insurance product are root nodes. The data corresponding to the root node may be preset by the user according to a knowledge graph that the user wants to construct, or may be obtained by analyzing the obtained plurality of insurance information, for example, extracting a keyword from the obtained plurality of insurance information to obtain an "insurance", and using the "insurance" as the data corresponding to the root node.

It should be noted that "insurance" or "insurance product" is merely a literal meaning, and is understood to be used to represent which type the constructed knowledge graph belongs to, and is not used to refer to a specific insurance or insurance product. Only one root node exists in an insurance knowledge graph, and in the embodiment, data corresponding to the root node can be insurance or insurance products. The description is given for illustrative purposes only and is not intended to be limiting.

In this embodiment, product information is taken as insurance information for example, accordingly, the product category is insurance category, the product name is insurance product, and the product knowledge map obtained by construction is insurance knowledge map. The description is given for illustrative purposes only and is not intended to be limiting.

An up-down relationship between a plurality of product categories, a plurality of product names, and a plurality of risk points is determined. Specifically, the superior-inferior relation among the insurance categories corresponding to all insurance information, all insurance products and all risk points is determined. The superior-inferior relation between the information can be determined according to the affiliated relation and the inclusion relation between the information. For example, if a includes B, then a is the upper position and B is the lower position; a includes B, B includes C, C is the lower position of B, B is the lower position of A, A is the upper position.

Illustratively, the insurance information is property insurance information and life insurance information, the insurance category corresponding to the property insurance information is property insurance, and the insurance category corresponding to the life insurance information is life insurance. The multiple insurance products corresponding to the property insurance are family property insurance, product responsibility insurance and employer responsibility insurance. The plurality of insurance products corresponding to the personal insurance are death insurance, accidental disability insurance and medical insurance. The risk points include household appliance damage, sofa damage, explosion in the use process of tires, toy allergy, injury caused by occupational diseases, disability caused by occupational diseases, normal death, accidental deafness, accidental broken legs, hyperthyroidism and coronary heart disease. The property insurance is the upper data of the home property insurance, the product responsibility insurance and the employer responsibility insurance, the home property insurance is the upper data of the damage of the household appliance and the damage of the sofa, and correspondingly, the damage of the household appliance and the damage of the sofa are the lower data of the home property insurance. The product responsibility insurance is upper data of explosion and toy allergy in the tire use process, and correspondingly, the explosion and toy allergy in the tire use process are lower data of the product responsibility insurance. And similarly, determining the upper and lower bit relations among the rest information.

Node attributes are determined based on the context. Specifically, all risk points belong to lower data, and therefore, node attributes corresponding to all risk points are leaf nodes (end nodes). The insurance classes all belong to upper data, and belong to upper relations in the relation with insurance products, so the node attributes corresponding to all the insurance classes are child nodes. The insurance products belong to lower data relative to insurance types and upper data relative to risk points, and because leaf nodes (end nodes) in the graph database can only be of one type, the node attributes corresponding to all the insurance products are child nodes.

Based on the context and node attributes, a storage location is determined. Specifically, according to a plurality of insurance categories corresponding to each insurance information, a plurality of insurance products, and a plurality of risk points, and node attributes corresponding to each insurance category, node attributes corresponding to each insurance product, and node attributes corresponding to each risk point, a corresponding storage location of the insurance category corresponding to each insurance information in a preset graphic database is determined, a corresponding storage location of each insurance product in the preset graphic database is determined, and a corresponding storage location of each risk point in the preset graphic database is determined.

Illustratively, the preset graph database is provided with storage locations corresponding to a root node, a child node and a leaf node. The approximate storage position of the information in a preset graph database can be preliminarily determined according to the node attribute, and the accurate storage position of the information is further determined according to the upper and lower relation. For example, the node attributes corresponding to all the risk points are leaf nodes, and the storage positions of the risk points in a preset graph database are preliminarily determined to be the storage positions corresponding to the leaf nodes on the bottom layer. And further analyzing the upper data corresponding to each risk point, wherein the upper data of the damaged household appliance and the damaged sofa is the home property insurance. The storage positions of the two risk points of household appliance damage and sofa damage correspond to leaf nodes below the child node of the family property insurance in the graph database. And determining the storage positions corresponding to the insurance category, the insurance product and the risk point corresponding to each insurance information in the same way. And storing the information to a corresponding position in a preset graph database to obtain the constructed insurance knowledge map.

The description is given for illustrative purposes only and is not intended to be limiting.

S104: and storing each product category, each product name and each risk point in a preset graph database based on the node attributes and the storage positions to obtain a product knowledge graph.

The preset graphic database is a prototype of the insurance knowledge map and can be understood as a framework of the insurance knowledge map, and insurance-related data can be stored in the preset graphic database in advance. For example, the data corresponding to the root node may be stored in a preset graph database in advance, that is, the graph database may be stored with "insurance" or "insurance product" in advance. And then based on the node attribute corresponding to each insurance type, the node attribute corresponding to each insurance product, the node attribute corresponding to each risk point, the storage position of the insurance type corresponding to each insurance information in a preset graphic database, the storage position of each insurance product in the preset graphic database, and the storage position of each risk point in the preset graphic database, the insurance type corresponding to each insurance information, each insurance product and each risk point are stored in the graphic database, and an insurance knowledge graph is obtained.

The data corresponding to the root node, the insurance category corresponding to each insurance information, each insurance product and each risk point are stored in the graph database based on the data corresponding to the root node, the node attribute corresponding to each insurance category, the node attribute corresponding to each insurance product, the storage position corresponding to the insurance category corresponding to each risk point in the preset graph database, the storage position corresponding to each insurance product in the preset graph database, the storage position corresponding to each risk point in the preset graph database and the storage position corresponding to the data corresponding to the root node, so that the insurance knowledge graph is obtained.

For example, in the present application, an insurance knowledge graph is constructed based on risk points corresponding to sample users, and the information at the bottom layer is also the risk points, so that the node attribute corresponding to each risk point is a leaf node (end node). Different storage positions corresponding to different node attributes are preset in the graph database. For example, the leaf node is stored in the graph database at the bottom of the graph database. The location of the root node in the graph database is the most central of the graph database. Searching the position where the leaf node should be stored in a preset graph database to obtain the storage position corresponding to each risk point, which is only an exemplary illustration here, and is not limited to this.

FIG. 2 is a schematic diagram of a partial insurance knowledgegraph provided in accordance with an embodiment of the present application. It is to be understood that the schematic diagram is merely provided to facilitate understanding, and a small portion of the contents of the insurance knowledgegraph is shown, and does not represent the entire insurance knowledgegraph as constructed.

As shown in fig. 2, the insurance product/insurance is data corresponding to a root node in the insurance knowledge graph, and the storage location of the insurance product/insurance product is also a location corresponding to the root node in the insurance knowledge graph. In the process of constructing the insurance knowledge graph, the insurance product/insurance is data corresponding to the root node in the preset graph database, and the storage position of the insurance product/insurance is also the position corresponding to the root node in the preset graph database. And storing the insurance products/insurance products into a preset graphic database according to the position corresponding to the root node. The node attributes corresponding to the property risk and the person risk are all child nodes and are all lower nodes of the root node, namely are all lower data of insurance products/insurance. Therefore, the storage positions of the property risk and the personal risk are the positions corresponding to the child nodes next to the root node in the preset graph database. And storing the property risk and the person risk to the position corresponding to the child node next to the root node in a preset graph database.

The family property insurance, the product responsibility insurance and the employer responsibility insurance are specific insurance products, and the node attributes corresponding to the insurance products are child nodes, namely the node attributes are all lower data of property insurance. Therefore, the storage positions corresponding to the family property insurance, the product responsibility insurance and the employer responsibility insurance are the positions corresponding to the child nodes below the property insurance in the preset graphic database, and the family property insurance, the product responsibility insurance and the employer responsibility insurance are stored in parallel to the positions corresponding to the three child nodes below the property insurance in the preset graphic database. The data corresponding to the node attributes belonging to the same level are stored in a non-limited order. For example, the storage order of the family property insurance, the product liability insurance and the employer liability insurance in the graphic database is not limited, and the family property insurance, the product liability insurance and the employer liability insurance may be stored from left to right, or the product liability insurance, the employer liability insurance and the family property insurance may be stored. The description is given for illustrative purposes only and is not intended to be limiting.

The household appliance damage and the sofa damage are risk points corresponding to the user and belong to the content of specific insurance of the family property insurance, and the node attributes corresponding to the risk points are leaf nodes and specifically belong to lower data of an insurance product, namely the family property insurance. Therefore, the storage positions corresponding to the damage of the household appliance and the damage of the sofa respectively are positions corresponding to leaf nodes under the home property insurance in the preset graphic database. And storing the damaged household appliance and the damaged sofa side by side to the positions corresponding to the leaf nodes below the home property insurance in the graph database.

Explosion and toy allergy caused in the use process of the tire are risk points corresponding to users, belong to the content of specific guarantee of product liability insurance, and the node attributes corresponding to the risk points are leaf nodes, and particularly belong to lower data of a product liability insurance product. Therefore, the storage positions corresponding to the explosion and the toy allergy in the use process of the tire respectively are the positions corresponding to the leaf nodes under the product responsibility insurance in the preset graphic database. And (4) storing the explosion and toy allergy in the use process of the tire to the position corresponding to the leaf node below the product responsibility insurance in the graphic database.

Occupational disease injury and occupational disease disability are risk points corresponding to users, belong to the content of employer responsibility insurance specific guarantee, and the node attributes corresponding to the risk points are leaf nodes, and particularly belong to lower data of an insurance product, namely employer responsibility insurance. Therefore, the storage locations corresponding to the injury and disability caused by the occupational disease respectively are the locations corresponding to the leaf nodes below the employer responsibility insurance in the preset graphic database.

Similarly, death insurance, accident disability insurance and medical insurance are specific insurance products, and the node attributes corresponding to the insurance products are all sub-nodes, namely all lower data of personal insurance. Therefore, the storage positions corresponding to death insurance, accident disability insurance and medical insurance are the positions corresponding to the sub-nodes below the personal insurance in the preset graphic database. And storing the death insurance, the accident disability insurance and the medical insurance in the graphic database in parallel at the position corresponding to the sub-node below the human body insurance.

The normal death is a risk point corresponding to the user and belongs to the content of the death insurance specific guarantee, and the node attribute corresponding to the risk point is a leaf node and specifically belongs to the lower data of the death insurance product. Therefore, the storage location corresponding to normal death is the location corresponding to the leaf node under the death insurance in the preset graphic database. Normal deaths are stored in the graphical database at locations corresponding to leaf nodes below the death insurance.

Unexpected deafness and accidental broken legs are risk points corresponding to users, belong to the content of specific insurance of the unexpected disability insurance, and the node attributes corresponding to the risk points are leaf nodes and specifically belong to lower data of the insurance product, namely the unexpected disability insurance. Therefore, the storage positions corresponding to the accidental deafness and the accidental broken leg are the positions corresponding to the leaf nodes below the accidental disability insurance in the preset graphic database. And storing the accidental deafness and the accidental broken legs side by side to the positions corresponding to the leaf nodes below the accidental disability insurance in the graph database.

Hyperthyroidism and coronary heart disease are risk points corresponding to a user, belong to the content of specific guarantee of medical insurance, and the node attributes corresponding to the risk points are leaf nodes, and specifically belong to lower data of an insurance product, namely medical insurance. Therefore, the storage positions corresponding to the hyperthyroidism and the coronary heart disease respectively are the positions corresponding to the leaf nodes below the medical insurance in the preset graphic database. And storing the hyperthyroidism and the coronary heart disease in the graphic database in parallel at the position corresponding to the leaf node below the Chinese medical insurance.

It is to be understood that the description is illustrative and not restrictive.

And storing the data corresponding to the root node, the insurance category corresponding to each insurance information, each insurance product and each risk point into a preset graphic database according to the mode to obtain the constructed insurance knowledge graph. It can be understood that the insurance knowledge graph can be updated at any time, that is, if a new risk point and new insurance information are obtained (the new insurance information includes an insurance category corresponding to the new insurance information and an insurance product corresponding to the insurance category), the new risk point, the insurance category corresponding to the new insurance information, node attributes corresponding to the insurance product corresponding to the insurance category and storage locations corresponding to the node attributes and the storage locations are determined by the method in S103, and the new risk point and the new insurance information are stored in the established insurance knowledge graph according to the newly determined node attributes and storage locations, so as to update the insurance knowledge graph.

In this embodiment, the construction terminal acquires a plurality of risk points for representing a plurality of risks of the sample user, and constructs and generates an insurance knowledge graph from bottom to top based on the plurality of risk points, insurance categories corresponding to insurance information, insurance products, and the like. The method for constructing the insurance knowledge map is fully considered from the perspective of the user, covers various types and degrees of risks which may occur to the user, and ensures that the coverage rate of insurance knowledge in the constructed and generated insurance knowledge map is wide, the types of risks which may occur to the user are complete, and the information is rich. Furthermore, if the insurance knowledge map is applied to the question answering system, the problem related to insurance can be accurately answered for the user, and the problem answering accuracy of the question answering system is improved; if the insurance knowledge map is applied to an insurance product recommendation system, the actual insurance application requirements of the user can be grasped, the most appropriate insurance product is recommended to the user, and the sales conversion rate and the sales volume of the insurance product are improved.

As shown in fig. 3, fig. 3 is a schematic flowchart of a method for constructing a product knowledge graph according to another embodiment of the present application, and optionally, in a possible implementation manner, the above S101 may include S1011 to S1013, specifically as follows:

s1011: the method comprises the steps of obtaining data to be analyzed, wherein the data to be analyzed comprises a conversation log and comment information, the conversation log comprises a conversation generated when a sample user consults insurance, and the comment information is information of the sample user comment insurance.

The data to be analyzed may include conversation logs, comment information, web page content in insurance websites, information in web magazines, and the like. Different data to be analyzed have different acquisition modes, and the acquisition modes are not limited by taking the actual situation as the standard.

The session log includes the session generated when the sample user consults the insurance, and it is understood that the session log is generated when the sample user dialogues with the insurance staff. For example, when a sample user consults insurance-related information (insurance products, insurance types, content of insurance guarantees, and the like) through a webpage, a mobile phone software (APP), an applet, and the like, chat information is generated, and the construction terminal integrates the chat records to obtain a session log.

The review information consists of information of the sample user review insurance. For example, a sample user can comment on some insurance products in a webpage, an APP and an applet, and the terminal integrates the comments of the sample user to obtain comment information.

For the webpage content in the insurance website and the information in the network magazine, the webpage content related to insurance can be extracted from the insurance website, and the information related to insurance can be collected in the network magazine.

It can be understood that the data to be analyzed may also be generated by other devices in advance, and the present construction terminal obtains the data to be analyzed sent by the other devices. And the user can upload the data to be analyzed to the construction terminal. The description is given for illustrative purposes only and is not intended to be limiting.

S1012: and extracting various risk information of the sample user from the data to be analyzed, and generating a plurality of candidate risk points based on the various risk information of the sample user.

And extracting various risk information which may occur to a sample user from the data to be analyzed, and generating a plurality of candidate risk points based on the various risk information which may occur to the sample user. Because some of the multiple risk information which may possibly occur to the sample user is extracted from the data to be analyzed, and some of the multiple risk information may be used as risk points corresponding to the sample user, and some of the multiple risk information may not be used as risk points corresponding to the sample user, multiple candidate risk points are generated based on the multiple risk information which may possibly occur to the sample user. And screening the candidate risk points to obtain the risk points corresponding to the sample users.

Specifically analyzing the data to be analyzed, namely extracting information such as risk information possibly generated by a sample user, risk factors possibly generated by the sample user, risk already generated by the sample user and the like in the data to be analyzed, and generating candidate risk points one by one according to the information.

Illustratively, when the data to be analyzed is a conversation log, analyzing specific chat information in the conversation log, and performing word segmentation processing on the chat information to obtain a plurality of words. And extracting information such as risk information which may occur to the sample user, risk factors which may occur to the sample user, risk which has occurred to the sample user and the like from the multiple participles, and generating candidate risk points one by one according to the information.

For example, the chat message is "i carelessly dropped the leg while going out to play", and the chat message is subjected to word segmentation processing to obtain "i/play/carelessly/dropped the leg". Through artificial judgment, if the 'broken leg' can be taken as a risk point, the information can be extracted and taken as a candidate risk point, namely the candidate risk point is the 'broken leg'.

For another example, the chat information is that when i sit at a bus and travel, a car accident breaks legs, the car, the bus, the travel, the car accident and the leg break are all factors that a sample user may have risks, and the information is extracted respectively to serve as candidate risk points, namely the candidate risk points are the car, the travel, the car accident and the leg break.

For another example, the chat information is that "the user travels out of the plane during the celebration in our country," the "plane" and "the" travel "are all possible risk factors of the sample user, and the information is extracted respectively to serve as candidate risk points, that is, the candidate risk points are" the plane "and" the "travel". The description is given for illustrative purposes only and is not intended to be limiting.

S1013: and determining a plurality of risk points corresponding to the sample user in the plurality of candidate risk points.

And screening the candidate risk points, wherein the screening result is used as a plurality of risk points corresponding to the sample user. The method comprises the steps of firstly carrying out duplicate removal processing on a plurality of candidate risk points, filtering out the same candidate risk points in the plurality of candidate risk points, and then filtering out the candidate risk points which cannot serve as the risk points corresponding to the final sample user in the plurality of candidate risk points after the duplicate removal processing, so as to obtain a plurality of risk points corresponding to the sample user.

Continuing with the example in S1012, the candidate risk points obtained in S1012 above are "broken leg", "bus", "travel", "car accident", "broken leg", "airplane", "travel". And carrying out duplication removal treatment on the candidate risk points to obtain the candidate risk points including 'broken legs', 'big bus', 'tourism', 'car accident' and 'plane'. And manually screening the candidate risk points obtained after the duplicate removal treatment to obtain a plurality of risk points corresponding to the sample user. For example, "broken leg" and "car accident" among these candidate risk points are taken as the risk points corresponding to the sample user.

In the embodiment, various types and degrees of risk information which may occur to a sample user is mined in data to be analyzed, risk points are generated based on the mined risk information, so that the types of the risk points are complete, the risk points are close to the user, and the risk coverage rate is wide, so that the insurance knowledge coverage rate in an insurance knowledge graph which is constructed and generated based on the risk points is wide, the types of risks which may occur to the user are complete, and the risk information is rich.

Optionally, in a possible implementation manner, S1014 to S1017 may be further included after S1012, specifically as follows:

s1014: and randomly combining the candidate risk points to obtain a plurality of combination results.

And randomly combining the candidate risk points, wherein the number of the combinations is not limited, and the combination can be performed by combining two candidate risk points or combining a plurality of candidate risk points, so that a plurality of combination results are obtained.

For example, the candidate risk points include "broken leg", "bus", "travel", "car accident", "airplane", and a plurality of combination results such as "broken leg by bus", "broken leg by bus accident", "broken leg by travel", "travel by airplane travel", "broken leg by airplane" can be obtained by randomly combining the candidate risk points. It will be appreciated that only individual combined results are illustrated here, and that many more are actually combined.

S1015: and processing each combined result through a preset language scoring model to obtain a score corresponding to each combined result.

The preset language scoring model is a trained model acquired in the network, and is used for evaluating the reasonability of the input statement and the fact expressed by the statement, and the training process and the specific processing process of the data can refer to the prior art and are only briefly described here.

Illustratively, each combined result is input into a language scoring model for scoring, and the language scoring model outputs a score corresponding to each combined result. For example, "leg was broken in travel", "leg was broken in airplane travel" was input into the language scoring model for scoring, the score corresponding to "leg was broken in travel" was output by the language scoring model to be 90, the "leg was broken in travel" was obviously not in accordance with the normal sentence order, the score corresponding to "leg was broken in travel" was output by the language scoring model to be 35 at this time, and the score corresponding to "leg was broken in airplane travel" was output by the language scoring model to be 85. The description is given for illustrative purposes only and is not intended to be limiting.

S1016: and acquiring a target combination result with the score larger than or equal to a preset threshold value.

The preset threshold is preset by a sample user and is used for judging whether the combined result can generate a corresponding risk point.

Comparing the score corresponding to each combined result with a preset threshold, and recording the combined result as a target combined result when the score is greater than or equal to the preset threshold; and when the score is smaller than the preset threshold value, the combined result is not processed or deleted.

For example, the preset threshold is 80. The score corresponding to the 'leg broken by tourism' is 90, the score is larger than a preset threshold value, and the combined result of 'leg broken by tourism' is marked as a target combined result. For another example, the score corresponding to "leg-broken travel" is 35, the score is smaller than the preset threshold, and the combined result of "leg-broken travel" is not processed or is deleted. The description is given for illustrative purposes only and is not intended to be limiting.

S1017: and generating a risk point corresponding to the sample user based on the target combination result.

And generating a plurality of risk points corresponding to the sample user based on all the target combination results. Wherein one target combination result generates one risk point.

For example, the preset threshold is 80. The score corresponding to the 'leg broken in travel' is 90, the score is larger than a preset threshold value, and the combined result of 'leg broken in travel' is a target combined result. A risk point corresponding to the sample user can be generated according to the 'leg broken during travel'. It is to be noted that, when generating the risk points, the concrete expression of the target combination result may be appropriately retouched. For example, the risk point corresponding to "the leg has been broken during travel" may be "the leg has been broken carelessly during travel", "the leg has been broken during travel", or the like.

The score corresponding to the 'leg breakage travel' is 35, the score is smaller than a preset threshold value, and at the moment, a risk point cannot be generated according to the 'leg breakage travel'.

The score corresponding to the condition that the legs are broken in the airplane tourism is 85, the score is larger than a preset threshold value, and the combined result that the legs are broken in the airplane tourism is the target combined result. A risk point corresponding to the sample user can be generated according to the fact that the legs are broken when the airplane travels. The method has the advantages that the 'legs are broken when the airplane travels' for color matching, the obtained risk points can be 'legs are broken when the airplane travels, the' legs are broken when the airplane travels and the airplane crashes 'when the airplane travels'. The description is given for illustrative purposes only and is not intended to be limiting.

In this embodiment, for the random combination of candidate risk points, a suitable risk is further screened as a risk point in the result of the random combination. The types of the risk points are further expanded, and the coverage rate of the risk points is wider. For example, the risk points obtained in S1013 are "broken legs" and "car accident", and by the implementation in the present embodiment, a risk point "broken legs are carelessly when traveling on an airplane" is also obtained.

Optionally, in a possible implementation manner, in order to make the generated insurance knowledgebase map more concise and facilitate finding answers in the actual application process, multiple generated risk points may be integrated, that is, risk points expressing the same concept and the same risk are merged into one risk point, it is worth explaining that, although merged into one risk point, when storing into the graph database, the risk points expressing the same concept and the same risk may all be stored in the storage location corresponding to the merged risk point.

For example, the acquired risk points include hyperthyroidism, and goiter, and the hyperthyroidism and the goiter are different names of hyperthyroidism, and the expression of the hyperthyroidism and the goiter is the same risk, so that the hyperthyroidism, the hyperthyroidism and the goiter can be combined into the same risk point. Hyperthyroidism is taken as a representative risk point after combination. It should be noted that, although hyperthyroidism, and goiter gas are taken as the same risk point, when storing in the graph database, hyperthyroidism and goiter gas are also stored in the storage location corresponding to hyperthyroidism. For example, it can be stored as hyperthyroidism/goiter. The thumbnail can also be embodied in the form of a thumbnail, namely, the information which can be visually seen in the leaf node is hyperthyroidism, the thumbnail corresponding to the leaf node can be expanded by clicking the leaf node, and the thumbnail shows hyperthyroidism and goiter. The description is given for illustrative purposes only and is not intended to be limiting.

Optionally, in a possible implementation manner, the obtaining of the plurality of risk points corresponding to the sample user may further include: acquiring different types of risk information preset by a sample user; and generating corresponding risk points of the sample users based on the different types of risk information.

Various types of risk information which can be generated by the conceivable user are recorded in the text, and different types of risk information preset by the sample user are obtained. And acquiring different types of risk information preset by the sample user and recorded in the text, and generating a plurality of risk points corresponding to the sample user according to the different types of risk information. Wherein one risk generates a corresponding one of the risk points.

Illustratively, different types of risk information preset by the sample user can be used for toys with unreasonable design to cause children to eat by mistake, causing operation or injury, having toxic components, household appliances damaged, sofa damaged, medicine-caused sequelae, food poisoning, accidental deafness, accidental blindness, accidental drowning, traffic accidents in travel, airplane wreck, being driven away, losing work, being ill by family people to cause urgent need for money, doing operation, infecting wind cold, coronary heart disease, lumbar vertebra strain, hypertension, hyperlipidemia, diseases caused by over-age, hereditary diseases, diseases caused by over-age, infertility, lung diseases caused by smoking, depression, natural disasters and the like. These are all from a sample user perspective, taking into account the various types of risks that a sample user may incur. According to the different types of risk information, a plurality of different risk points can be correspondingly generated.

In this embodiment, in order to make the risk points closer to the user and better embody the multi-directional consideration to the user, the risk of expressing the comparative spoken language can be collected, and then the comparative spoken language risk points are generated. For example, the risk information corresponding to the sample user may be that the ear is not audible and the eyes are not visible, the written language is expressed as deafness and blindness, and the corresponding risk points are also expressed by "ear is not audible" and "eyes are not visible" for the user to get closer to the sample user. Another advantage of representing the risk points in this way is that when the insurance knowledge graph generated based on the expressed risk points is applied to a question-answering system or an insurance product recommendation system, the questions of the target users (which can be understood as consumers, clients and the like) can be solved more quickly and accurately, and more accurate insurance products are recommended for the target users. Because most target users tend to use the spoken language expression when consulting the problem, answers matched with the spoken language expression of the target users can be quickly and accurately found based on the insurance knowledge graph, and then insurance products more conforming to the target users are selected and recommended to the target users, so that the experience of the target users is improved, and the friendliness of intelligent interaction is improved.

Optionally, in a possible implementation manner, after different types of risk information preset by a sample user are collected and risk points corresponding to the sample user are generated based on the different types of risk information, a plurality of risk points generated at this time may be randomly combined to obtain a plurality of combined results, and each combined result is processed through a preset language scoring model to obtain a score corresponding to each combined result; and acquiring a combined result with the score larger than or equal to a preset threshold, and generating a plurality of new risk points corresponding to the sample user based on the combined result with the score larger than or equal to the preset threshold. For specific implementation, reference may be made to descriptions in S1014 to S1017, which are not described herein again. This further enriches the types of risk points, extending the coverage of risks that a user may incur.

Fig. 4 is a schematic flow chart of a method for constructing a product knowledge graph according to another embodiment of the present application, as shown in fig. 4, and optionally, in one possible implementation, the method for constructing a product knowledge graph shown in fig. 4 may include S201 to S206. The steps S201 to S204 shown in fig. 4 may refer to the above description of S101 to S104, and are not repeated herein for brevity. The following will specifically explain steps S205 to S206.

S205: and acquiring a natural query statement input by a target user.

In this embodiment, a product knowledge graph is taken as an insurance knowledge graph for example.

Illustratively, the target user is a user who wants to consult insurance-related information, and the natural query statement is an insurance-related question that the target user wants to consult.

For example, when a user wants to consult information related to insurance, a natural query statement may be input in an input interface such as a browser, an Application (APP) system, an applet, and the like on a terminal, and the terminal may be constructed to obtain the natural query statement. The natural query statement input by the target user may be a question sentence containing logic or a common natural query statement, which is not limited herein. For example, a natural query statement may be "what insurance should be bought at normal death is appropriate", "what insurance should be bought if there is a concern that the eyes cannot see", and so on. The description is given for illustrative purposes only and is not intended to be limiting.

S206: and searching the product name matched with the natural query sentence in the product knowledge graph, and recommending the product name to a target user.

Searching a node matched with the natural query statement in the insurance knowledge graph according to the natural query statement, determining whether the information stored in the node is an insurance product, and recommending the insurance product to a target user if the information stored in the node is the insurance product. If the information stored in the node is not an insurance product, searching an upper node or a lower node adjacent to the node based on the node, judging which node is stored as the insurance product, and recommending the insurance product to a target user.

For example, the natural query statement is "what insurance should be bought if there is a fear that the ear cannot hear", the node matching the natural query statement is "accidental deafness", the information stored in the upper node adjacent to the "accidental deafness" is "accidental disability insurance", and the "accidental disability insurance" is recommended to the target user.

In the embodiment, the insurance knowledge map is applied to an insurance product recommendation system, the actual insurance application requirements of users can be grasped, the most appropriate insurance products are recommended to the users, and the sales conversion rate and the sales volume of the insurance products are improved.

Optionally, in a possible implementation manner, after step S104, or after step S204, or after step S206, the method may further include: and uploading the product knowledge graph to a block chain.

In this embodiment, uploading the insurance knowledgegraph into the blockchain can ensure its security and fair transparency to the user. And the insurance knowledge graph is uploaded to the block chain, so that the insurance knowledge graph can be prevented from being maliciously distorted by means of the characteristic that files on the block chain cannot be randomly distorted, the problem related to insurance can be solved for the user accurately according to the insurance knowledge graph in the follow-up process, and the most appropriate insurance product is recommended for the user.

The blockchain referred to in this example is a novel application mode of computer technologies such as distributed data storage, point-to-point transmission, consensus mechanism, encryption algorithm, and the like. A block chain (Blockchain), which is essentially a decentralized database, is a series of data blocks associated by using a cryptographic method, and each data block contains information of a batch of network transactions, so as to verify the validity (anti-counterfeiting) of the information and generate a next block. The blockchain may include a blockchain underlying platform, a platform product service layer, an application service layer, and the like.

Referring to fig. 5, fig. 5 is a schematic diagram of a product knowledge graph constructing apparatus according to an embodiment of the present application. The device comprises units for performing the steps in the embodiments corresponding to fig. 1, 3, 4. Please refer to the related descriptions in the embodiments corresponding to fig. 1, fig. 3, and fig. 4. For convenience of explanation, only the portions related to the present embodiment are shown. Referring to fig. 5, it includes:

a first obtaining unit 310, configured to obtain a plurality of risk points corresponding to a sample user, where the plurality of risk points are used to represent a plurality of risks corresponding to the sample user, and one risk point corresponds to one risk;

a second obtaining unit 320, configured to obtain a plurality of product information, where each product information includes a product category and a plurality of product names corresponding to the product category;

a determining unit 330, configured to determine a top-bottom relationship among a plurality of product categories, a plurality of product names, and a plurality of risk points, and determine node attributes and storage locations based on the top-bottom relationship, where the node attributes include a node attribute corresponding to each product category, a node attribute corresponding to each product name, and a node attribute corresponding to each risk point, and the storage locations include a storage location corresponding to each product category, a storage location corresponding to each product name, and a storage location corresponding to each risk point;

a storage unit 340, configured to store each product category, each product name, and each risk point in a preset graph database based on the node attribute and the storage location, so as to obtain a product knowledge graph.

Optionally, the first obtaining unit 310 is specifically configured to:

acquiring data to be analyzed, wherein the data to be analyzed comprises a session log and comment information, the session log comprises a session generated when a sample user consults insurance, and the comment information is information of the sample user comment insurance;

extracting various risk information of a sample user from the data to be analyzed, and generating a plurality of candidate risk points based on the various risk information of the sample user;

and determining the corresponding risk point of the sample user in the candidate risk points.

Optionally, the first obtaining unit 310 is further configured to:

randomly combining the candidate risk points to obtain a plurality of combination results;

processing each combined result through a preset language scoring model to obtain a score corresponding to each combined result;

acquiring a target combination result with the score larger than or equal to a preset threshold value;

and generating a risk point corresponding to the sample user based on the target combination result.

Optionally, the first obtaining unit 310 is further configured to:

acquiring different types of risk information preset by the sample user;

and generating the corresponding risk points of the sample users based on the different types of risk information.

Optionally, the apparatus further comprises:

the third acquisition unit is used for acquiring a natural query sentence input by a target user;

and the recommending unit is used for searching the product name matched with the natural query sentence in the product knowledge graph and recommending the product name to the target user.

Optionally, the apparatus further comprises:

and the uploading unit is used for uploading the product knowledge graph to a block chain.

Referring to fig. 6, fig. 6 is a schematic diagram of a terminal for building a product knowledge graph according to another embodiment of the present application. As shown in fig. 6, the product knowledge graph building terminal 4 of the embodiment includes: a processor 40, a memory 41, and computer instructions 42 stored in the memory 41 and executable on the processor 40. The processor 40, when executing the computer instructions 42, implements the steps in the various product knowledge graph construction method embodiments described above, such as S101-S104 shown in fig. 1. Alternatively, the processor 40, when executing the computer instructions 42, implements the functions of the units in the embodiments described above, such as the functions of the units 310 to 340 shown in fig. 5.

Illustratively, the computer instructions 42 may be divided into one or more units that are stored in the memory 41 and executed by the processor 40 to accomplish the present application. The one or more elements may be a series of computer instruction segments capable of performing certain functions that describe the execution of the computer instructions 42 in the building terminal 4 of the product knowledge graph. For example, the computer instructions 42 may be divided into a first acquisition unit, a second acquisition unit, a determination unit, and a storage unit, each unit having the specific functions as described above.

The terminal for building the product knowledge graph may include, but is not limited to, a processor 40 and a memory 41. Those skilled in the art will appreciate that FIG. 6 is merely an example of a product knowledge graph build terminal 4 and does not constitute a limitation of a product knowledge graph build terminal, and may include more or fewer components than shown, or some components in combination, or different components, e.g., the product knowledge graph build terminal may also include input output terminals, network access terminals, buses, etc.

The Processor 40 may be a Central Processing Unit (CPU), other general purpose Processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), an off-the-shelf Programmable Gate Array (FPGA) or other Programmable logic device, discrete Gate or transistor logic, discrete hardware components, etc. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.

The memory 41 may be an internal storage unit of the product knowledge graph constructing terminal, such as a hard disk or a memory of the product knowledge graph constructing terminal. The memory 41 may also be an external storage terminal of the product knowledge graph building terminal, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card), and the like equipped on the product knowledge graph building terminal. Further, the memory 41 may also include both an internal storage unit of the construction terminal of the product knowledge graph and an external storage terminal. The memory 41 is used for storing the computer instructions and other programs and data required by the terminal. The memory 41 may also be used to temporarily store data that has been output or is to be output.

The embodiment of the present application further provides a computer storage medium, where the computer storage medium may be nonvolatile or volatile, and the computer storage medium stores a computer program, and the computer program, when executed by a processor, implements the steps in the above-mentioned method for constructing a product knowledge graph.

The above-mentioned embodiments are only used for illustrating the technical solutions of the present application, and not for limiting the same; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; such modifications and substitutions do not cause the essential features of the corresponding technical solutions to depart from the spirit scope of the technical solutions of the embodiments of the present application, and are intended to be included within the scope of the present application.

Claims

1. A method for constructing a product knowledge graph is characterized by comprising the following steps:

2. The construction method according to claim 1, wherein the obtaining of the plurality of risk points corresponding to the sample user comprises:

3. The construction method according to claim 2, wherein after extracting the plurality of types of risk information of the sample user from the data to be analyzed and generating the plurality of candidate risk points based on the plurality of types of risk information of the sample user, the construction method further comprises:

4. The construction method according to claim 2, wherein before obtaining the plurality of risk points corresponding to the sample user, the construction method further comprises:

acquiring different types of risk information preset by the sample user;

5. The construction method according to any one of claims 1 to 4, wherein after storing each product category, each product name, and each risk point in a preset graph database based on the node attributes and the storage locations, and obtaining a product knowledge graph, the construction method further comprises:

acquiring a natural query statement input by a target user;

and searching the product name matched with the natural query statement in the product knowledge graph, and recommending the product name to the target user.

6. The building method according to claim 1, wherein after storing each product category, each product name, and each risk point in a preset graph database based on the node attributes and the storage locations, and obtaining a product knowledge graph, the building method further comprises:

and uploading the product knowledge graph to a block chain.

7. An apparatus for building a product knowledge graph, comprising:

8. The build device of claim 7, further comprising:

9. A product knowledge graph building terminal comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the method according to any one of claims 1 to 6 when executing the computer program.

10. A computer-readable storage medium, in which a computer program is stored which, when being executed by a processor, carries out the method according to any one of claims 1 to 6.