CN114860916A

CN114860916A - Knowledge retrieval method and device

Info

Publication number: CN114860916A
Application number: CN202210643778.9A
Authority: CN
Inventors: 牛任恺; 张艳丽; 郭伟; 张鑫磊; 刘继鹏; 王利赛; 岳虎; 王玉君; 余红; 王凯; 石磊; 赵志刚; 毕明; 郑思达; 郭皎; 王亚超; 王磊; 李文慧; 路鑫; 妙红英
Original assignee: State Grid Corp of China SGCC; State Grid Jibei Electric Power Co Ltd
Current assignee: State Grid Corp of China SGCC; State Grid Jibei Electric Power Co Ltd
Priority date: 2022-06-09
Filing date: 2022-06-09
Publication date: 2022-08-05

Abstract

The application provides a knowledge retrieval method and a knowledge retrieval device, which can be used in the field of electric power marketing inspection or other fields, and the method comprises the following steps: acquiring management and control knowledge of batch power marketing business; constructing a target knowledge graph according to the batch power marketing service control knowledge, wherein nodes in the target knowledge graph represent entities in the batch power marketing service control knowledge, and a directional edge between two nodes represents an incidence relation between the entities; and receiving a knowledge retrieval request, and applying a preset retrieval rule and the target knowledge graph to obtain a retrieval result corresponding to the knowledge retrieval request. The method and the device can improve accuracy and efficiency of knowledge retrieval, and further can improve reliability and efficiency of electric power marketing inspection.

Description

Knowledge retrieval method and device

Technical Field

The application relates to the technical field of electric power marketing inspection, in particular to a knowledge retrieval method and a knowledge retrieval device.

Background

The electric power marketing inspection means that a power supply enterprise monitors and checks the operation management work of an electric power marketing link in order to block the management loophole, prevent operation loss, further complete and perfect a self-constraint mechanism, promote the standardization and systematization of marketing work and ensure the normal operation of a marketing system; the basic task is to check and supervise the construction, marketing work quality, behavior regulation and the like of the relevant marketing system of the units and individuals engaged in the electric power marketing work according to relevant laws, administrative laws, policies and relevant regulations and regulations of enterprises.

However, relevant laws and regulations at the present stage are distributed in documents such as marketing inspection management and control business specifications, power laws and regulations, industry standards and the like, and when relevant knowledge needs to be searched and quoted in marketing inspection work, the relevant knowledge can only be browsed and inquired one by one, which brings great inconvenience to marketing inspection workers to retrieve professional knowledge and seriously restricts inspection work efficiency.

Disclosure of Invention

Aiming at the problems in the prior art, the knowledge retrieval method and the knowledge retrieval device are provided, so that the accuracy and the efficiency of knowledge retrieval can be improved, and further, the reliability and the efficiency of electric power marketing inspection can be improved.

In order to solve the technical problem, the present application provides the following technical solutions:

in a first aspect, the present application provides a knowledge retrieval method, including:

acquiring management and control knowledge of batch power marketing business;

constructing a target knowledge graph according to the batch power marketing service control knowledge, wherein nodes in the target knowledge graph represent entities in the batch power marketing service control knowledge, and a directional edge between two nodes represents an incidence relation between the entities;

and receiving a knowledge retrieval request, and applying a preset retrieval rule and the target knowledge graph to obtain a retrieval result corresponding to the knowledge retrieval request.

Further, the constructing a target knowledge graph according to the batch power marketing service management and control knowledge includes:

extracting the batch entities from the batch electric power marketing business control knowledge according to a preset entity extraction model;

obtaining the target knowledge graph according to the batch of entities and a pre-constructed graph knowledge layer;

the preset entity extraction model is a Bert-Bilstm-CRF model obtained through pre-training.

Further, the obtaining a retrieval result corresponding to the knowledge retrieval request by applying a preset retrieval rule and the target knowledge graph includes:

determining a to-be-processed entity corresponding to the knowledge retrieval request according to a preset mapping relation table of search contents and entities, and obtaining a query statement corresponding to the to-be-processed entity based on a preset power marketing business rule;

executing the query statement, and determining a plurality of entity combinations corresponding to the query statement, wherein each entity combination comprises: at least one entity and its corresponding attribute information;

and sequencing the entity combinations according to a preset intelligent recommendation model to obtain a retrieval result corresponding to the knowledge retrieval request, and outputting and displaying the retrieval result.

Further, the obtaining the target knowledge graph according to the batch entities and the pre-constructed graph knowledge layer comprises:

obtaining an intermediate knowledge graph according to the batch entities and a pre-constructed graph knowledge layer;

and carrying out entity disambiguation and entity alignment treatment on the intermediate knowledge graph to obtain the target knowledge graph.

In a second aspect, the present application provides a knowledge retrieval apparatus comprising:

the acquisition module is used for acquiring the management and control knowledge of the batch power marketing business;

the system comprises a construction module, a data processing module and a data processing module, wherein the construction module is used for constructing a target knowledge graph according to batch power marketing service control knowledge, nodes in the target knowledge graph represent entities in the batch power marketing service control knowledge, and directional edges between the two nodes represent incidence relations between the entities;

and the retrieval module is used for receiving a knowledge retrieval request, and obtaining a retrieval result corresponding to the knowledge retrieval request by applying a preset retrieval rule and the target knowledge map.

Further, the building module includes:

the extraction unit is used for extracting batch entities from the batch electric power marketing business management and control knowledge according to a preset entity extraction model;

the construction unit is used for obtaining the target knowledge graph according to the batch of entities and a pre-constructed graph knowledge layer;

Further, the retrieval module includes:

the determining unit is used for determining the entity to be processed corresponding to the knowledge retrieval request according to a preset search content and entity mapping relation table, and obtaining a query statement corresponding to the entity to be processed based on a preset power marketing business rule;

an execution unit, configured to execute the query statement, and determine a plurality of entity combinations corresponding to the query statement, where each entity combination includes: at least one entity and its corresponding attribute information;

and the obtaining unit is used for sequencing the entity combinations according to a preset intelligent recommendation model, obtaining a retrieval result corresponding to the knowledge retrieval request and outputting and displaying the retrieval result.

Further, the building unit is configured to:

In a third aspect, the present application provides an electronic device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor implements the knowledge retrieval method when executing the program.

In a fourth aspect, the present application provides a computer-readable storage medium having stored thereon computer instructions that, when executed, implement the knowledge retrieval method.

According to the technical scheme, the application provides a knowledge retrieval method and a knowledge retrieval device. Wherein, the method comprises the following steps: acquiring management and control knowledge of batch power marketing business; constructing a target knowledge graph according to the batch power marketing service control knowledge, wherein nodes in the target knowledge graph represent entities in the batch power marketing service control knowledge, and a directional edge between two nodes represents an incidence relation between the entities; receiving a knowledge retrieval request, and applying a preset retrieval rule and the target knowledge map to obtain a retrieval result corresponding to the knowledge retrieval request, so that the accuracy and efficiency of knowledge retrieval can be improved, and the reliability and efficiency of electric power marketing inspection can be further improved; specifically, the method can assist the user in completing professional knowledge retrieval and inspection theme creation, save the working time of inspection workers, reduce the complexity of knowledge acquisition and reference, and reduce the cost.

Drawings

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

FIG. 1 is a schematic flow chart diagram of a knowledge retrieval method in an embodiment of the present application;

FIG. 2 is a schematic diagram of a target knowledge graph in one example of the present application;

FIG. 3 is a diagram of a search results page in an example of the present application;

FIG. 4 is a flow chart

illustrating steps

401 and 402 of a knowledge retrieval method in an embodiment of the present application;

FIG. 5 is a flow chart illustrating steps 501-503 of a knowledge retrieval method in an embodiment of the present application;

FIG. 6 is a schematic diagram of a target knowledge graph in another example of the present application;

FIG. 7 is a schematic structural diagram of a knowledge retrieval device in an embodiment of the present application;

fig. 8 is a schematic block diagram of a system configuration of an electronic device according to an embodiment of the present application.

Detailed Description

In order to make those skilled in the art better understand the technical solutions in the present specification, 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 order to solve the problems in the prior art, the scheme provides a method and a device for retrieving electric power marketing management and control rules based on a knowledge graph, wherein the method comprises the steps of constructing an inspection business management and control knowledge graph, mining and associating business rules which are dispersedly split in a marketing inspection data set and correction measures in a historical worksheet data set, forming a complete marketing business knowledge system, supporting knowledge learning and retrieval requirements in inspection work, and enabling the marketing business management and control rules to be equivalent to electric power marketing business management and control knowledge; based on the intelligent retrieval technology of the knowledge graph controlled by the inspection service, semantic analysis and conversion are carried out on the knowledge query input of the inspectors, semantic expressions which can be understood and executed by the knowledge graph are generated, entity link reasoning research is carried out aiming at the incidence relation among knowledge bodies such as inspection subjects, service items, knowledge points, clauses and the like, knowledge recommendation weight is defined according to cosine distance between entities, and knowledge retrieval is carried out; dynamically analyzing the input inspection target and service knowledge point, reading the inspection service control knowledge map through the NLP processes of semantic understanding, slot position identification, constraint identification and the like, and feeding back service control rules, service knowledge, related service files and other contents for the user.

Based on this, in order to improve accuracy and efficiency of knowledge retrieval and further improve reliability and efficiency of electric power marketing inspection, the embodiment of the present application provides a knowledge retrieval device, which may be a server or a client device, where the client device may include a smart phone, a tablet electronic device, a network set-top box, a portable computer, a desktop computer, a Personal Digital Assistant (PDA), a vehicle-mounted device, an intelligent wearable device, and the like. Wherein, intelligence wearing equipment can include intelligent glasses, intelligent wrist-watch and intelligent bracelet etc..

In practical applications, the part for knowledge retrieval may be performed on the server side as described above, or all operations may be performed in the client device. The selection may be specifically performed according to the processing capability of the client device, the limitation of the user usage scenario, and the like. This is not a limitation of the present application. The client device may further include a processor if all operations are performed in the client device.

The client device may have a communication module (i.e., a communication unit), and may be communicatively connected to a remote server to implement data transmission with the server. The server may include a server on the task scheduling center side, and in other implementation scenarios, the server may also include a server on an intermediate platform, for example, a server on a third-party server platform that has a communication link with the task scheduling center server. The server may include a single computer device, or may include a server cluster formed by a plurality of servers, or a server structure of a distributed apparatus.

The server and the client device may communicate using any suitable network protocol, including network protocols not yet developed at the filing date of this application. The network protocol may include, for example, a TCP/IP protocol, a UDP/IP protocol, an HTTP protocol, an HTTPS protocol, or the like. Of course, the network Protocol may also include, for example, an RPC Protocol (Remote Procedure Call Protocol), a REST Protocol (Representational State Transfer Protocol), and the like used above the above Protocol.

It should be noted that the knowledge retrieval method and apparatus disclosed in the present application can be used in the technical field of power marketing inspection, and can also be used in any field except the technical field of power marketing inspection.

The following examples are intended to illustrate the details.

In order to improve the accuracy and efficiency of knowledge retrieval and further improve the reliability and efficiency of power marketing inspection, the present embodiment provides a knowledge retrieval method in which the executing subject is a knowledge retrieval device, the knowledge retrieval device includes but is not limited to a server, as shown in fig. 1, the method specifically includes the following contents:

step 101: and acquiring the management and control knowledge of the batch power marketing business.

Specifically, the batch power marketing business management and control knowledge may include: business documents and historical inspection work orders of enterprises related to the power grid, and the like.

Step 102: and constructing a target knowledge graph according to the batch power marketing service control knowledge, wherein nodes in the target knowledge graph represent entities in the batch power marketing service control knowledge, and a directional edge between the two nodes represents an association relation between the entities.

Specifically, the entities may include: the method comprises the following steps of (1) checking the inspection subject, abnormal reasons, rectification measures and the like in a work order by using a clause name, service items, file names to which the clauses belong, knowledge points in the clauses; the association relationship may include: the method comprises the following steps of containing, corresponding, generating, relating, covering, relating, leading, descending, setting, basis and the like; the entity in the target knowledge graph may be provided with corresponding attribute information, where the attribute information may be entity content, or may be used to explain the entity, and the like, for example, a knowledge point: the 'electric charge clearing and quitting' attribute is as follows: the' clearing of the electric charge refers to that a user holds valid identity documents and related evidence materials to a business hall in charge to handle a charge returning procedure under the condition that the user does not have arrears and the electric charge account has a balance, and a power supply enterprise processes the charges according to charge returning management rules; knowledge points are as follows: the low-voltage non-resident has the following properties: "commercial electricity consumption, special industrial electricity consumption, and other illumination electricity consumption, as well as air conditioner and electric heater electricity consumption; the attribute information of the clause may be the content of the clause, and the attribute information of the service file may be the entire content of the service file.

For example, a power marketing business control knowledge is a file of "xxx company electric charge checking and charging management method", the file includes: the thirty-fifth article, the article content is: thirty-fifth article strengthens error management of electric quantity and electricity charge, standardizes a refund process, needs to refund the electric quantity and electricity charge due to meter reading errors, errors of charging parameters, faults of metering devices, default electricity consumption, electricity stealing and the like, initiates an electric quantity and electricity charge refund process in a marketing business application system by a responsibility department, shows the refund reason, calculates the process and uploads related data, and the marketing business application system needs to set an electricity charge refund approval link, and finishes refund approval, and issue by an accounting center (team) after stage-by-stage approval. "the entity in the electric marketing service management and control knowledge may include: the file name of the article: the management method for reading and checking the electric charge of xxx company, the article name: the thirty-fifth article, business matters: error backoff management, knowledge points in the article: meter reading errors, default electricity consumption and the like; the correlation relationship is as follows: the management method for reading and checking the electricity charge of the xxx company comprises thirty-fifth clauses, the thirty-fifth clauses comprise meter reading errors, illegal electricity utilization and the like, and the basis of error compensation management is the thirty-fifth clauses. The file also includes: the 'fifty-eighth article' is the article content, namely 'fifty-eighth article' executes the electric charge default fund system strictly according to the convention of the supply and utilization contract, so that the electric charge default fund cannot be reduced or avoided at will, the electric charge default fund cannot be used for charging the electric charge, and the maximum amount of the electric charge for the default fund cannot exceed 30 percent of the principal fund. The electric charge default fund caused by the following reasons can be approved and approved to implement the electric charge default fund non-collection; and (I) the power supply business personnel make meter reading errors or calculate the electric charge, so that the electric customers pay the electric charge on time. And (II) … … for non-power customer reasons. "the entity in the electric marketing service management and control knowledge may further include: the article names: fifty-eight articles, knowledge points in articles: meter reading errors, non-electric power customer reasons and the like, the basis of the electricity fee default fund is fifty-eight clauses, and the finally obtained corresponding knowledge graph can be shown in fig. 2.

Step 103: and receiving a knowledge retrieval request, and applying a preset retrieval rule and the target knowledge graph to obtain a retrieval result corresponding to the knowledge retrieval request.

Specifically, the knowledge retrieval request may include search contents, inspection targets or service knowledge points input by the user from the front end, such as "residential electricity grading standard" or "pricing rate".

As can be seen from the above description, the knowledge retrieval method provided in this embodiment can improve accuracy and efficiency of knowledge retrieval, and further improve reliability and efficiency of power marketing inspection, can perform semantic analysis on an inspection target or a service knowledge point input by a user, and display contents such as relevant legal provisions, service management methods, professional files, inspection subject rules and the like matched with the inspection target, thereby assisting the user in completing professional knowledge retrieval and inspection subject creation. The function can save the working time for the inspection staff and reduce the complexity of knowledge acquisition and reference; during the generation of the inspection scheme, the processing and the auditing of the inspection work order, the working efficiency can be obviously improved. The method is characterized in that files, articles, knowledge points, inspection subjects, typical cases, risk points, abnormal points, business items and the like are associated, the policy basis of business type analysis, processing and auditing is rapidly mastered through knowledge fuzzy query, the traceability analysis, comprehensive study and judgment and business reasoning capability of inspectors are greatly improved, the knowledge management cost can be saved, in one example, the copying and storage cost of each file in a company in a district under jurisdiction is about 50 yuan, 253 files are managed through the knowledge retrieval mode provided by the embodiment, and 1.265 ten thousand yuan can be saved in the estimated year.

In one example, the search content can be presented in the form of a knowledge question and answer by four modules of noun explanation, recommendation subject, related knowledge and full text retrieval; search content related to the volume price fee can be input in the front-end search box, and name explanations, recommended subjects, knowledge points, files and the like related to the search content are displayed after searching. The 'hot search list' of the front-end page can show 10 common hot search problems, and a user can directly click the corresponding problems and directly jump to the corresponding contents, so that the search link is omitted. As shown in fig. 3, the front page after searching for the question shows the answers in four modules, i.e., "synthesis", "recommended subject", "related knowledge", and "full text retrieval".

The "synthesis" module may contain all the contents of "noun interpretation", "recommended subject", "related knowledge", and "full text retrieval".

The "noun interpretation" may show the meaning of interpreting a specific knowledge or noun.

The "recommendation topic" module may display related audit topics recommended according to search contents, audit rules, and the like.

The "relevant knowledge" module may display relevant content of the search content within the knowledge-graph, the content being presented in an order of relevance of the files in the knowledge-graph.

The full-text retrieval module can display the search content to perform full-text retrieval in all document libraries, and the number of times of occurrence of a certain document is used as relevance display.

The right side of the front end page may show "noun recommendations", "frequently asked questions", "atlas visualizations", and "wrong knowledge feedback" related to the search content.

The noun recommendation shows noun recommendation related to search content, and the noun related knowledge can be inquired by clicking; for example, the search content is: the resident electricity grading standard, noun recommendation can show: the type of electricity consumption, school teaching and student life electricity consumption, resident life electricity consumption, electricity price system and the like.

The 'common questions' show some common questions related to the search content, and the click can directly jump to the corresponding query page.

The 'atlas visualization' displays a knowledge atlas visualization thumbnail related to search content, and clicks can skip to a visualization page of the search content.

The 'error knowledge feedback' is used for the user to feed back error content when the user searches knowledge, if the content recommended by the system is wrong or the searched content is inconsistent with the displayed content, the 'error knowledge feedback' button can be clicked. And prompting after the error feedback is successful: the addition is successful; if the search content system can not identify, prompting: not good meaning, the question you search for has not been answered for a while.

In another example, after receiving the knowledge retrieval request, the retrieval result may be presented in the form of a graph; clicking a screening button, screening the layer number, the entity and the relation of the visual content displayed on the current page, and after checking the corresponding content, only displaying the checked content on the page; clicking a refreshing button, and reloading the current page; and clicking a history button to inquire the name of the center point of the history display, and clicking the center point to quickly jump the content of the history center point.

To further improve the reliability of constructing the knowledge-graph and further improve the accuracy of knowledge retrieval by applying the reliable knowledge-graph, referring to fig. 4, in an embodiment of the present application, step 102 includes:

step 401: and extracting the batch entities from the batch electric power marketing business management and control knowledge according to a preset entity extraction model.

Step 402: obtaining the target knowledge graph according to the batch of entities and a pre-constructed graph knowledge layer; the preset entity extraction model is a Bert-Bilstm-CRF model obtained through pre-training.

Specifically, the pre-constructed schema of the map knowledge layer may be obtained by combing the expert experience, the industry dictionary, the knowledge base and other contents in an early stage based on the control knowledge of the batch power marketing service.

In order to further improve the retrieval efficiency, referring to fig. 5, in an embodiment of the present application, the applying a preset retrieval rule and the target knowledge graph in step 103 to obtain a retrieval result corresponding to the knowledge retrieval request includes:

step 501: and determining the entity to be processed corresponding to the knowledge retrieval request according to a preset mapping relation table of the search content and the entity, and obtaining a query statement corresponding to the entity to be processed based on a preset power marketing business rule.

Specifically, the preset mapping relationship table of search contents and entities may include: the mapping between the search content and the entities in the target knowledge-graph, for example, mapping the search content "two-system electricity prices" to the entities in the target knowledge-graph or mapping the search content "photovoltaic" to the entities in the target knowledge-graph "distributed photovoltaic", may be performed using a simple string matching approach.

Specifically, a large number of service rules can be injected in the process of constructing the SPARQL query by applying the nl2sql technology, the category grammar is combined to complete the tree node merging work from bottom to top, and finally, a root node, namely a semantic parsing result, is generated.

Examples are: and inquiring photovoltaic, and after mapping according to the electric power marketing professional vocabulary, associating the photovoltaic with distributed photovoltaic in the map, which is a root node, so that a SPARQL query is directly formed according to the mapping of the character strings, and information such as policy clause information, inspection subject information, relevant knowledge point recommendation and the like corresponding to the inquiry is output through a query result.

In an example, the preset power marketing business rule may be set as follows:

completing inquiry and retrieval of the inspection subject based on a depth-first search method; namely:

searching for content according to attribute information such as checking rules and checking points in the checking subject entity and the checking subject attribute, traversing the target knowledge graph based on a depth-first search method, searching whether the target knowledge graph has the checking subject matched with the knowledge search request, if so, returning the checking subject and the related attribute of the checking subject, wherein the related attribute of the checking subject can comprise: the method comprises the following steps of checking subject names, checking rules, checking points, data sources and access fields, checking methods, problems and modification standards, namely, the determined entity combination can comprise: "audit subject" and audit subject related attributes. The business rules of the query retrieval are independently executed, and the retrieval results participate in the next sequencing. For example, if an inspection theme is the same as the to-be-processed entity corresponding to the knowledge retrieval request, the inspection theme is determined to be the inspection theme matched with the knowledge retrieval request, and if the attribute information of the inspection theme contains the to-be-processed entity corresponding to the knowledge retrieval request, the inspection theme can also be determined to be the inspection theme matched with the knowledge retrieval request.

Secondly, based on a depth-first search method, the query and retrieval of 'business items' and 'knowledge points' are completed; namely:

and traversing the target knowledge graph based on a depth-first search method, and retrieving the service items and the knowledge points which are matched with the query statement in the target knowledge graph, wherein the weight of the service items is greater than that of the knowledge points. For example, if a business item is the same as the to-be-processed entity corresponding to the knowledge retrieval request, it is determined that the business item is the business item matching the knowledge retrieval request, and if the attribute information of the business item includes the to-be-processed entity corresponding to the knowledge retrieval request, it is also determined that the business item is the business item matching the knowledge retrieval request, and "knowledge points" are the same.

And (III) performing query retrieval again based on the query retrieval result of the previous step.

Step 1: and inquiring whether the target knowledge graph has the inspection subject in one-degree association or not based on the matched service items, and returning the combination of the service items and the inspection subject if the target knowledge graph has the inspection subject directly associated with the inspection subject.

Step 2: inquiring whether a 'clause' entity in one-degree association exists in the target knowledge graph or not based on the matched 'business item', if so, returning the combination of the 'business item' and the 'clause', namely the determined entity combination can comprise: "business events", "articles", and attribute information of each of the two.

And step 3: inquiring whether the entity of the risk point in the first-degree association exists in the target knowledge graph or not based on the matched business item, if so, returning the combination of the business item and the risk point, namely the determined entity combination can comprise: "business event", "risk point", and attribute information of each of the two.

And 4, step 4: inquiring whether an entity of the abnormal point in the first-degree association exists in the target knowledge graph or not based on the matched service item, if so, returning the combination of the service item and the abnormal point, namely the determined entity combination can comprise: "business transaction", "exception point", and attribute information of each of the two.

And 5: inquiring whether a service file in one-degree association exists in the target knowledge graph or not based on the inspection subjects, the clauses, the risk points and the abnormal points in the steps 1 to 4, if so, returning a service file entity and attribute information, namely, the determined entity combination can comprise: the "service file" entity and its attribute information.

Furthermore, the combination from the service items to the service files and the shortest path information can be returned, and the combination of the service items and the service files can be grouped and judged according to the query retrieval result in the step 5, so as to screen out the service file entity with the score reaching a certain threshold value.

Step 6: and performing one-time association based on the matched knowledge points to obtain the clause entity and the clause attribute information corresponding to the knowledge points.

Further, the entity of the article and the knowledge point can be further identified and scored, and the entity of the article with the score reaching a certain threshold value is screened out.

And 7: inquiring and retrieving the clause entity obtained in the step 6 to obtain a 'service file' corresponding to the clause, and grouping the clauses according to the 'service file', namely, the determined entity combination can comprise: the "article" entity and its attribute information.

Specifically, when the "articles" are grouped according to the "service file", the weight value of the "service file" may be the highest weight value of the "articles" protected by the "service file".

Specifically, the knowledge retrieval process is traversal search of the power marketing service control knowledge on a depth tree level, and the knowledge retrieval can be performed by adopting a depth-first search algorithm; Depth-First Search (DFS) is a process of searching from a vertex down each possible path until Depth is no longer possible, and each node can only be visited once; the specific description is as follows:

(1) and performing intention identification and confirmation on search contents in the knowledge retrieval request, constructing a Query statement, and finding an initial node A in the target knowledge graph, wherein the initial node A can be equivalent to the root node.

(2) And sequentially starting from the adjacent points which are not accessed by the A, and performing depth-first traversal on the target knowledge graph.

(3) If the node is not accessed, backtracking to the node and continuing to perform depth-first traversal.

(4) Until all nodes that communicate with vertex a path have been visited once.

Step 502: executing the query statement, and determining a plurality of entity combinations corresponding to the query statement, wherein each entity combination comprises: at least one entity and its corresponding attribute information.

Step 503: and sequencing the entity combinations according to a preset intelligent recommendation model to obtain a retrieval result corresponding to the knowledge retrieval request, and outputting and displaying the retrieval result.

Specifically, the retrieval of the clause content is emphasized, and the returned clause content is subjected to advanced sorting; fig. 6 is a schematic view of a target knowledge graph in an example of the application, where fig. 6 is only a schematic view, only entity types are displayed, and no specific entity is displayed, it may be assumed that an initial node a in fig. 6 is an audit subject, and the preset intelligent recommendation model may be used to order entity combinations from large to small according to a weight value of each entity combination to obtain a retrieval result corresponding to a knowledge retrieval request; the entities in the same type of entity combination are sorted according to the weight value except for the clauses, the clauses can be firstly grouped according to the service files, the clauses in the same group are sorted according to the weight value after being grouped, and the clauses are sorted among the groups according to the weight value of the service files.

In one example, the weight values of the respective entity combinations may be determined according to table 1:

TABLE 1

And processing the clause information and the service file information pushed by the page based on the weight setting of the service rule, and intelligently sequencing according to the calculation result of the intelligent recommendation index.

In order to improve the accuracy of the knowledge-graph and save the occupied storage space, in an embodiment of the present application, step 302 includes: obtaining an intermediate knowledge graph according to the batch entities and a pre-constructed graph knowledge layer; and carrying out entity disambiguation and entity alignment treatment on the intermediate knowledge graph to obtain the target knowledge graph.

Specifically, the descriptions of the same knowledge in different data sources will differ, and the definitions of the knowledge attributes will also differ, so that the integration of the audit knowledge can be realized by adopting an entity alignment technology and an entity disambiguation technology. For example, "electric charge payment urging" and "electric charge payment urging" occupy two positions in the map, exist as two different knowledge points, and need to be combined.

Through the extraction and representation of the knowledge of the power marketing inspection business (namely the power marketing business control knowledge), considerable formalized knowledge can be obtained, but due to the difference of knowledge sources, the quality of the knowledge is different, conflicts or overlaps exist among the knowledge, and the quantity and the quality of the knowledge are to be improved; the traditional knowledge fusion is to carry out rule processing on knowledge in the knowledge graph in a data fusion mode, and is not suitable for the construction of the knowledge graph in the power industry. The embodiment can process multi-source knowledge by applying knowledge fusion technology, and realize knowledge fusion of the knowledge map by entity disambiguation and entity alignment, thereby improving the professional quality of the knowledge map on one hand and enriching the stock of the knowledge on the other hand.

a) Entity disambiguation

Knowledge fusion model training can be performed based on a deep learning local entity link technique.

1. And performing atomization segmentation processing on the entity to be disambiguated and the text information corresponding to the clause where the entity is located and the inspection work order to form an entity set, and meanwhile, performing standardized processing on all entities by using knowledge in the constructed intermediate knowledge map.

2. Check in the knowledge base whether the same entity designation exists. If so, a set of candidate entities is constructed.

3. And judging the candidate entity set, wherein a plurality of candidate entities exist in the candidate entity set, and calculating indexes such as semantic similarity, text similarity and expression similarity of each candidate entity and the entity to be disambiguated by adopting a conditional random field model (CRF). A CRF model is a conditional probability distribution model of another set of output random variables given a set of input random variables, and also adds some constraints to the final prediction tags to ensure that they are valid. And resolving the ambiguity of the entity with the highest output probability according to the model calculation result. And (4) evaluating the result, wherein the overall evaluation index mainly comprises the F value and the running time of the whole algorithm.

b) Entity alignment

In the inspection service management and control knowledge map, the synonymy different entities exist, for example, the "payment for electric charge" and the "payment for electric charge" correspond to one knowledge entity, and the inspection service management and control knowledge map can be equivalent to the target knowledge map. Entity alignment is the study of alignment techniques to direct these entities to the same objective knowledge.

The entity alignment method based on the adaptive attribute selection of the decision tree can divide all entity sets into entity subsets which are irrelevant to each other, ensure that all potentially aligned entity pairs are divided into the same subset in the dividing process, and ensure that the entity number of each entity subset is uniformly distributed by using a load balancing strategy. And expressing the entity subsets in a vector space through joint learning, performing information gain calculation on the vector space corresponding to the entity subsets to select optimal constraint attributes, calculating the similarity of the optimal constraint attributes and the similarity of entity semantics, establishing an entity alignment model, and realizing the entity alignment of the knowledge graph.

To further explain the scheme, the application provides an application example of the electric power marketing management and control rule retrieval method, which is specifically described as follows:

construction inspection business control knowledge graph

An inspection business control knowledge graph belongs to a domain knowledge graph and is generally constructed in a top-down mode. The main construction process comprises the following steps: knowledge modeling, knowledge acquisition, knowledge fusion, knowledge storage and knowledge updating.

1) Knowledge modeling

The key technology of knowledge graph modeling is researched, and a semi-automatic construction scheme of the inspection service control knowledge graph schema is generated by combining the logic of knowledge retrieval, so that the modeling work of the inspection service control knowledge graph is guided. Specific objectives include:

a) the method is characterized in that policy files and historical worksheets of power grid related enterprises are combed, expert experience and content such as an industry dictionary and a knowledge base are combined to comb an initial knowledge map, and the policy files can be equivalent to the business files.

b) And the information extraction and verification work of part of policy documents is realized by researching the open information extraction technology.

2) Knowledge acquisition

Knowledge acquisition is used as a basic technology for constructing a knowledge graph, so that the acquisition of a structured named entity and attributes or associated information thereof from large-scale data is realized, the application field and scene of a knowledge extraction technology are expanded, and the recognition degree of the value and the necessity of the research of the knowledge extraction technology is improved.

The application example faces to the professional subdivision field of electric marketing inspection, and based on two professional data sets of a marketing data set and a historical work order set, knowledge acquisition is achieved, the knowledge acquisition comprises an entity extraction method, a relation extraction method, an attribute extraction method and the like, and finally knowledge extraction model training and verification aiming at unstructured data are achieved.

1. Firstly, an algorithm program is compiled by adopting a Python computing framework to realize model construction. The process needs to combine the Bert algorithm, the Bilstm algorithm and the CRF algorithm in a program and realize calculation logic through preliminary parameter setting.

2. Authentication data is input. The training data is information which can be identified by a machine formed by utilizing a BIO labeling system through secondary conversion based on labeling data. The verification data needs to be split according to the actual situation in the research and divided into a training set and a verification set, and under the general situation, the splitting work can be performed according to the ratio of 8:2, namely 80% of the data is used as the training set and 20% of the data is used as the verification set.

3. And (5) training a model. The deep learning neural network framework adopted by the model needs to call GPU resources to carry out model training work.

4. And (6) evaluating the model. After the model training is finished, the effect evaluation work of the model is required to be carried out according to different evaluation indexes, the important evaluation index is an accuracy index, and if AUC is more than 80%, the model prediction can be directly carried out; if the AUC is less than 80%, model optimization can be carried out through parameter adjustment under the condition that the labeled data have no problem, and if the labeled data have the problem, the labeled data directly return to a labeling link to label the problem data again.

5. And (5) model prediction. And performing model prediction by using the split verification set, and evaluating a prediction result. If the empty-tagging rate of the model is greater than 30%, the model training is proved not to cover all data, and the labeling is required to be carried out again; if the idle labeling rate is less than 30% and the AUC is less than 80%, returning to a model training link, and performing model optimization by adjusting parameters, and if the idle labeling rate is less than 30% and the AUC is greater than 80%, the model training result reaches the expectation, and the research is finished.

3) Knowledge fusion

The description of the same knowledge in different data sources will be different, and the definition of the knowledge attribute will also be different, so that the integration of the inspection knowledge can be realized by adopting an entity alignment technology and an entity disambiguation technology.

4) Knowledge storage

For knowledge storage, the main idea is to use different storages to realize storage and use of different types of data, and the data put into the knowledge graph is data rich in entities and having a large number of entity relations. The knowledge storage adopts a mixed storage mode of a graph database and a relational database.

5) Knowledge update

The knowledge of the electric power marketing inspection professional has the characteristics of short service change period, quick policy file updating and the like, and the research on the operation and maintenance of the knowledge is an essential link in the construction of the inspection service management and control knowledge map. The adopted knowledge updating technology mainly comprises the following steps: updating of overdue knowledge, finding and complementing of missing knowledge and finding and correcting of wrong knowledge.

(1) Expired knowledge updates

Knowledge is dynamically changed, so that the discovery of outdated knowledge in the knowledge graph and the timely update are important parts of quality control after the knowledge graph is constructed. The update mode of the knowledge graph can be divided into: active update and passive update. The active update is initiated by the knowledge-graph platform and the passive update is initiated by the data source. Most updates to the knowledge-graph are active updates. The early knowledge graph updating mainly adopts a mechanism of regular global updating in active updating, but the cost of each updating of the method is very high, so that a local updating mechanism is subsequently provided, namely, only local knowledge of the knowledge graph is updated each time. The mechanism of local update is divided into the following categories: update mechanisms based on update frequency prediction, update mechanisms based on time tags, and update mechanisms based on hotspot event discovery.

(2) Discovery and completion of missing knowledge

And the completion of the missing knowledge comprises the completion of entity types, the completion of relationships (or attribute information) among entities and the completion of missing attribute values of the entities.

Entity type completion is the completion of concepts missing from the construction of the knowledge graph, and is also commonly referred to as Entity Typing. Entity typing typically requires knowledge of the attributes and relationships already in the knowledge graph in the entity, as well as the entire knowledge graph. In addition, additional information may also be obtained by means of external knowledge sources. According to the technical route used for distinguishing, common entity judgment methods can be roughly divided into a heuristic probability model based on knowledge graph information and an entity classification model.

The relationship completion among the entities is to complete incomplete relationship triples (head entities, relationships and tail entities) in the knowledge graph. Among them, the scene of three-tuple head entity loss is rarely studied, and most research works mainly consider the relation or the scene of the tail entity. The relationship completion is roughly divided into: internal knowledge based relational completion and external data based relational completion.

Attribute value completion is the discovery and completion of missing attributes, and the problem of attribute value completion of knowledge maps is similar to but different from the classical attribute value completion problem in the field of relational databases. The attribute value deletion of the relational database is explicit, while the attribute value deletion in the knowledge graph is implicit, and firstly, the missing attribute of an entity in the knowledge graph needs to be found, and then the corresponding missing attribute value is filled. Entity missing attribute value discovery problems tend to translate into mandatory attribute discovery problems. The completeness of checking attribute information or attribute values may be generally established based on some specific assumptions: importance of attributes, reference to other entities under the same concept, reference to similar entities, pattern matching, partial integrity of attribute values, and the like.

(3) Discovery and correction of erroneous knowledge

No matter how the quality control is in place in the construction process, error knowledge is inevitably generated in the automatic construction of the knowledge graph, so that the discovery and correction of the error knowledge are crucial to the quality assurance of the knowledge graph. In the knowledge graph, concepts of entities, relationships between entities, and attribute information values may all be in error.

1. Error entity type detection

The detection of the error test question types in the corresponding knowledge graph is also related in the knowledge graph construction process, and the difference between the two methods is that the error entity types enough for the knowledge graph construction are performed after the knowledge extraction, the related information in the knowledge extraction process is unavailable, but the basic principle of the error type detection is still available in the concept mutual exclusion relationship and the like. Knowledge in the knowledge graph may also be relied upon to infer the type of entity that may be in error.

2. Erroneous entity relationship detection

Methods for discovering erroneous entity relationships are broadly divided into two categories: a method for detection based on data internal to the knowledge-graph and a method for detection by means of data external to the knowledge-graph. The first method establishes error entity judgment rules through the data association relation in the file knowledge graph, or establishes a probability model for error entity relation detection through analyzing the data distribution characteristics in the knowledge graph. The latter method usually finds the wrong entity relationship by means of an external data source such as the internet.

3. Error attribute value detection

Wrong attribute values are a common knowledge graph quality problem and are also of great interest. The error attribute value detection method still has two main ideas of 'internal detection' and 'appearance'. A common "inlining" approach is outlier detection, i.e., taking outliers that do not match the relevant data distribution as possible errors. The "look" method is similar to other error correction methods, and uses an external information source to find and correct the error attribute value. But the way in which attributes are referenced in the internet is quite diverse and therefore is subject to special consideration.

(II) intelligent retrieval based on inspection service control knowledge graph

The knowledge map of the electric marketing inspection professional (which can be equivalent to the target knowledge map) is stored in a Neo4j database in a map structure mode, and knowledge can be visually and flexibly represented and stored. The standard graph query algorithm is relatively complex. Aiming at the service scene of the inspection subject, the intelligent retrieval process mainly comprises the following four parts:

1) semantic parsing based on power marketing knowledge question-answer

Carrying out knowledge retrieval based on a Semantic Parsing (Semantic Parsing) method in a knowledge base question and answer; semantic analysis operation is carried out on natural language mainly through NL2SQL technology, a knowledge question (NL) is converted into semantic representation (logic form) which can be understood by a knowledge base, namely, an SPARQL query statement, and then reasoning and query are carried out on knowledge in a knowledge graph through inspection business control, so as to obtain a final answer; the semantic parsing process is a process of constructing a syntax tree from bottom to top according to the power marketing business rule, and the root node of the syntax tree is the final logical form expression of the natural language problem; the specific description is as follows:

1. and (5) mapping vocabularies.

The vocabulary mapping is a process of constructing nodes of a bottom syntax tree, and is a process of mapping a single natural language phrase or word to an entity of the inspection service control knowledge graph or a logic form corresponding to an entity relation; the vocabulary mapping can be completed by constructing a power marketing professional vocabulary, which can be equivalent to the above-mentioned search content and entity mapping relation table.

The electric power marketing professional vocabulary is single-point mapping of entity relations in the knowledge map controlled by the natural language and the inspection business, the mapping process is simple, and the mapping process can be carried out in a simple character string matching mode if two power generation prices are mapped into two power generation prices in the map or photovoltaic is mapped into distributed photovoltaic in the map.

2. Constructing SPARQL query.

A large number of service rules can be injected in the process of constructing the SPARQL query by applying the nl2sql technology, the category grammar is combined to complete the tree node merging work from bottom to top, and finally, a root node, namely a semantic parsing result, is generated.

2) Knowledge reasoning based on marketing business rules

The knowledge reasoning process based on the business rules is that on the premise of inputting question and answer knowledge, a knowledge retrieval method based on depth-first search (DFS) performs a graph traversal search process for inspecting business control knowledge graphs, meanwhile, weight setting of entities and relations is performed according to marketing business rules, calculation is performed based on a graph traversal algorithm, and all retrievable knowledge corresponding to the question and answer content is output.

3) Using intelligent recommendation index models

And (4) sorting the knowledge reasoning results based on the power marketing business rules, performing intelligent recommendation index processing calculation on answers according to the weight setting of each business rule, and sequencing entity combinations.

Through intelligent retrieval of the inspection service management and control knowledge graph, the input inspection target and service knowledge points are dynamically analyzed, through the NLP processes of semantic understanding, slot position identification, constraint identification and the like, the inspection service management and control knowledge graph is read, the contents of inspection theme rules, relevant legal regulations, service management methods, professional files and the like matched with the inspection target are displayed, and a user is assisted in achieving inspection theme creation and professional knowledge semantic retrieval.

In terms of software, in order to improve accuracy and efficiency of knowledge retrieval and further improve reliability and efficiency of power marketing inspection, the present application provides an embodiment of a knowledge retrieval device for implementing all or part of the contents in the knowledge retrieval method, and referring to fig. 7, the knowledge retrieval device specifically includes the following contents:

the acquisition module 71 is configured to acquire management and control knowledge of the batch power marketing service;

the building module 72 is configured to build a target knowledge graph according to the batch power marketing service control knowledge, where nodes in the target knowledge graph represent entities in the batch power marketing service control knowledge, and a directional edge between two nodes represents an association relationship between the entities;

and the retrieval module 73 is configured to receive a knowledge retrieval request, and obtain a retrieval result corresponding to the knowledge retrieval request by applying a preset retrieval rule and the target knowledge graph.

In one embodiment of the present application, the building module includes:

In one embodiment of the present application, the retrieval module includes:

In an embodiment of the present application, the building unit is configured to: obtaining an intermediate knowledge graph according to the batch entities and a pre-constructed graph knowledge layer; and carrying out entity disambiguation and entity alignment treatment on the intermediate knowledge graph to obtain the target knowledge graph.

The embodiment of the knowledge retrieval device provided in this specification may be specifically used to execute the processing flow of the embodiment of the knowledge retrieval method, and the functions of the embodiment are not described herein again, and reference may be made to the detailed description of the embodiment of the knowledge retrieval method.

In terms of hardware, in order to improve accuracy and efficiency of knowledge retrieval and further improve reliability and efficiency of power marketing inspection, the present application provides an embodiment of an electronic device for implementing all or part of contents in the knowledge retrieval method, where the electronic device specifically includes the following contents:

a processor (processor), a memory (memory), a communication Interface (Communications Interface), and a bus; the processor, the memory and the communication interface complete mutual communication through the bus; the communication interface is used for realizing information transmission among the knowledge retrieval device, the user terminal and other related equipment; the electronic device may be a desktop computer, a tablet computer, a mobile terminal, and the like, but the embodiment is not limited thereto. In this embodiment, the electronic device may be implemented by referring to the embodiment of the embodiment for implementing the knowledge retrieval method and the embodiment for implementing the knowledge retrieval apparatus, and the contents of the embodiments are incorporated herein, and repeated descriptions are omitted.

Fig. 8 is a schematic block diagram of a system configuration of an electronic device 9600 according to an embodiment of the present application. As shown in fig. 8, the electronic device 9600 can include a central processor 9100 and a memory 9140; the memory 9140 is coupled to the central processor 9100. Notably, this FIG. 8 is exemplary; other types of structures may also be used in addition to or in place of the structure to implement telecommunications or other functions.

In one or more embodiments of the present application, the knowledge retrieval functionality can be integrated into the central processor 9100. The central processor 9100 may be configured to control as follows:

From the above description, the electronic device provided by the embodiment of the application can improve accuracy and efficiency of knowledge retrieval, and further improve reliability and efficiency of electric power marketing inspection.

In another embodiment, the knowledge retrieval device may be configured separately from the central processor 9100, for example, the knowledge retrieval device may be configured as a chip connected to the central processor 9100, and the knowledge retrieval function is realized by the control of the central processor.

As shown in fig. 8, the electronic device 9600 may further include: a communication module 9110, an input unit 9120, an audio processor 9130, a display 9160, and a power supply 9170. It is worthy to note that the electronic device 9600 also does not necessarily include all of the components shown in fig. 8; further, the electronic device 9600 may further include components not shown in fig. 8, which may be referred to in the art.

As shown in fig. 8, a central processor 9100, sometimes referred to as a controller or operational control, can include a microprocessor or other processor device and/or logic device, which central processor 9100 receives input and controls the operation of the various components of the electronic device 9600.

The memory 9140 can be, for example, one or more of a buffer, a flash memory, a hard drive, a removable media, a volatile memory, a non-volatile memory, or other suitable device. The information relating to the failure may be stored, and a program for executing the information may be stored. And the central processing unit 9100 can execute the program stored in the memory 9140 to realize information storage or processing, or the like.

The input unit 9120 provides input to the central processor 9100. The input unit 9120 is, for example, a key or a touch input device. Power supply 9170 is used to provide power to electronic device 9600. The display 9160 is used for displaying display objects such as images and characters. The display may be, for example, but is not limited to, an LCD display.

The memory 9140 can be a solid state memory, e.g., Read Only Memory (ROM), Random Access Memory (RAM), a SIM card, or the like. There may also be a memory that holds information even when power is off, can be selectively erased, and is provided with more data, an example of which is sometimes called an EPROM or the like. The memory 9140 could also be some other type of device. Memory 9140 includes a buffer memory 9141 (sometimes referred to as a buffer). The memory 9140 may include an application/function storage portion 9142, the application/function storage portion 9142 being used for storing application programs and function programs or for executing a flow of operations of the electronic device 9600 by the central processor 9100.

The memory 9140 can also include a data store 9143, the data store 9143 being used to store data, such as contacts, digital data, pictures, sounds, and/or any other data used by an electronic device. The driver storage portion 9144 of the memory 9140 may include various drivers for the electronic device for communication functions and/or for performing other functions of the electronic device (e.g., messaging applications, contact book applications, etc.).

The communication module 9110 is a transmitter/receiver 9110 that transmits and receives signals via an antenna 9111. The communication module (transmitter/receiver) 9110 is coupled to the central processor 9100 to provide input signals and receive output signals, which may be the same as in the case of a conventional mobile communication terminal.

Based on different communication technologies, a plurality of communication modules 9110, such as a cellular network module, a bluetooth module, and/or a wireless local area network module, may be provided in the same electronic device. The communication module (transmitter/receiver) 9110 is also coupled to a speaker 9131 and a microphone 9132 via an audio processor 9130 to provide audio output via the speaker 9131 and receive audio input from the microphone 9132, thereby implementing ordinary telecommunications functions. The audio processor 9130 may include any suitable buffers, decoders, amplifiers and so forth. In addition, the audio processor 9130 is also coupled to the central processor 9100, thereby enabling recording locally through the microphone 9132 and enabling locally stored sounds to be played through the speaker 9131.

According to the description, the electronic equipment provided by the embodiment of the application can improve the accuracy and efficiency of knowledge retrieval, and further improve the reliability and efficiency of electric power marketing inspection.

Embodiments of the present application also provide a computer-readable storage medium capable of implementing all the steps in the knowledge retrieval method in the above embodiments, where the computer-readable storage medium stores thereon a computer program, and when the computer program is executed by a processor, the computer program implements all the steps in the knowledge retrieval method in the above embodiments, for example, when the processor executes the computer program, the processor implements the following steps:

From the above description, it can be seen that the computer-readable storage medium provided in the embodiments of the present application can improve accuracy and efficiency of knowledge retrieval, and further improve reliability and efficiency of electric power marketing inspection.

In the present application, each embodiment of the method is described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments. Reference is made to the description of the method embodiments.

As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.

The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable power marketing auditing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable power marketing auditing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable power marketing inspection device to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be loaded onto a computer or other programmable power marketing inspection device to cause a series of operational steps to be performed on the computer or other programmable device to produce a computer implemented process such that the instructions which execute on the computer or other programmable device provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

The principle and the implementation mode of the present application are explained by applying specific embodiments in the present application, and the description of the above embodiments is only used to help understanding the method and the core idea of the present application; meanwhile, for a person skilled in the art, according to the idea of the present application, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present application.

Claims

1. A knowledge retrieval method, comprising:

acquiring management and control knowledge of batch power marketing services;

2. The knowledge retrieval method of claim 1, wherein the constructing a target knowledge graph according to the batch power marketing business management and control knowledge comprises:

3. The knowledge retrieval method of claim 1, wherein the applying preset retrieval rules and the target knowledge graph to obtain a retrieval result corresponding to the knowledge retrieval request comprises:

4. The knowledge retrieval method of claim 2, wherein the obtaining the target knowledge graph according to the batch of entities and a pre-constructed graph knowledge layer comprises:

5. A knowledge retrieval apparatus, comprising:

6. The knowledge retrieval device of claim 5, wherein the building module comprises:

7. The knowledge retrieval device of claim 5, wherein the retrieval module comprises:

8. The knowledge retrieval device of claim 6, wherein the construction unit is configured to:

9. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the knowledge retrieval method of any one of claims 1 to 4 when executing the program.

10. A computer-readable storage medium having computer instructions stored thereon, wherein the instructions, when executed, implement the knowledge retrieval method of any one of claims 1 to 4.