CN114461781A

CN114461781A - Data storage method, data query method, server and storage medium

Info

Publication number: CN114461781A
Application number: CN202111659502.1A
Authority: CN
Inventors: 管牧华; 周航; 戴健; 葛灿辉
Original assignee: Alibaba Cloud Computing Ltd
Current assignee: Alibaba Cloud Computing Ltd
Priority date: 2021-12-30
Filing date: 2021-12-30
Publication date: 2022-05-10

Abstract

The embodiment of the application provides a data storage method, a data query method, a server and a storage medium, wherein the data storage method comprises the following steps: acquiring a target text of a defined target and object data of the target; identifying conditions corresponding to the targets and condition indexes associated with the conditions at least from the target texts; the condition index defines a related object index of the object and the condition and a rule which needs to be met by the related object index; obtaining an object index from the object data; at least constructing a condition entity corresponding to the condition, a condition index entity corresponding to the condition index and an object index entity corresponding to the object index in the knowledge graph; and constructing an incidence relation between the condition index entity and the object index entity based on the incidence object index defined by the condition index. The embodiment of the application can improve the expansibility of data storage and provide support for high-performance associated query.

Description

Data storage method, data query method, server and storage medium

Technical Field

The embodiment of the application relates to the technical field of data processing, in particular to a data storage method, a data query method, a server and a storage medium.

Background

At present, data is mainly stored through a relational database, and the relational database is a database for storing data in a relational model. Although the relational database has many advantages, the relational database manages data in rows and columns of tables, and is difficult to apply to a scenario of relational query. Therefore, how to provide a novel data storage scheme to provide support for associated query becomes a technical problem that needs to be solved urgently by those skilled in the art.

Disclosure of Invention

In view of this, embodiments of the present application provide a data storage method, a data query method, a server, and a storage medium to provide support for association query.

In order to achieve the above object, the embodiments of the present application provide the following technical solutions.

In a first aspect, an embodiment of the present application provides a data storage method, including:

acquiring a target text defining a target and object data of the target;

at least identifying a condition corresponding to the target and a condition index associated with the condition from the target text; the condition index defines a related object index of the object and the condition and a rule which needs to be met by the related object index;

obtaining an object index of the object from the object data;

constructing a condition entity corresponding to the condition, a condition index entity corresponding to the condition index and an object index entity corresponding to the object index in a knowledge graph;

and constructing an association relation between the condition index entity and the object index entity based on an associated object index defined by the condition index.

In a second aspect, an embodiment of the present application provides a data query method, including:

acquiring a data query request;

calling a knowledge graph, wherein the knowledge graph stores data based on the data storage method of the first aspect;

determining a data query result based on the knowledge-graph.

In a third aspect, an embodiment of the present application provides a data query method, including:

displaying an object index submission page, wherein a plurality of object index items are displayed on the object index submission page;

determining object indexes of the first object corresponding to the object index items, and sending a data query request based on the object indexes corresponding to the object index items;

acquiring a data query result determined based on a knowledge graph, and displaying the data query result on a data result page; wherein the knowledge-graph stores data based on the data storage method of the first aspect.

In a fourth aspect, embodiments of the present application provide a server, including at least one memory and at least one processor, the memory storing one or more computer-executable instructions, and the processor invoking the one or more computer-executable instructions to perform the data storage method according to the first aspect, and/or the data query method according to the second aspect.

In a fifth aspect, embodiments of the present application provide a storage medium storing one or more computer-executable instructions that, when executed, implement a data storage method as described in the first aspect above, or a data query method as described in the second aspect above, or a data query method as described in the third aspect above.

In a sixth aspect, an embodiment of the present application provides a computer program, which when executed, implements the data storage method according to the first aspect, or the data query method according to the second aspect, or the data query method according to the third aspect.

It can be seen that, in the embodiment of the present application, the knowledge graph is formed by at least an entity on the target side, an entity on the object side, and an association relationship between the entities, where the entity on the target side includes at least a condition entity and a condition index entity, and the entity on the object side includes at least an object index entity. The method and the device for identifying the target text and the object data of the target can use the target text of the target and the object data of the target as data sources of knowledge graph storage targets and object data, so that at least conditions corresponding to the target and condition indexes associated with the conditions are identified from the target text, wherein the condition indexes define associated object indexes associated with the conditions and rules required to be met by the associated object indexes; and, an object index of the object is acquired from the object data. The entity is represented by a node based on the knowledge graph, and the embodiment of the application can at least construct a condition entity corresponding to the condition, a condition index entity corresponding to the condition index and an object index entity corresponding to the object index in the knowledge graph so as to store information of the entity at the target side and the entity at the object side in the knowledge graph; based on the knowledge graph, the relation between the entities is represented by edges, the embodiment of the application can construct the incidence relation between the condition entities and the condition index entities in the knowledge graph, and construct the incidence relation between the condition index entities and the object index entities based on the incidence object indexes defined by the condition indexes, so that the information of the entities at the target side and the object side stored in the knowledge graph is correlated at least through the incidence relation between the condition index entities and the object index entities, and the data of the correlated target and the object is realized. Therefore, in the embodiment of the application, the data of the target and the object are stored and described in a mode that the knowledge graph represents the entities by the nodes and represents the relationship between the entities by the edges, the high expansibility of data storage can be realized by using the flexible mode definition capability of the knowledge graph, and the relationship is established at the node level of the entities instead of the table level relationship of the relational database, so that support is provided for realizing high-performance and quick association query.

Drawings

In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly introduced below, it is obvious that the drawings in the following description are only embodiments of the present application, and for those skilled in the art, other drawings can be obtained according to the provided drawings without creative efforts.

Fig. 1 is a block diagram of a system architecture according to an embodiment of the present application.

Fig. 2 is a flowchart of a data storage method according to an embodiment of the present application.

Fig. 3A is an exemplary diagram of a knowledge-graph provided by an embodiment of the present application.

Fig. 3B is another example diagram of a knowledge-graph provided by an embodiment of the present application.

FIG. 3C is an exemplary diagram of a target text.

FIG. 4 is a diagram of yet another example of a knowledge graph provided by an embodiment of the present application.

Fig. 5 is another flowchart of a data storage method according to an embodiment of the present application.

Fig. 6A is a diagram of another example of a knowledge-graph provided by an embodiment of the present application.

Fig. 6B is yet another example diagram of a knowledge-graph provided by an embodiment of the present application.

Fig. 7 is a further flowchart of a data storage method according to an embodiment of the present application.

Fig. 8A is a diagram of yet another example of a knowledge graph provided by an embodiment of the present application.

Fig. 8B is another example of a knowledge graph provided in an embodiment of the present application.

Fig. 9 is a flowchart of a data query method according to an embodiment of the present application.

Fig. 10A is another flowchart of a data query method according to an embodiment of the present application.

Fig. 10B is an exemplary diagram of an object index submission page provided in the embodiment of the present application.

Fig. 10C is an exemplary diagram of a data result page provided in the embodiment of the present application.

Fig. 11 is a block diagram of a data storage device according to an embodiment of the present application.

Fig. 12 is a block diagram of a server.

Detailed Description

The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.

In the fields of intelligent data recommendation, intelligent data analysis and the like, frequent association query is required, and the association query refers to querying associated data, for example, in a government scene, querying an associated policy through an enterprise, querying an associated enterprise through an enterprise, querying an associated policy through a policy, and the like. In the relational database, the relational database manages data in rows and columns of tables, which results in that when the relational database is used for performing association query, the relational database needs to perform table join (connection) operations for many times, which undoubtedly seriously affects the performance of association query and causes too high delay of association query, and therefore a novel data storage scheme needs to be provided to support the realization of high-performance and fast association query. Furthermore, the relational database manages data in rows and columns of tables, which also results in poor extensibility of data storage, so that the novel data storage scheme also meets the requirement of high extensibility of data storage.

In the embodiment of the application, the association query mainly relates to the association query between the target and the object; goals may be action criteria and requirements that are tailored to achieve a particular purpose, such as policies, election rules, assessment rules, and the like; the subject may be any type of subject, such as a subject of a business, individual user, business, school, home, and so forth. Further, the association query of the embodiment of the present application may also relate to an association query between objects (e.g., an association query between a business and a business), and an association query between targets (e.g., an association query between a policy and a policy).

In order to facilitate the association query between the target and the object, in the embodiment of the present application, data of the target and the object may be stored in a way of a knowledge graph, that is, data of the target and the object may be described in a way that the knowledge graph represents entities by nodes and represents an association relationship between the entities by edges, so that a flexible schema definition capability of the knowledge graph is used to implement high extensibility of data storage, and an association relationship established at a node level by the knowledge graph is used instead of a table-level relationship of a relational database to implement association query supporting high performance and high speed.

The knowledge graph is a knowledge management storage structure based on a graph data structure, in the knowledge graph, entities are represented by nodes, association relations between the entities are represented by edges between the nodes, and the entities can be regarded as formal representations of knowledge concepts and knowledge relations.

As an alternative implementation, fig. 1 schematically shows a block diagram of a system architecture of an embodiment of the present application. As shown in fig. 1, the system architecture may include: a terminal device 11 and a server 12. The terminal device 11 may be a user device used by a user, such as a smart phone, a notebook computer, and so on. The server 12 may be a server device disposed on the network side, and may be implemented by a single server or a server cluster; in an alternative implementation, the server 12 may be a cloud server based on cloud computing. In the embodiment of the present application, the server may store at least data of the target and the object by means of a knowledge graph, and when the terminal device 11 sends a data query request, the server may perform an association query of the data based on the knowledge graph, so as to feed back a data query result to the terminal device 11.

In the embodiment of the application, in order to realize the association query between the target and the object based on the knowledge graph, the server needs to store at least data of the target and the target object in a knowledge graph mode. Based on this, the embodiment of the application provides a novel data storage scheme, so that at least data of a target and an object are stored through a knowledge graph, support is provided for realizing high-performance and quick associated query, and the expansibility of data storage is further improved.

As an alternative implementation, fig. 2 exemplarily shows an alternative flowchart of a data storage method provided by an embodiment of the present application. The method flow may be applied to a server, and referring to fig. 2, the method flow may include the following steps.

In step S210, a target text defining a target and object data of the target are acquired.

Based on the application, at least the data of the target and the object are stored in the knowledge graph, the target text can be regarded as a data source of the data of the knowledge graph storage target, and the object data can be regarded as a data source of the data of the knowledge graph storage object. In some embodiments, the target text of the target may describe the full-text content of the target, such as policy text or the like. As an alternative implementation, the target text of the target may be obtained from a data source such as a website, a database, etc. of an organization issuing the target (e.g., a government organization issuing policies). In some embodiments, the object data of the object may be data describing the condition of the object, such as enterprise data or the like.

In step S211, at least a condition corresponding to the target and a condition indicator associated with the condition are identified from the target text, where the condition indicator defines an associated object indicator associated with the condition and a rule that the associated object indicator needs to satisfy.

In step S212, an object index of the object is acquired from the object data.

In step S213, at least a condition entity corresponding to the condition, a condition index entity corresponding to the condition index, and an object index entity corresponding to the object index are constructed in the knowledge graph.

Since the knowledge graph represents the entities by nodes, it is necessary to construct the entities on the object side and the entities on the object side in a node manner in the knowledge graph to store data of the objects and the objects in the knowledge graph. For example, storing data of policies and businesses in a knowledge graph, it is necessary to construct entities on the policy side and entities on the business side in a node manner in the knowledge graph.

In the embodiment of the present application, the entities on the target side may include at least a condition entity and a condition index entity; therefore, the condition corresponding to the target and the condition index associated under the condition can be identified from the target text, so that the condition entity corresponding to the condition and the condition index entity corresponding to the condition index are constructed by the nodes in the knowledge graph. As an optional implementation, the embodiment of the application may perform intelligent identification on the target text, so as to identify a condition corresponding to the target and a condition index associated with the condition from the target text.

It should be noted that, in the embodiment of the present application, the target text of the target may define a condition that the object meets the target requirement (for example, the policy text may define a condition that the enterprise meets the policy requirement), and the condition indicator may be associated with the condition indicator under the condition, and the condition indicator may define an object indicator (referred to as an associated object indicator) associated with the condition in the object, and a rule that the associated object indicator needs to meet. The target index may be index data indicating a specific condition of the target, such as an enterprise index of a annual sales amount, a annual tax amount, a staff amount, and the like of the enterprise. In one example, the policy text may define a condition that the object satisfies the policy requirements, and a condition indicator may be associated with the condition, the condition indicator may define an associated business indicator associated with the condition in the business, and a rule that the associated business indicator is to satisfy. As an alternative implementation, the condition indicator may define a rule that the associated object indicator needs to satisfy through a regular expression, for example, the regular expression may define a formula that the associated object indicator needs to satisfy, such as the annual sales amount of an enterprise is greater than 100 ten thousand, the number of people is greater than 50, and the like.

In the embodiment of the present application, the entity on the object side may include at least an object index entity; therefore, in the embodiment of the present application, an object index of an object is obtained from object data of the object (for example, an enterprise index of an enterprise is obtained from enterprise data of the enterprise), so that an object index entity corresponding to the object index is constructed with nodes in a knowledge graph (for example, an enterprise index entity corresponding to the enterprise index is constructed with nodes in the knowledge graph).

In step S214, an association relationship between a condition index entity and a condition entity is constructed in the knowledge graph, and an association relationship between the condition index entity and an object index entity is constructed based on an associated object index defined by the condition index.

Based on the knowledge graph, the relation between the entities is represented by sides, the embodiment of the application needs to construct the incidence relation between the condition entities and the condition index entities in the knowledge graph by sides, construct the incidence relation between the condition index entities and the object index entities by sides, and realize the association between the data of the target and the data of the object stored in the knowledge graph through the incidence relation between the condition index entities at the target side and the object index entities at the object side.

In some embodiments, the embodiment of the present application may construct an association relationship from a condition entity to a condition index entity with an edge in a knowledge graph based on an associated condition index of a condition. And based on the association condition of the condition index entity, establishing the association relationship from the condition index entity to the condition entity in the knowledge graph by sides so as to realize the establishment of the association relationship between the condition entity and the condition index entity in the knowledge graph. As an alternative implementation, the associated condition index of the condition may be recorded in the attribute of the condition entity and the associated condition of the condition index may be recorded in the attribute of the condition index entity.

In some embodiments, the association relationship from the condition index entity to the object index entity may be established by edges in the knowledge graph based on the associated object index of the condition index; for example, through the associated object index defined by the condition index, the object index entity corresponding to the associated object index is found in the object index entity, so as to construct the association relationship from the condition index entity to the object index entity. Furthermore, the embodiment of the application can construct the incidence relation from the object index entity to the condition index entity in the knowledge graph by using the edge based on the incidence condition index of the object index; for example, through the associated condition indicators defined by the object indicators, the condition indicator entities corresponding to the associated condition indicators are found in the condition indicator entities, so as to construct the association relationship from the object indicator entities to the condition indicator entities. As an alternative implementation, the associated object index of the condition index may be recorded in the attribute of the condition index entity, and the associated condition index of the object index may be recorded in the attribute of the object index entity.

In some further embodiments, the information of the entity may be stored by the node attribute, for example, the attribute of the entity may be stored by the node attribute.

In an example, fig. 3A illustrates an alternative example diagram of a knowledge graph provided by an embodiment of the present application, and as shown in fig. 3A, after a condition corresponding to a target and a condition index associated with the condition are identified from a target text, an embodiment of the present application may construct a condition entity and a condition index entity in a node in the knowledge graph, and construct an association relationship between the condition entity and the condition index entity by associating the condition index of the condition entity and an association condition of the condition index entity. Meanwhile, after the object indexes are obtained from the object data of the object, the object index entity can be constructed in the knowledge graph by using the nodes, and the association relationship between the condition index entity and the object index entity is constructed by using the associated object indexes of the condition index entity and the associated condition indexes of the object index entity, so that the data of the target and the object are stored in the knowledge graph.

In further some embodiments, fig. 3B is another exemplary diagram illustrating an example of a knowledge graph provided by an embodiment of the present application, and as shown in fig. 3A and fig. 3B, in the knowledge graph illustrated in fig. 3B, the entity on the target side may further include: a target entity, a sub-target entity, and a target tag entity. In the embodiment of the present application, the target text of the target may describe one or more sub-targets, and one sub-target may have one or more target tags; for example, the implementation target comprises a plurality of sub-targets, the content under the sub-targets can be abstracted and summarized into one or more content abstraction points, and the label information of each content abstraction point can be used as a target label under the sub-targets; and one or more conditions can be associated in the content abstraction point corresponding to each target label, and one or more condition indexes can be associated under each condition. As an example, fig. 3C exemplarily shows an exemplary diagram of the target text, and as shown in fig. 3C, the target text may record the content of at least one sub-target included in the target, where the content of each sub-target is associated with at least one target tag, each target tag is associated with at least one condition, and each condition is associated with at least one condition indicator.

In some embodiments, sub-targets, target tags associated with the sub-targets, conditions associated with the target tags, and condition indicators associated with the conditions may be identified from a target text of the target, so that a target entity corresponding to the target, a sub-target entity corresponding to the sub-target, a target tag entity corresponding to the target tag, a condition entity corresponding to the conditions, and a condition indicator entity corresponding to the condition indicators are respectively constructed by nodes in a knowledge graph. Furthermore, the embodiment of the application can respectively construct the incidence relation between the target entity and the sub-target entity, the incidence relation between the sub-target entity and the target label entity, the incidence relation between the target label entity and the condition entity, and the incidence relation between the condition entity and the condition index entity by using edges in the knowledge graph.

As an optional implementation, referring to fig. 3B, in the embodiment of the present application, an association relationship between a target entity and a sub-target entity may be constructed based on the sub-targets included in the target and the targets to which the sub-targets belong; the sub-targets contained in the target can be recorded in the attribute of the target entity, and the targets to which the sub-targets belong can be recorded in the attribute of the sub-target entity. Further, based on the associated target tags of the sub-targets and the associated sub-targets of the target tags, constructing an association relationship between the sub-target entities and the target tag entities; wherein the associated target tag of the sub-target can be recorded in the attribute of the sub-target entity, and the associated sub-target of the target tag can be recorded in the attribute of the target tag entity. Further, based on the condition determined by the target label and the target label to which the condition belongs, constructing an association relation between the target label entity and the condition entity; the condition determined by the target tag can be recorded in the attribute of the target tag entity, and the target tag to which the condition belongs can be recorded in the attribute of the condition entity. The association relationship between the condition entity and the condition index entity can be constructed as described in the corresponding part above.

In some embodiments, the embodiment of the application may identify information such as sub-targets, target tags, conditions, condition indexes, and the like from the target text by intelligently identifying the target text. As an optional implementation, the target text may be encoded to obtain text features of the target text; extracting text features corresponding to the sub-targets from the text features; decoding text features corresponding to the sub-targets, identifying text phrases of label categories from decoding results, standardizing the text phrases and refining the text phrases into target labels; and identifying a data phrase from the decoding result, and extracting a condition under a target label and a condition index under the condition from the data phrase.

As an alternative implementation, the above process of intelligently recognizing the target text may be implemented based on an intelligent recognition model, for example, a bertm (Bidirectional Encoder Representation from converters) -LSTM (Long Short-Term Memory, Long Short Term Memory network) -CRF (Conditional Random Field) -based intelligent recognition model. In a possible implementation, the method and the device for encoding the target text can use a large-scale text corpus (e.g., a large-scale target text) to pre-train a BERT model, so that the BERT model is used to encode the target text to obtain text features of the target text; and then, performing text feature extraction by using a recurrent neural network in an LSTM form, and extracting text features corresponding to the sub-targets. And decoding the text features corresponding to the sub-targets (for example, decoding the text features based on CRF), so as to identify text phrases of at least one label category from the decoding result, where the identified text phrases may be standardized and refined into the target labels. After the text phrases of at least one tag category are identified, the text phrases are not directly refined into the target tags in the embodiment of the application, but the text phrases are further normalized and then refined into the target tags according to the normalized result, which mainly solves the problem of data redundancy, for example, "computer science research and development center" and "computer science research center" can be merged into a standardized "computer science research and development center". In an alternative implementation of normalizing text phrases, considering that the text phrases are short in length, the embodiments of the present application may employ a string similarity algorithm to perform clustering of the text phrases.

Further, after the text features corresponding to the sub-targets are decoded, the data phrases can be identified from the decoding results, so that the conditions and condition indexes under the conditions of the target tags are extracted from the data phrases. As an alternative implementation, the data phrases may include numeric phrases and claim language phrases; the numerical phrases are generally attribute values of annual report type attributes, for example, "annual development cost reaches 500 ten thousand", so the characteristic modes of the numerical phrases are "numerical value" and "unit", and by using the characteristic modes of the numerical phrases, the embodiment of the application can perform characteristic extraction on the numerical phrases through the characteristic template to obtain conditions and condition indexes of the numerical phrases; for example, for the numerical phrase "annual development cost reaches 500 ten thousand" in the above example, the embodiment of the present application may extract the condition "annual development cost reaches 500 ten thousand" and the condition index "annual development cost", and define a regular expression of "annual development cost ≧ 500 ten thousand" in the condition index. As an alternative implementation, the required language phrase may be a phrase defining requirements, such as "enterprise needs to be listed" and the like, and the embodiment of the present application may extract the required language phrase from the decoding result of the text feature by using required words (such as "should", "needs", "meets", and the like), and intelligently identify conditions and condition indexes under the conditions from the required language phrase.

In further embodiments, fig. 4 is a diagram illustrating still another alternative example of a knowledge-graph provided by embodiments of the present application. In the knowledge-graph shown in fig. 4, in combination with fig. 3B and fig. 4, the entity on the object side may further include an object entity. According to the embodiment of the application, for an object (such as an enterprise), an object entity corresponding to the object can be constructed by nodes in a knowledge graph, and information of the entity can be stored by node attributes based on the knowledge graph.

It should be noted that the object indicator may be a special attribute, and the object indicator may be inherited to a type (type, property, etc.) of an attribute built in the knowledge graph. In a possible implementation, the object index may describe the attribute content of the object itself, but there is also a difference between the object index and the attribute of the object: the object index may be used by the condition index as an element. For example, the annual sales amount of an enterprise is used as an enterprise index, and the enterprise index may correspond to a specific condition index in which the annual sales amount reaches a certain value in the condition, for example, the enterprise index of the annual sales amount may correspond to a condition index of "annual sales amount > 100 ten thousand" in the condition, and thus the target index may be used as an element of the condition index.

Further, as shown in fig. 4, in the embodiment of the present application, an association relationship between an object entity and a target tag entity may also be constructed in a knowledge graph. For example, the embodiment of the present application may construct an association relationship from the object entity to the target tag entity based on the associated target tag of the object entity.

The knowledge graph constructed in the foregoing embodiment of the present application is implemented based on existing data, and the existing data may not fully and completely represent the relationship between the entity on the target side and the entity on the object side, which may result in a missing relationship between the entity on the target side and the entity on the object side, so that the embodiment of the present application needs to supplement the missing relationship between the entity on the target side and the entity on the object side. According to the embodiment of the application, the missing relationship between the entity on the supplementary target side and the entity on the object side can be realized through the missing path. In some embodiments, embodiments of the present application may determine missing paths between entities in a knowledge-graph, including missing paths for object entities and conditional entities, as illustrated in conjunction with fig. 4. Further, the missing path may further include a missing path of the object entity and the target tag entity, and/or the target entity, or a missing path of the missing object index entity and the condition index entity. The missing path may be in an entity form in the knowledge graph, and the embodiment of the application may implement the association relationship between the entity on the supplementary object side and the entity on the target side by defining the association relationship between the missing path and the entity on the object side and the association relationship between the missing path and the entity on the target side.

In some further embodiments, as shown in fig. 4, the missing path and the condition indicator entity may be regarded as a connection path (for example, the connection path includes the missing path and the condition indicator entity), and a confidence level of the connection path is determined, where the confidence level may indicate a credibility of an association relationship between entities connected by the connection path.

In further embodiments, as shown in connection with FIG. 4, the condition indicator entity and the missing path may be a CVT (Compound Value Type) entity that allows multiple entities to be bound together. Further, the object entity may have a plurality of composite attributes in addition to the normal attributes, and the plurality of composite attributes may be represented by a plurality of entities in the form of a CVT, for example, as shown in connection with fig. 4, the object entity may have composite attributes F1 and F2 in addition to the normal attributes P1 and P2, and the composite attributes F1 and F2 may be represented by the entities of the CVT, respectively.

It should be noted that the knowledge graph shown in fig. 3B and fig. 4 is only an optional implementation of extension based on the knowledge graph shown in fig. 3A, and based on the principle of the method flow shown in fig. 2 and the basic structure of the knowledge graph shown in fig. 3A, the embodiment of the present application can completely perform adaptive adjustment and extension of the knowledge graph structure according to the specific object and situation, which should also be considered to be within the scope of the embodiment of the present application.

The data storage method provided by the embodiment of the application stores and describes data of targets and objects in a mode that the knowledge graph represents the entities by nodes and represents the relationship between the entities by edges, can realize high expansibility of data storage by using flexible mode definition capability of the knowledge graph, establishes the relationship at the node level of the entities and provides support for realizing high-performance and quick association query.

In an alternative application example, the data storage method provided by the embodiment of the application can be applied to storing data of enterprises and policies so as to support association query of the enterprises and the policies. It should be noted that in the government scene, the government organization has issued many favorable policies, and it is necessary to digitize these policies and the information of the enterprise itself to facilitate the inquiry of the policies and the enterprise data. In this context, if a relational database is used to store data of policies and enterprises, the relational database (e.g., mysql) is used to create tables to manage data of enterprises and policies, which may cause the following problems:

the flexibility is poor, the relational database is a strong-mode database, when the mode is changed, columns need to be added and deleted, and the table is locked in the process, so that the service is interrupted;

the relation between any two types of data among the enterprise basic information, the enterprise extension information, the policy key points, the conditions and the policy labels needs to maintain a relation table, and the quantity of the relation table is expanded;

due to the requirements on performance and the requirements on associated query, a specific table needs to be established for each type of extended information (such as annual sales amount, industry information, main personnel information and associated tags) of an enterprise, and the maintenance cost is high;

the correlation query performance is weak, the overhead of each table-linked query is the Cartesian product of two table data, and the method cannot be applied to a real-time recommendation scene with low delay requirement;

meanwhile, due to the fact that policy contents in various regions and requirements for enterprises are different, a strong-mode relational database is used for modeling, and large expenses are brought to the relational database for storing data due to changes of modes.

In order to solve the problem of storing enterprise and policy data in a relational database, the data storage method provided by the embodiment of the application can be used for describing and storing the enterprise and policy data in a mode that nodes of a knowledge graph represent entities and edges represent the relationships between the entities, so that the high expansibility of enterprise and policy data storage is realized by using the flexible mode definition capability of the knowledge graph, and the high-performance and quick association query is supported by establishing the relationships at the node level. That is to say, the embodiment of the application can provide a lighter-weight and more flexible storage scheme for enterprise and policy data storage.

Based on the data storage scheme provided by the embodiment of the application, in the data storage scene of enterprises and policies, aiming at a target side, the previously described target comprises a policy, the target text comprises a policy text, the sub-targets comprise main points of the policy under the policy, the target label comprises a policy label under the main points of the policy, the policy label is associated with a condition, and the condition index is associated with the condition under the condition; for the object side, the object comprises an enterprise, and the object index comprises an enterprise index.

As an alternative implementation, fig. 5 exemplarily shows another alternative flowchart of the data storage method provided by the embodiment of the present application. The method flow may be implemented by a server, and as shown in fig. 5, the method flow may include the following steps.

In step S510, a policy text of the policy is acquired.

Policies are action criteria and requirements set by policy agencies to achieve certain goals, and policies may be issued by government agencies. The full content of the policy may be described in the policy text. In some embodiments, the policy text of the policy may be obtained from a data source such as a website, a database, etc. of the administration institution.

In step S511, the policy text is analyzed to determine a policy point from the policy text.

To achieve the policy goal of a policy, multiple sub-objectives may need to be implemented, one sub-objective may be considered as one policy point, and a policy may contain multiple policy points. For example, a policy aiming at accelerating talent development of an enterprise may have sub-objectives of "accelerating intelligent manufacturing", "stabilizing talent team of the enterprise", "giving a certain reward", etc., and these sub-objectives may be regarded as the main points of the policy.

In some embodiments, the policy text may be encoded to obtain text features of the policy text, and then the text features corresponding to the policy key points are extracted to obtain the policy key points. As an alternative implementation, the embodiment of the application can intelligently identify the policy text of the policy through an intelligent identification model, such as the intelligent identification model based on BERT-LSTM-CRF. As an optional implementation, in the embodiment of the present application, a BERT model may be pre-trained using a large-scale text corpus (e.g., a large-scale policy text), so that the BERT model is used to encode the policy text to obtain text features of the policy text; and then, extracting text features by using a recurrent neural network in an LSTM form, thereby extracting text features corresponding to policy key points to obtain the policy key points.

In step S512, a policy tag is determined according to the policy point.

The policy tag can be tag information which is related to the main point of the policy and is obtained by refining the content of the main point of the policy; for example, one policy tag is tag information of one content abstraction point of the policy gist. In the embodiment of the application, one policy tag can be abstracted from a plurality of conditions, and one policy principal point can be associated with a plurality of different policy tags. Through the policy label of the policy, the embodiment of the application can inquire similar policies.

In some embodiments, after extracting the text features corresponding to the policy principal points, the text features corresponding to the policy principal points may be decoded (for example, the text features are decoded based on CRF), so as to identify text phrases of at least one policy tag category from the decoding result, and the text phrases identified may be normalized and refined into policy tags

In step S513, the condition corresponding to the policy tag and the condition index of the condition are determined.

After determining the policy tag, embodiments of the present application may determine the associated conditions under the policy tag. In some embodiments, the condition may be a condition content that needs to be met to meet the requirement of the policy label, for example, whether a policy label meets the condition content such as a specific requirement of annual sales volume and a specific requirement of annual tax amount for the enterprise, and these condition contents may be used as a condition for describing the policy label, i.e., a condition associated under the policy label.

After the condition is determined, the embodiment of the present application may determine the content of the rule describing the condition, so as to obtain a condition indicator associated under the condition, for example, if a condition requires that the annual sales volume of the enterprise reach a certain specific value, the condition is associated with the condition indicator corresponding to the annual sales volume, and the condition indicator defines a rule (for example, a rule expression of the specific value that the annual sales volume needs to reach) that the annual sales volume needs to meet. The embodiment of the application can make specific description on the conditions and rules that an enterprise needs to meet under the policy label through the conditions of the policy label and the condition indexes of the conditions, for example, a policy text of a policy describes that the condition that the annual sales volume of an excellent enterprise should meet is up to 5 million yuan, the condition that the annual sales volume is up to 5 million yuan is provided under the policy label of the excellent enterprise, the condition is associated with the condition index corresponding to the annual sales volume, and the condition index defines the rule expression of the annual sales volume.

In some embodiments, after decoding the text features corresponding to the policy key points, the embodiments of the present application may identify data phrases from the decoding result, so as to extract conditions under the policy label and condition indicators under the conditions from the data phrases.

In step S514, the policy focus, the policy label, the condition, and the condition indicator are respectively used as policy-side entities in the knowledge graph, and the policy-side entities are represented by nodes in the knowledge graph and the association relationship between the policy-side entities is represented by edges.

The knowledge graph is a data structure based on a graph and comprises nodes and edges, wherein the nodes represent entities, and the edges represent incidence relations between the entities; an entity refers to something in the real world, such as a person, place name, company, animal, etc.; the incidence relation is used for representing the relation between different entities; the knowledge graph is thus essentially a semantic network. When the knowledge graph is used for storing the information of the policy side, the entity corresponding to the information of the policy side in the knowledge graph (called the entity of the policy side) needs to be determined, and the association relationship between the entities of the policy side is represented by the edge.

In some embodiments, the policy-side entities constructed in the knowledge graph according to the embodiments of the present application may include: policy entity, policy key entity, policy label entity, condition index entity. Therefore, after acquiring information of the policy, the policy gist, the policy label, the condition, and the condition index from the policy text, the embodiment of the present application may respectively construct the policy entity, the policy gist entity, the policy label entity, the condition entity, and the condition index entity by using nodes in the knowledge graph, and represent the association relationship between the entities on the policy side.

In one example, fig. 6A is a diagram illustrating another example of a knowledge graph provided by an embodiment of the present application, and as shown in fig. 6A, in the knowledge graph, the embodiment of the present application may construct a policy entity, a policy gist entity, a policy tag entity, a condition entity, and a condition index entity in nodes in the knowledge graph. It should be noted that the policy entity, the policy gist entity, the policy tag entity and the condition entity may be general Type entities in the knowledge map, the condition indicator entity may be a CVT (Compound Value Type) entity, and the CVT entity allows binding of multiple entities together, for example, a row in the "cast table" includes information of actors, played roles, and the like, and may be understood as a CVT.

Meanwhile, the entities on the policy side of the association relationship exist among each other, and the association relationship among the entities is represented by edges (directed edges). In some embodiments, the attribute of the entity on the policy side may define the entity on the policy side with the association relationship, so that the embodiment of the present application may represent the association relationship between the entities on the policy side in the form of an edge based on the attribute of the entity on the policy side.

In some embodiments, the policy may contain one or more policy points, and the attributes of the policy entity may include at least one of: policy name, issuing time, issuing department, effective time, expiration time, contents of required nodes and the like. In some embodiments, the attributes of the policy gist entity may include at least one of: associating policy labels, declaration starting time, declaration ending time, whether the declaration can be redeemed or not, whether the declaration can be declared, counties and districts, cities and districts, provinces, policies to which the policies belong, policy key points, full text, policy key point numbers and the like. In some embodiments, the attributes of the policy tag entity may include at least one of: tag name, tag id, associated policy point, condition of decision, etc. In some embodiments, the property of the conditional entity may include at least one of: associated condition indicators, affiliated policy tags, condition codes, and the like. In some embodiments, the property of the condition indicator entity may comprise at least one of: associating enterprise indexes, association conditions, confidence degrees, regular expressions and the like; the regular expression of the condition index defines that the enterprise index meets the requirement which the condition should meet.

With reference to fig. 6A, the attributes of the policy entity may define the main points of the policy contained in the policy, and the attributes of the main point of the policy entity may define the policy to which the main points of the policy belong, so that the embodiment of the present application may show the association relationship between the main points of the policy contained in the policy and the main points of the policy belonging to the policy through the edge in the knowledge graph. The attribute of the policy main point entity can define the associated policy tag of the policy main point, and the attribute of the policy tag entity can define the associated policy main point of the policy tag, so that the embodiment of the application can show the association relationship between the policy main point associated policy tag (multi-value) and the policy main point (multi-value) associated with the policy tag by the edge in the knowledge graph, and the multi-value and the single value can show the corresponding relationship between multiple values and single values. In the embodiment of the present application, the association relationship between the condition (multiple values) determined by the policy tag and the policy tag (single value) to which the condition belongs may be represented by an edge in the knowledge graph. The attribute of the condition entity may define an associated condition index of the condition, and the attribute of the condition index entity may define an associated condition of the condition index, so that the embodiment of the present application may indicate, by an edge, an associated relationship between the associated condition index (multiple values) of the condition and the associated condition (single value) of the condition index in the knowledge graph.

Furthermore, in the embodiment of the present application, the industry and region tags included in the policy tags may be extracted from the existing enterprise and policy data through NLP (Natural Language Processing), manual review, and the like.

In step S515, an association relationship between the condition index entity and the entity on the enterprise side is established in the knowledge graph.

After an entity on the policy side is constructed in the knowledge graph, in order to associate the entity on the policy side with the entity on the enterprise side, in the embodiment of the present application, at least a condition indicator entity may be used as a connection entity for connecting an enterprise and the policy in the knowledge graph, for example, an associated enterprise indicator defined based on an attribute of the condition indicator entity. In a case that the condition indicator entity is used as the connection entity, the embodiment of the present application may construct, based on the attribute of the condition indicator entity, an edge of the condition indicator entity pointing to the entity on the enterprise side (for example, an edge of the enterprise indicator entity corresponding to the condition indicator entity pointing to the associated enterprise indicator), so as to establish an association between the entity on the policy side and the entity on the enterprise side.

In one example, fig. 6B is a diagram illustrating yet another example of a knowledge graph provided by an embodiment of the present application, which is shown in conjunction with fig. 6A and 6B, and the embodiment of the present application may construct an edge from a condition indicator entity to an enterprise indicator entity in the knowledge graph by associating an enterprise indicator (single value) indicated in an attribute of the condition indicator entity.

Fig. 7 schematically shows a further alternative flowchart of a data storage method provided by an embodiment of the present application. The method flow may be implemented by a server, and as shown in fig. 7, the method flow may include the following steps.

In step S710, an enterprise index is obtained from enterprise data of an enterprise.

The enterprise data for the enterprise may be provided by a government agency or enterprise. The enterprise data of the enterprise bears enterprise indexes of the enterprise, and the enterprise indexes can be obtained from the enterprise data.

The enterprise indexes such as enterprise business scope, address, tax amount, sales amount and the like are used for index data for representing the enterprise condition. As an alternative implementation, in the knowledge graph, the enterprise index may be a special attribute, and the enterprise index may be inherited to the type (type, property, etc.) of the attribute built in the knowledge graph. In a possible implementation, the enterprise metrics may describe the content of the attributes of the enterprise itself, but the enterprise metrics are different from the attributes of the enterprise: the business index may be used as an element by the condition index. For example, the annual sales volume of a business is used as a business index, which may correspond to a condition index that requires the business to achieve a certain amount of sales under the condition.

In step S711, the enterprise and the enterprise index are respectively used as the enterprise-side entity in the knowledge-graph, and the enterprise-side entity is represented as a node in the knowledge-graph.

In step S712, the attributes of the business entity and the business index entity are saved in the knowledge graph as node attributes, and the association relationship between the business entity and the business index entity and the node attributes is constructed.

In some embodiments, the enterprise-side entities constructed in the knowledge graph according to the embodiments of the present application may include enterprise entities and enterprise index entities. Therefore, the enterprise and enterprise indexes can be respectively used as the entities on the enterprise side in the knowledge graph, the entities on the enterprise side are represented by nodes in the knowledge graph, and the association relation between the entities on the entity side is represented by edges.

In an example, fig. 8A is a diagram illustrating still another example of a knowledge graph provided by an embodiment of the present application, which is shown in fig. 6B and fig. 8A, and in the knowledge graph, an enterprise and an enterprise index are respectively regarded as entities on an enterprise side and are respectively represented by nodes, so that an enterprise entity and an enterprise index entity are constructed in the knowledge graph. In some further embodiments, the information of the entity may be stored by the node attribute, for example, the attribute of the entity may be stored by the node attribute, and the attribute of the entity includes two types, one is a relational type, and the other is a basic type; the attributes of the basic type can be represented by the basic type (the basic type includes character strings, Boolean, numbers and the like); a relational attribute means that the attribute points to another entity. Therefore, the embodiment of the application can further construct node attributes in the knowledge graph to store the attributes of the enterprise entity and the enterprise index entity, and construct an association relationship between the enterprise entity and the node attributes (the enterprise entity owns the node attributes), and an association relationship between the enterprise index entity and the node attributes (the enterprise index entity includes the node attributes, and the enterprise index entity inherits the built-in attribute types of the knowledge graph).

In some embodiments, the attributes of the business entity may include at least one of: the system comprises a unique credit code of an enterprise, an affiliated industry, enterprise qualification, an associated policy label, an operation range, a detailed address, regional information, net assets, a production value, net profits, earnings, tax amount, the highest investment amount of a single project, equipment investment, research and development expenses, regional information, whether foreign trade exists and the like. The attributes of the business indicator entity may include at least one of: enterprise index identification, associated condition index, candidate value, etc.

In step S713, an association relationship of the enterprise index entity to the condition index entity is established in the form of an edge in the knowledge graph.

In the embodiment of the present application, when the association relationship from the condition index entity to the entity on the enterprise side is established in step S515, the association relationship from the condition index entity to the enterprise index entity may be established; in order to complement the association relationship between the enterprise index entity and the conditional index entity, in step S713, the embodiment of the present application may establish the association relationship between the enterprise index entity and the associated conditional index in the form of a side in the knowledge graph based on the associated conditional index indicated in the attribute of the enterprise index entity.

In an example, after extracting data phrases from a policy text, the embodiment of the present application may perform data extraction on the data phrases through a feature template, so that subjects in the data phrases are regarded as enterprise indexes needing to be associated, and numerical conditions in the data phrases are regarded as condition indexes of conditions. Thereby defining the associated condition indicators of the enterprise indicators (which can be defined in the attributes of the enterprise indicators) and the associated enterprise indicators of the condition indicators (which can be defined in the attributes of the condition indicators) based on the data phrases in the policy text, so as to realize that the enterprise indicators are associated to specific conditions through specific condition indicators. For example, if the data phrase in the policy text is "annual development cost reaches 500 ten thousand", the association relationship between the enterprise index of "annual development cost" and the condition index of "500 ten thousand or more" may be extracted.

In step S714, the association relationship between the business entity and the policy tag entity is established in the form of an edge in the knowledge graph.

Referring to fig. 8A, in the embodiment of the present application, associating the entity on the enterprise side and the entity on the policy side may be further implemented through association between the enterprise entity and the policy tag entity. In some embodiments, the embodiments of the present application may establish the association relationship from the business entity to the policy tag entity in the form of an edge in the knowledge graph based on the associated policy tag (multivalue) indicated in the attribute of the business entity.

In step S715, the missing paths and the confidence levels of the missing paths of the enterprise-side entity and the policy-side entity are determined.

After the association relationship between the condition indicator entity and the enterprise indicator entity and the association relationship between the enterprise entity and the policy tag entity are constructed in the form of an edge, the established association relationship between the entities is implemented based on existing data (for example, existing enterprise indicators), and the existing data may not completely represent the association relationship between the entity on the enterprise side and the entity on the policy side, which may result in the absence of the association relationship between the entity on the enterprise side and the entity on the policy side, so the embodiment of the present application needs to supplement the missing relationship between the entity on the enterprise side and the entity on the policy side. In the embodiment of the present application, a missing path between an entity on the enterprise side and an entity on the policy side may be constructed to supplement the missing relationship between the entity on the enterprise side and the entity on the policy side. In some embodiments, the missing path may be in the form of an entity in the knowledge-graph.

As an alternative implementation, the missing path between the entity on the enterprise side and the entity on the policy side may include: a missing path between a business entity and a policy-side entity, such as a conditional entity, and/or a policy tag entity, and/or a policy entity. In other possible implementations, the missing path between the enterprise-side entity and the policy-side entity may include: missing paths of the missing enterprise index entities and the missing condition index entities; for example, if there is a missing enterprise index determined based on existing data, the user is required to supplement the missing enterprise index, so that after an enterprise index entity corresponding to the missing enterprise index is constructed in the knowledge graph, a missing path with the conditional index entity is required to be supplemented based on the missing enterprise index entity. In one example, as shown in connection with fig. 8A, embodiments of the present application may supplement a missing path from a business entity to a conditional entity in a knowledge-graph, where the missing path may be an entity in the form of a CVT. Further, the embodiment of the present application may determine an association relationship between the missing path and the business entity (e.g., a multi-valued missing path CVT between the business entity and the missing path shown in fig. 8A), and an association relationship between the missing path and the conditional entity (e.g., a single-valued association condition between the missing path and the conditional entity shown in fig. 8A), so as to implement linking the business entity and the conditional entity.

In some embodiments, as the connection path connecting the entity on the enterprise side and the entity on the policy side, the condition index entity and the missing path may determine a confidence level for the connection path (e.g., determine a confidence level of the condition index entity, determine a confidence level of the missing path), where the confidence level may indicate a confidence level of a relationship between the enterprise side and the entity on the policy side where the connection path communicates. As shown in fig. 8A, in the embodiment of the present application, confidence levels are respectively determined for the missing path and the condition indicator entity, and are recorded in the attributes of the missing path and the condition indicator entity.

It should be noted that determining the missing path in step S715 is only an optional means for improving the comprehensiveness of the data, and the embodiment of the present application may not perform step S715.

In further embodiments, a business entity may have multiple composite attributes in addition to the common attributes, which may be represented by multiple entities in the form of CVTs. In one example, as shown in connection with fig. 8A, in addition to the common attributes P1 and P2, the entity of the enterprise may represent the information of the enterprise over multiple years, respectively, through the entity of the composite attribute F1 and the entity of the composite attribute F2 in the form of CVT.

As an example, taking a specific enterprise scenario as an example, fig. 8B exemplarily shows another example of the knowledge graph provided in the embodiment of the present application, which may be referred to, and the construction of the knowledge graph shown in fig. 8B may refer to the description of the corresponding part in the foregoing similarly, which is not described again here.

The embodiment of the application divides concepts related to enterprises and policies into entities represented by nodes, such as enterprise entities, indicator entities, policy main point entities, conditional entities and the like, saves information of the entities by node attributes, represents various relationships (such as associated enterprise indicators, associated condition indicators and the like) between the entities by edges, and describes types of the relationships by the attributes of the edges, so that data of the enterprises and the policies can be stored and described in a knowledge graph in a mesh structure instead of a list-column structure of a table, the expandability of data storage of the enterprises and the policies is improved, and support is provided for supporting high-performance and quick associated query of the enterprises and the policies.

The embodiment of the present application further provides a data query method, which can be implemented based on the data storage method provided in the embodiment of the present application, for example, after the data storage method provided in the embodiment of the present application stores the data of the target and the object by using the knowledge graph, the embodiment of the present application can implement a target associated with the query object, an object associated with the object, a target associated with the target, and the like in the knowledge graph. In some embodiments, the server may obtain a data query request sent by the terminal device; calling a knowledge graph (the construction process of the knowledge graph can be realized based on the data storage method provided by the embodiment of the application, and the specific content can refer to the description of the corresponding parts in the foregoing); determining a data query result based on the knowledge-graph.

As an alternative implementation, taking the target of query object association as an example, fig. 9 exemplarily shows an alternative flowchart of the data query method provided by the embodiment of the present application. The method flow may be implemented by a server, and referring to fig. 9, the method flow may include the following steps.

In step S910, a data query request for querying a target associated with the first object is obtained.

The data query request may be sent to the server by the terminal device, for example, when a user of the terminal device needs to query a target that the first object conforms to (for example, query a policy that an enterprise conforms to), the data query request may be sent to the server by the terminal device to request to query a target that the first object is associated with. The first object may be a user-specified object, such as a business where the user is located or a specified business, etc.

In step S911, a knowledge map is called.

The server may store data of the target and the object through the knowledge graph, and the manner of storing data through the knowledge graph may refer to the description of the corresponding parts, which is not described herein again. And after the server acquires the data query request sent by the terminal equipment, the knowledge graph of the data of the storage target and the object can be taken out in an adjustable mode.

In step S912, a conditional entity and a conditional index entity associated with the first object are determined in the knowledge-graph.

In the case that the knowledge graph stores data of a target and an object, and records an association relationship between the target and the object, based on a first object needing to be queried, the embodiment of the application may determine a condition entity associated with the first object and a condition index entity from the knowledge graph. In some further embodiments, the target tag entity, the object entity, and the association relationship between the target tag entity and the object entity are provided in the knowledge graph, and in the embodiments of the present application, the associated target tag entity is determined in the knowledge graph through the object entity of the first object, so as to determine the condition entity associated with the target tag entity, and the condition indicator entity associated with the condition entity, thereby obtaining the condition entity and the condition indicator entity associated with the first object.

In an implementation example, a policy that an enterprise conforms to may be queried, in which case, the first object may be a first enterprise to be queried, and after the server obtains the data query request, the server may determine, in the knowledge graph, an associated policy tag entity from an enterprise entity corresponding to the first enterprise, and then determine a condition entity associated with the policy tag entity and a condition indicator entity associated with the condition entity, so as to determine, in the knowledge graph, the condition entity and the condition indicator entity associated with the first enterprise to be queried.

In step S913, a matching degree of the first object with the associated condition index entity is determined according to the object index of the first object and the rule defined by the condition index entity associated with the first object.

After determining the condition entity and the condition index entity associated with the first object, embodiments of the present application may determine a matching degree of the first object and the associated condition index entity based on the object index of the first object and a rule defined by the associated condition index entity. In some embodiments, the condition index entity associated with the first object may define a regular expression, and the embodiment of the present application may bring the object index of the first object into the regular expression to determine a matching degree of the object index of the first object and the regular expression, so as to obtain a matching degree of the first object and the associated condition index entity.

As an optional implementation, the regular expression defined by the condition indicator entity may include operators of multiple operators, and the operators perform different operator calculations by using the operands as parameters and using codes; an operator can be understood as a function, such as "include/greater than/less than/equal to/sum/average" or the like. Based on this, the object index of the first object is brought into the regular expression in the embodiment of the application, and then the object index can be used as a parameter to be calculated, so that the matching degree of the object index and the regular expression is obtained.

In step S914, a matching target of the first object is determined from a plurality of targets according to the matching degree, and the first object has at least one associated conditional index entity under any target.

In step S915, a data query result is derived according to the determined target.

After the matching degree of the first object and the associated conditional index entities is obtained, data of a plurality of targets can be stored based on the knowledge graph, and the first object can have at least one associated conditional index entity under any target; therefore, for any target, the matching degree of the first object in the condition index entity associated with the target can be determined, and the matching degree of the first object in the target is obtained by combining the matching degree of the first object in the condition index entity associated with the target; furthermore, the matching degree of the first object in each of the multiple targets can be obtained, and according to the matching degree of the first object in the multiple targets, the embodiment of the application can determine the target matched with the first object from the multiple targets, and obtain the data query result based on the determined target and feed the data query result back to the terminal device. In some embodiments, the multiple targets may be ranked according to the matching degree of the first object in the multiple targets, so that a set number of targets ranked in the top are determined as targets matched by the first object (for example, a target with the highest matching degree ranked in the top is determined as a target matched by the first object), and thus the determined targets are taken as targets matched and associated by the first object, carried in the data query result, and fed back to the terminal device.

In an example, after determining a condition entity and a condition index entity associated with a first enterprise to be queried from a knowledge graph, the embodiment of the present application may bring enterprise indexes of the first enterprise into a regular expression of the associated condition index entity, so as to determine a matching degree between the first enterprise and the associated condition index entity; based on that the first enterprise has at least one associated condition index entity under a plurality of policies, the embodiment of the application can determine the matching degree of the first enterprise under each policy according to the matching degree of the condition index entity associated with the first enterprise under each policy, so that one or more policies with the highest matching degree are used as the policies which are met and associated with the first enterprise, and the policies which are met and associated with the first enterprise are loaded in the data query result and fed back to the terminal device.

In further embodiments, the embodiments of the present application may also implement, through a knowledge graph, an object-to-object association query (e.g., an enterprise-to-enterprise association query), a target-to-target association query (e.g., a policy-to-policy association query), and the implementation principle is similar to that described above, and may be extended. As an alternative implementation, the embodiment of the present application may determine, from a business entity of an enterprise, a policy-side entity corresponding to a specific type of industry tag, area tag, and policy tag in a knowledge graph, and then reversely find a business entity of another related enterprise from the policy-side entity, so as to search for the related enterprise through the specific industry, the specific area, and the specific policy tag.

According to the embodiment of the application, elements such as objects (such as enterprises), targets (such as policies), conditions and target labels (such as policy labels) are formed into the mesh association in a mode that the nodes of the knowledge graph represent the entities and the edges represent the relationships among the entities, and the storage mode of rows and columns of a table is replaced by the relational database, so that the query cost of high-frequency association query in an intelligent scene is reduced from full-table multi-time cross multiplication to query of specific edges and nodes of the knowledge graph, and the performance and the speed of the association query are improved. Meanwhile, when the data storage is expanded, the expansion of nodes, edges and attributes in the knowledge graph can be realized without increasing rows and columns of a table, the expansion of the data storage can be simply and efficiently realized, and the expansibility of the data storage is improved.

In some further embodiments, the object index provided by the object may be supported, and after the object provides the object index, the embodiment of the present application may automatically query the object for the associated target or the associated object based on the foregoing knowledge graph scheme, and feed back the query result to the object, so that the object can obtain the recommended associated target or the associated object. For example, after an enterprise provides enterprise indexes, the embodiment of the application may automatically query the associated policies for the enterprise based on the association relationship between the enterprise and the policies stored in the knowledge graph, and feed the query result back to the enterprise.

As an alternative implementation, fig. 10A illustrates another alternative flowchart of a data query method provided in an embodiment of the present application. The method flow can be implemented by the terminal device, for example, when the first object queries the associated object, the terminal device of the first object can implement the method flow. As shown in fig. 10A, the method flow may include the following steps.

In step S101, an object index submission page is displayed, where a plurality of object index items are displayed.

When the first object queries the associated target or the associated object, the terminal device of the first object can display the object index submission page, so that the first object provides the object index of the first object. In some embodiments, the object index submission page may present a plurality of object index items, and an object index item may correspond to an object index. As an optional implementation, the object indexes corresponding to the part of the object index items in the object index submission page may be automatically determined based on the server, for example, the server may automatically determine the object indexes corresponding to the part of the object index items based on object data related to the first object in the database; some object index items possibly exist in the plurality of object index items, the object index cannot be automatically determined, and the first object is required to be manually input. In other possible implementations, the object metrics corresponding to the plurality of object metric items of the object metric submission page may each be input by the first object (e.g., none of the object metrics of the plurality of object metric items can be automatically determined by the server). In another possible implementation, the plurality of object index items of the object index submission page may all be object indexes that have been automatically filled in by the server, but the first object is required to confirm whether the object indexes filled in by the object index items are correct.

In an example, fig. 10B exemplarily shows an optional example diagram of an object index submission page, as shown in fig. 10B, the object index submission page may be an enterprise index submission page showing a plurality of enterprise index items to be filled in the enterprise index (the enterprise index items may be regarded as optional examples of the object index items), where the enterprise index corresponding to a part of the enterprise index items is automatically determined by the server based on the enterprise data entered in the enterprise database and the government database, and the enterprise index corresponding to another part of the enterprise index items needs to be filled in by the enterprise itself because there is no data base.

In step S102, object indexes of the first object corresponding to the plurality of object index items are determined, and a data query request is sent based on the object indexes corresponding to the plurality of object index items.

In some embodiments, if some or all of the object index items exist in the plurality of object index items, and the object index cannot be automatically determined, the embodiment of the present application may implement determining the object index of the first object corresponding to the plurality of object index items after the first object manually fills the object index of the some or all of the object index items and confirms submission of the object index of each object index item by the first object. In other embodiments, if a plurality of object index items are all automatically determined by the server, the embodiments of the present application may implement determining the object index of the first object corresponding to the plurality of object index items after the first object confirms and submits the object index of each object index item.

As an optional implementation, based on the first object confirming the object indexes corresponding to the submitted multiple object index items, the terminal device may send a data query request to the server to request the server to query and recommend a target associated with the first object or an associated object.

In one example, as shown in fig. 10B, after the enterprise fills the enterprise index for the enterprise index items that have not filled the enterprise index, if the terminal device detects that the confirmation button is clicked, the terminal device may determine that the enterprise confirms submission of the enterprise index corresponding to each enterprise index item, so that the terminal device may send a query request for querying the associated policy of the enterprise or associating the enterprise to the server.

In step S103, a data query result determined based on the knowledge graph is acquired, and the data query result is displayed on a data result page.

After the server obtains the data query request sent by the terminal device of the first object, the server can construct a knowledge graph based on the data storage method described above, and store the associated target or the associated object of the first object by using the constructed knowledge graph, so that the server can determine the associated target or the associated object of the first object based on the constructed knowledge graph, and further obtain a data query result.

The server can feed back the data query result to the terminal equipment, so that the terminal equipment can obtain the data query result determined based on the knowledge graph. The terminal device may display a data result page on which the data query result is presented so that the first object knows the associated target or the associated object.

In one example, FIG. 10C illustrates an example diagram of a data results page, which may be referenced. After the enterprise submits the enterprise indexes corresponding to the enterprise index items through the enterprise index submission page shown in fig. 10B, the server may query the policy associated with the enterprise based on the knowledge graph and feed back the policy associated with the enterprise to the terminal device of the enterprise, so that the terminal device of the enterprise may display the data result page shown in fig. 10C to display the policy associated with the enterprise, thereby implementing recommendation of the associated policy to the enterprise.

Further, based on different query requirements of the first object, for example, different query requirements such as querying a related target or a related object, the terminal device of the first object may send a data query request adapted to the query requirements to the server when the first object submits object indexes of the plurality of object index items, so that the server feeds back corresponding data query results and displays the data query results on a data result page, thereby satisfying different requirements of the first object for querying the related target or the related object.

When the data of the target and the object are stored, the flexibility and the expansibility of knowledge graph modeling are fully utilized, the unification of multi-element heterogeneous data is facilitated, and the software development cost is reduced. And the complex association of elements such as enterprises, policies, conditions, policy labels and the like is described through a mesh structure, so that the problem of linked list calculation amount of a relational database is solved, and the low-delay association query service is provided. Furthermore, the network structure adopting the knowledge graph can also fully utilize graph algorithm theoretical knowledge (such as a shortest path algorithm), and can more effectively extract data value. For example, enterprise and policy data are precipitated into the knowledge graph, so that the multivariate knowledge ecology built in the knowledge graph can be effectively utilized, and the multivariate knowledge ecology and other data jointly play a role in making a resultant force, thereby generating a higher value.

In the following, the data storage device provided by the embodiment of the present application is described, the data storage device described below may be regarded as a functional module that is required by a server to implement the data storage method provided by the embodiment of the present application, and the contents of the device described below may be referred to in correspondence with the contents described above.

As an alternative implementation, fig. 11 is a block diagram schematically illustrating a data storage apparatus provided in an embodiment of the present application, where the apparatus is applicable to a server, and as shown in fig. 11, the apparatus may include:

the data acquisition module 011 is used for acquiring a target text defining a target and object data of the target;

a condition and condition index obtaining module 012, configured to identify at least a condition corresponding to the target and a condition index associated with the condition from the target text; the condition index defines a related object index of the object and the condition and a rule which needs to be met by the related object index;

a target index obtaining module 013, configured to obtain a target index of the target from the target data;

an entity construction module 014, configured to construct at least a condition entity corresponding to the condition, a condition index entity corresponding to the condition index, and an object index entity corresponding to the object index in a knowledge graph;

the relationship building module 015 is configured to build an association relationship between a condition entity and a condition index entity in a knowledge graph, and build an association relationship between the condition index entity and an object index entity based on an associated object index defined by the condition index.

In some embodiments, the target comprises at least one sub-target; a sub-target is associated with at least one target label, and one target label is label information of a content abstraction point of the sub-target; one target tag is associated with at least one condition and one condition is associated with at least one condition indicator.

In some embodiments, the condition and condition index obtaining module 012 is configured to identify, from at least the target text, a condition corresponding to the target, and the condition index associated with the condition includes:

coding the target text to obtain text characteristics of the target text;

extracting text features corresponding to the sub-targets from the text features;

decoding the text features corresponding to the sub-targets, identifying text phrases of the label categories from decoding results, standardizing the text phrases and refining the text phrases into target labels;

and identifying a data phrase from the decoding result, and extracting a condition under the target label and a condition index under the condition from the data phrase.

In some embodiments, the condition and condition index obtaining module 012 is configured to identify a data phrase from the decoding result, and extract the condition under the target tag and the condition index under the condition from the data phrase, where the condition and condition index include:

and identifying the numerical phrases from the decoding result, and performing characteristic extraction on the numerical phrases through a characteristic template to obtain conditions and condition indexes of the numerical phrases.

In some further embodiments, the apparatus may be further configured to: constructing a target entity corresponding to the target, a sub-target entity corresponding to the sub-target and a target label entity corresponding to the target label in the knowledge graph, and constructing an association relationship between the target entity and the sub-target entity, an association relationship between the sub-target entity and the target label entity and an association relationship between the target label entity and a condition entity;

constructing an object entity corresponding to the object in the knowledge graph, storing the attributes of the object entity and the object index entity through node attributes, and constructing an association relationship between the object entity, the object index entity and the node attributes;

and constructing an incidence relation between the object entity and the target tag entity in the knowledge graph.

In some further embodiments, the apparatus may be further configured to: determining a missing path between entities in the knowledge-graph, wherein the missing path comprises a missing path of an object entity and a condition entity, and/or a target tag entity, and/or a target entity, or a missing path of a missing object index entity and a missing condition index entity;

determining a confidence level of a connection path, the connection path comprising the missing path, and a condition indicator entity.

In some embodiments, the relationship building module 015, configured to build, based on the associated object indicator defined by the condition indicator, an association relationship between the condition indicator entity and the object indicator entity, where the association relationship includes:

constructing an incidence relation from the condition index entity to the object index entity in a knowledge graph according to the incidence object index indicated in the attribute of the condition index;

and constructing the association relation from the object index entity to the condition index entity in a knowledge graph according to the association condition index indicated in the attribute of the object index entity.

In some embodiments, in a government scenario, the goal comprises a policy, the goal text comprises a policy text, the sub-goals comprise key points of policies under the policy, the goal tags comprise policy tags under key points of policies, the object comprises a business, and the object metrics comprise business metrics.

An embodiment of the present application further provides a data query device, where the data query device is configured to: acquiring a data query request; calling a knowledge graph, wherein the knowledge graph stores data based on the data storage method provided by the embodiment of the application; determining a data query result based on the knowledge-graph.

In some embodiments, the data query request is for querying a target associated with the first object; the data query device is used for determining a data query result based on the knowledge graph and comprises the following steps:

determining a condition entity and a condition indicator entity associated with the first object in the knowledge-graph;

determining the matching degree of the first object and the associated condition index entity according to the object index of the first object and a rule which needs to be met by the associated object index defined by the condition index entity associated with the first object;

determining a target matched with the first object from a plurality of targets according to the matching degree, wherein the first object has at least one associated conditional index entity under any target;

and obtaining a data query result according to the determined target.

The embodiment of the present application further provides a server, where the server may implement the data storage method provided in the embodiment of the present application through the data storage device, and/or implement the data query method provided in the embodiment of the present application through the data query device.

As an alternative implementation, fig. 12 shows an alternative block diagram of a server. As shown in fig. 12, the server may include: at least one processor 1, at least one communication interface 2, at least one memory 3 and at least one communication bus 4.

In the embodiment of the present application, the number of the processor 1, the communication interface 2, the memory 3, and the communication bus 4 is at least one, and the processor 1, the communication interface 2, and the memory 3 complete mutual communication through the communication bus 4.

Alternatively, the communication interface 2 may be an interface of a communication module for performing network communication.

Alternatively, the processor 1 may be a CPU, a GPU (Graphics Processing Unit), an NPU (embedded neural network processor), an FPGA (Field Programmable Gate Array), a TPU (tensor Processing Unit), an AI chip, an asic (application Specific Integrated circuit), or one or more multi-valued Integrated circuits configured to implement the embodiments of the present application. The memory 3 may comprise a high-speed RAM memory and may also comprise a non-volatile memory, such as at least one disk memory. The memory 3 stores one or more computer-executable instructions, and the processor 1 calls the one or more computer-executable instructions to execute the data storage method provided by the embodiment of the present application and/or the data query method provided by the embodiment of the present application.

The embodiment of the present application further provides a storage medium, where the storage medium stores one or more computer-executable instructions, and the one or more computer-executable instructions, when executed, implement the data storage method provided in the embodiment of the present application and/or the data query method provided in the embodiment of the present application.

The embodiments of the present application further provide a computer program, where the computer program is executed to implement the data storage method provided in the embodiments of the present application and/or the data query method provided in the embodiments of the present application.

While the embodiments of the present application have been described above as providing multiple-valued embodiments, the various alternatives described in the embodiments can be combined and cross-referenced without conflict, thereby extending the variety of possible embodiments that can be considered as disclosed and disclosed in the embodiments of the present application.

Although the embodiments of the present application are disclosed above, the present application is not limited thereto. Various changes and modifications may be effected by one skilled in the art without departing from the spirit and scope of the application, and the scope of protection is defined by the claims.

Claims

1. A method of data storage, comprising:

acquiring a target text defining a target and object data of the target;

obtaining an object index of the object from the object data;

2. The method of claim 1, wherein the target comprises at least one sub-target; a sub-target is associated with at least one target label, and one target label is label information of a content abstraction point of the sub-target; one target tag is associated with at least one condition and one condition is associated with at least one condition indicator.

3. The method of claim 2, wherein the identifying at least a condition corresponding to the target from the target text, and the condition indicator associated with the condition comprises:

coding the target text to obtain text characteristics of the target text;

4. The method of claim 3, wherein the identifying the data phrase from the decoding result and extracting the condition under the target label and the condition index under the condition from the data phrase comprises:

5. The method of claim 2, further comprising:

constructing a target entity corresponding to the target, a sub-target entity corresponding to the sub-target and a target tag entity corresponding to the target tag in the knowledge graph, and constructing an association relationship between the target entity and the sub-target entity, an association relationship between the sub-target entity and the target tag entity and an association relationship between the target tag entity and a condition entity;

6. The method of claim 5, further comprising: determining a missing path between entities in the knowledge-graph, wherein the missing path comprises a missing path of an object entity and a condition entity, and/or a target tag entity, and/or a target entity, or a missing path of a missing object index entity and a missing condition index entity;

7. The method of claim 1, wherein the constructing the association of the condition index entity with an object index entity based on the associated object index defined by the condition index comprises:

8. The method of any of claims 2-6, wherein the goal comprises a policy, the goal text comprises a policy text, the sub-goal comprises a policy focus under the policy, the goal tag comprises a policy tag under the policy focus, the object comprises a business, and the object indicator comprises a business indicator.

9. A data query method, comprising:

acquiring a data query request;

retrieving a knowledge graph storing data based on the data storage method of any one of claims 1-8;

determining a data query result based on the knowledge-graph.

10. The method of claim 9, wherein the data query request is for querying a target associated with a first object; the determining a data query result based on the knowledge-graph comprises:

and obtaining a data query result according to the determined target.

11. A data query method, comprising:

acquiring a data query result determined based on a knowledge graph, and displaying the data query result on a data result page; wherein the knowledge-graph stores data based on the data storage method of any one of claims 1-8.

12. A server comprising at least one memory storing one or more computer-executable instructions and at least one processor invoking the one or more computer-executable instructions to perform a data storage method according to any one of claims 1-8 and/or a data query method according to any one of claims 9-10.

13. A storage medium, wherein the storage medium stores one or more computer-executable instructions that, when executed, implement a data storage method as claimed in any one of claims 1-8, or a data query method as claimed in any one of claims 9-10, or a data query method as claimed in claim 11.