CN116719946B

CN116719946B - Work order recommending method, work order recommending device, storage medium and electronic equipment

Info

Publication number: CN116719946B
Application number: CN202311004362.3A
Authority: CN
Inventors: 冯晨; 陈子鹏; 孙佩霞
Original assignee: China Telecom Corp Ltd
Current assignee: China Telecom Corp Ltd
Priority date: 2023-08-10
Filing date: 2023-08-10
Publication date: 2023-11-28
Anticipated expiration: 2043-08-10
Also published as: CN116719946A

Abstract

The disclosure relates to the technical field of data processing, in particular to a work order recommending method and device, a computer readable storage medium and electronic equipment. The method comprises the following steps: determining a first distance between the entity characteristics of a first target field and the cluster center characteristics of the existing entity cluster aiming at the first target field of the work order to be processed, and determining a matched entity cluster of the first target field according to the first distance; determining a matching work order list of a first target field based on the processed work orders to which the entities included in the matching entity cluster belong; according to the matching work order list of the first target field, determining a candidate work order list of the work order to be processed; calculating the similarity between the to-be-processed work order and the candidate work order in the candidate work order list so as to determine a reference recommended work order of the to-be-processed work order from the candidate work orders; the existing entity clustering clusters are obtained by clustering the entities in the processed worksheets in advance. The method and the system can improve accuracy and efficiency of work order recommendation.

Description

Work order recommending method, work order recommending device, storage medium and electronic equipment

Technical Field

The disclosure relates to the technical field of data processing, in particular to a work order recommending method, a work order recommending device, a computer readable storage medium and electronic equipment.

Background

In the related art, a technical expert summarizes work order cases according to experience to form a case base, and then instructs service personnel to analyze fault information and provide solutions according to the case base.

However, because of the redundancy of work order information, locating the cause of the failure and finding the matching reference work order case is time consuming, the related art has low processing efficiency and low accuracy.

It should be noted that the information disclosed in the above background section is only for enhancing understanding of the background of the present disclosure and thus may include information that does not constitute prior art known to those of ordinary skill in the art.

Disclosure of Invention

The disclosure aims to provide a work order recommending method, a work order recommending device, a computer readable storage medium and electronic equipment, so that the efficiency and the accuracy of work order recommending are improved at least to a certain extent.

According to a first aspect of the present disclosure, there is provided a work order recommendation method, including: determining a first distance between an entity characteristic of a first target field and a clustering center characteristic of an existing entity cluster aiming at the first target field of a work order to be processed, and determining a matching entity cluster of the first target field according to the first distance; determining a matching work order list of the first target field based on the processed work orders to which the entities included in the matching entity cluster belong; determining a candidate work order list of the work order to be processed according to the matched work order list of the first target field; calculating the similarity between the work order to be processed and the candidate work orders in the candidate work order list, and determining a reference recommended work order of the work order to be processed from the candidate work orders according to the similarity; the existing entity clustering clusters are obtained by clustering the entities in the processed worksheet in advance.

In an exemplary embodiment, based on the foregoing embodiment, the clustering of the existing entity clusters by clustering the entities in the processed worksheet in advance includes: extracting the entity from the processed work order, and carrying out text coding on the extracted entity through a language coding model to obtain an entity characteristic vector; based on the distance between the entity feature vectors, the entity features are clustered according to a clustering algorithm to obtain the existing entity clustering clusters.

In an exemplary embodiment, based on the foregoing embodiment, the method further includes: extracting the entity from the second target field of the processed work order to obtain the entity corresponding to the second target field of the processed work order; generating an entity relation triplet according to one or more of the subordinate relation between the processed work order and the second target field, the causal relation between the entities corresponding to the second target field and the time sequence relation between the entities corresponding to the second target field; and generating a work order knowledge graph according to the entity relation triplet.

In an exemplary embodiment, based on the foregoing embodiment, the determining, based on the processed worksheets to which the entities included in the matched entity cluster belong, the matched worksheet list of the first target field includes: searching a processed work order of the entity included in the matched entity cluster in the work order knowledge graph; and determining a matching work order list of the first target field according to the searched processed work orders.

In an exemplary embodiment, based on the foregoing embodiment, the determining, according to the matching worksheet list of the first target field, the candidate worksheet list of the to-be-processed worksheet includes: performing multiple intersection operations on the matched worksheet list of each first target field according to a first preset intersection operation sequence, wherein the first preset intersection operation sequence is used for indicating the sequence of intersection operations between the first target fields; and determining a candidate work order list of the work orders to be processed according to the intersection operation result.

In an exemplary embodiment, based on the foregoing embodiment, the determining, according to the matching worksheet list of the first target field, the candidate worksheet list of the to-be-processed worksheet includes: determining an initial candidate work order list of the work order to be processed from the work order knowledge graph according to a preset rule, wherein the preset rule is determined according to a third target field; and determining a candidate work order list of the work order to be processed according to the initial candidate work order list and the matched work order list of each first target field.

In an exemplary embodiment, based on the foregoing embodiment, the determining the candidate list of the to-be-processed work order according to the initial candidate list of work orders and the matching list of work orders of the first target field includes: performing multiple intersection operations on the initial candidate list and the matched list of each first target field according to a second preset intersection operation sequence, wherein the second preset intersection operation sequence is used for indicating the sequence of the intersection operations between the matched list of each first target field and the initial candidate list; and determining a candidate work order list of the work orders to be processed according to the intersection operation result.

In an exemplary embodiment, based on the foregoing embodiment, the determining the candidate list of the work order to be processed according to the result of the intersection operation includes: and when the result of the current intersection operation is empty, determining a candidate work order list of the work order to be processed according to the result of the previous intersection operation of the current intersection operation.

In an exemplary embodiment, based on the foregoing embodiment, the calculating the similarity between the work order to be processed and the candidate work orders in the candidate work order list includes: calculating a first similarity between an entity of a fourth target field of the work order to be processed and an entity of a fourth target field of a candidate work order in the candidate work order list; and determining the similarity between the work order to be processed and the candidate work orders in the candidate work order list according to the first similarity.

In an exemplary embodiment, based on the foregoing embodiment, the calculating the similarity between the work order to be processed and the candidate work orders in the candidate work order list includes: determining a work order pair sub-graph according to a first sub-knowledge graph of the candidate work order and a second sub-knowledge graph of the work order to be processed aiming at each candidate work order in the candidate work order list; inputting the work order sub-graph into a pre-training graph neural network model, and encoding node characteristics and network topology information in the work order sub-graph through the pre-training graph neural network model to obtain a second similarity between a first sub-knowledge graph and a second sub-knowledge graph in the work order sub-graph; and determining the similarity between the work order to be processed and the candidate work order according to the second similarity.

In an exemplary embodiment, based on the foregoing embodiment, the calculating the similarity between the work order to be processed and the candidate work orders in the candidate work order list includes: determining a first target similarity between the work order to be processed and the candidate work order in the candidate work order list according to the similarity between the entity of the fourth target field of the work order to be processed and the entity of the fourth target field of the candidate work order in the candidate work order list; determining a work order pair sub-graph according to a first sub-knowledge graph of the candidate work order and a second sub-knowledge graph of the work order to be processed aiming at each candidate work order in the candidate work order list; inputting the work order sub-graph into a pre-training graph neural network model, and encoding node characteristics and network topology information in the work order sub-graph through the pre-training graph neural network model to obtain a second target similarity between the work order to be processed and the candidate work order; determining fusion weights of the first target similarity and the second target similarity according to an ensemble learning algorithm; and fusing the first target similarity and the second target similarity based on the fusion weight so as to determine the similarity between the work order to be processed and the candidate work order in the candidate work order list according to the fusion result.

In an exemplary embodiment, based on the foregoing embodiment, the determining, according to the similarity, the reference recommended worksheets of the worksheets to be processed from the candidate worksheets includes sorting the similarities in a descending order, and determining, according to the candidate worksheets indicated by the similarities sorted in the first N, the reference recommended worksheets of the worksheets to be processed.

According to a second aspect of the present disclosure, there is provided a work order recommendation device including: the matching entity cluster determining module is configured to determine, for a first target field of a work order to be processed, a first distance between an entity feature of the first target field and a cluster center feature of an existing entity cluster, and determine a matching entity cluster of the first target field according to the first distance; the matching work order list determining module is configured to determine a matching work order list of the first target field based on the processed work orders of the entities included in the matching entity cluster; the candidate work order list determining module is configured to determine a candidate work order list of the work order to be processed according to the matched work order list of the first target field; the reference recommended work order determining module is configured to calculate the similarity between the work order to be processed and the candidate work orders in the candidate work order list, and determine the reference recommended work order of the work order to be processed from the candidate work orders according to the similarity; the existing entity clustering clusters are obtained by clustering the entities in the processed worksheet in advance.

According to a third aspect of the present disclosure, there is provided a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements the work order recommendation method described in the first aspect.

According to a fourth aspect of the present disclosure, there is provided an electronic apparatus, comprising: one or more processors; and a memory for storing one or more programs which, when executed by the one or more processors, cause the one or more processors to implement the work order recommendation method described in the first aspect.

According to a fifth aspect of the present disclosure there is provided a computer program product comprising instructions which, when run on a computer, cause the computer to perform the steps of the work order recommendation method as described in the first aspect.

The technical scheme of the present disclosure has the following beneficial effects:

in the disclosure, on one hand, similar entities are clustered through the entity cluster to realize entity alignment, so that when entity matching is performed to determine a matched work order list, only the current entity characteristics and the entity characteristics of the cluster center are required to be matched, the efficiency of entity matching can be improved, and the efficiency of work order recommendation is further improved; on the other hand, similar entities are merged through the entity cluster, so that the accuracy of entity matching can be improved, and the accuracy of worksheet recommendation is further improved.

It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the disclosure and together with the description, serve to explain the principles of the disclosure. It will be apparent to those of ordinary skill in the art that the drawings in the following description are merely examples of the disclosure and that other drawings may be derived from them without undue effort. In the drawings:

fig. 1 shows a flowchart of a work order recommendation method in an exemplary embodiment of the present disclosure.

Fig. 2 shows a flow diagram of a method of obtaining an existing entity cluster in an exemplary embodiment of the present disclosure.

FIG. 3 illustrates a schematic diagram of a processed work order in an exemplary embodiment of the present disclosure.

Fig. 4 illustrates a schematic diagram of encoding physical features in an exemplary embodiment of the present disclosure.

Fig. 5 shows a schematic diagram of a resulting existing entity cluster in an exemplary embodiment of the present disclosure.

Fig. 6 shows a flow diagram of a method of generating a worksheet knowledge graph in an exemplary embodiment of the present disclosure.

Fig. 7 shows a schematic diagram of a generated work order knowledge graph in an exemplary embodiment of the present disclosure.

Fig. 8 shows a schematic diagram of a matched entity cluster in an exemplary embodiment of the present disclosure.

FIG. 9 illustrates a flow chart of a method of determining a list of candidate worksheets for the worksheets to be processed in an exemplary embodiment of the present disclosure.

Fig. 10 shows a flow chart of a method of computing similarity in an exemplary embodiment of the present disclosure.

Fig. 11 shows a flow diagram of another method of calculating similarity in an exemplary embodiment of the present disclosure.

Fig. 12 shows a schematic diagram of determining similarity in an exemplary embodiment of the present disclosure.

Fig. 13 shows a flow diagram of another worksheet recommendation method in an exemplary embodiment of the present disclosure.

Fig. 14 shows a flow diagram of yet another worksheet recommendation method in an exemplary embodiment of the present disclosure.

Fig. 15 shows a composition diagram of a work order recommendation device in an exemplary embodiment of the present disclosure.

Fig. 16 shows a schematic diagram of an electronic device to which the exemplary embodiments of the present disclosure may be applied.

Detailed Description

Example embodiments will now be described more fully with reference to the accompanying drawings. However, the exemplary embodiments may be embodied in many forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the concept of the example embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

Furthermore, the drawings are merely schematic illustrations of the present disclosure and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and thus a repetitive description thereof will be omitted. Some of the block diagrams shown in the figures are functional entities and do not necessarily correspond to physically or logically separate entities. These functional entities may be implemented in software or in one or more hardware modules or integrated circuits or in different networks and/or processor devices and/or microcontroller devices.

A work order, i.e., a work order, is a simple maintenance or manufacturing plan consisting of one or more jobs. The worksheet may be independent or may be part of a large item.

The operation and maintenance work order refers to a problem list which is initiated by a client or other internal users and requests IT operation and maintenance personnel to process. The worksheets typically include fault names, fault causes, fault reports, service requests, change requests, fault solutions, and the like.

There are a large number of work order cases in the current network operation and maintenance scene, and these work order cases can be understood as processed work orders, i.e. faults or requests in the work order cases have been successfully processed, and technical specialists can summarize the work order cases according to experience, so as to generate a work order case library, and the work order case library is used for guiding business personnel to analyze fault information and provide solutions.

However, in the process of processing the work order, because the work order information is redundant, locating the fault cause and finding the matched reference case are very time-consuming, so that the recommending efficiency of the work order reference case is low and the recommending accuracy is low.

In view of the foregoing, exemplary embodiments of the present disclosure provide a work order recommendation method.

A work order recommendation method in the present exemplary embodiment will be described below with reference to fig. 1, and fig. 1 shows an exemplary flow of the work order recommendation method, which may include:

step S110, determining a first distance between an entity characteristic of a first target field and a clustering center characteristic of an existing entity cluster aiming at the first target field of a work order to be processed, and determining a matching entity cluster of the target field according to the first distance;

step S120, determining a matching work order list of the first target field based on the processed work orders of the entities included in the matching entity cluster;

step S130, determining a candidate work order list of the work order to be processed according to the matched work order list of the first target field;

step S140, calculating the similarity between the work order to be processed and the candidate work orders in the candidate work order list, and determining a reference recommended work order of the work order to be processed from the candidate work orders according to the similarity;

The existing entity clustering clusters are obtained by clustering the entities in the processed worksheet in advance.

Based on the method, on one hand, similar entities are clustered through the entity cluster to realize entity alignment, so that when entity matching is carried out to determine a matched work order list, only the current entity characteristics and the entity characteristics of the cluster center are required to be matched, the entity matching efficiency can be improved, and the work order recommending efficiency is further improved; on the other hand, similar entities are merged through the entity cluster, so that the accuracy of entity matching can be improved, and the accuracy of worksheet recommendation is further improved.

The steps shown in fig. 1 are specifically described below.

Step S110, determining a first distance between an entity characteristic of a first target field and a cluster center characteristic of an existing entity cluster aiming at the first target field of a work order to be processed, and determining a matched entity cluster of the first target field according to the first distance.

In an exemplary embodiment, the existing entity cluster is obtained by clustering entities in the processed worksheet in advance. In other words, each of the existing entity clusters may include a plurality of entities having similar characteristic representations therein.

The processed worksheets may include, among other things, case worksheets, i.e., worksheets that already have solutions. The work order to be processed may include work orders without solutions. The reference recommended worksheet of the to-be-processed worksheet may be determined from the processed worksheets, such that a solution of the to-be-processed worksheet is determined by referencing the solutions in the recommended worksheets.

Illustratively, fig. 2 shows a flow chart of a method for obtaining an existing entity cluster in an exemplary embodiment of the present disclosure. Referring to fig. 2, the method may include steps S210 to S220. Wherein:

in step S210, entity extraction is performed on the processed work order, and text encoding is performed on the extracted entity through a language encoding model to obtain an entity feature vector.

Illustratively, FIG. 3 shows a schematic diagram of a processed work order in an exemplary embodiment of the present disclosure. As previously described, the processed worksheets may include case worksheets, as shown in FIG. 3, and each processed worksheet (e.g., case 1, case 2, case 3, etc. in FIG. 3) may include a plurality of fields, such as fields for a fault major class, fault minor class, alarm information, fault description, etc.

The entity extraction may be performed on each field in the processed worksheet, and the entity extraction may be performed on a preset field in the processed worksheet, where the preset field may be determined according to requirements, which is not particularly limited in this exemplary embodiment. The extracted entities are then encoded to obtain entity feature vectors.

For example, for complex fields that describe more information and include multiple types of entities, such as a fault description field, a neural network model UIE (Universal Information Extraction, general information extraction) text entity extraction model may be used to extract various types of entities in the complex field. The UIE model adopts a transducer as a network structure, and is subjected to training fine adjustment through labeling data extracted by a work order entity. For simple fields, such as fault major field and fault minor field, which are directly described by simple words, the value of the field can be directly taken as the extracted entity. Finally, a single worksheet can be extracted in each field to obtain one or more entities, such as a plurality of fault cause entities and a plurality of fault entities.

For each entity extracted, the extracted entity may be encoded by a language encoding model, such as Bert (Bidirectional Encoder Representation from Transformers, bi-directional encoded representation based on a transformer), to obtain a one-dimensional entity feature vector for a corresponding fixed number of feature bits.

Illustratively, FIG. 4 shows a schematic diagram of encoding physical features in one exemplary embodiment of the present disclosure. Referring to fig. 4, the entity 2 may be input into the language encoding model 41 in fig. 4, thereby obtaining the entity characteristics of the entity 2, i.e., the characteristics 2.

In an exemplary embodiment, both the processed worksheets and the to-be-processed worksheets of the present disclosure may be in text form, such as the form shown in fig. 3. Of course, the processed worksheet and the to-be-processed worksheet in the present disclosure may also be in an image format, such as a processed worksheet image and a to-be-processed worksheet image obtained by photographing the table in fig. 3. Under the condition that the work order to be processed and the work order to be processed are in an image format, the image character recognition can be carried out on the work order to be processed and the work order to be processed, and the structured text processed work order and the text work order to be processed are generated according to the recognized characters. If the character recognition is carried out on the to-be-processed work order image, the work order identification of the to-be-processed work order, each field in the to-be-processed work order and text information corresponding to each field are obtained, and therefore the structured to-be-processed work order text is generated.

In step S220, the entity features are clustered according to a clustering algorithm based on the distances between the entity feature vectors, so as to obtain an existing entity cluster.

In an exemplary embodiment, the clustering algorithm may include a Kmeans (K-means) clustering algorithm. In other words, the entity features can be clustered by Kmeans, and the optimal setting needs to be selected for the current data because the number of clusters is a specified super parameter. In the disclosure, the contour coefficient (silhouette cofficient) can be used as an evaluation index to select the optimal clustering quantity in a certain optional parameter range in a grid search mode, and the entity characteristics are clustered based on the optimal clustering quantity, so that the existing entity clustering clusters of the optimal clustering quantity are obtained. The cluster center feature vector of each existing entity cluster can also be stored.

Illustratively, fig. 5 shows a schematic diagram of an existing entity cluster obtained in one exemplary embodiment of the present disclosure. Referring to fig. 5, 3 existing entity clusters are shown. Wherein 51 represents the extracted entity, and 52, 53, 54 are respectively the cluster centers of 3 existing entity clusters.

Due to factors such as the difference of the expression of the worksheet text, the granularity of the entity extracted by different worksheets is different, namely the co-pointing phenomenon that the entity with fine difference has the same ideas exists. Through the steps S210 to S220, entities with similar ideas can be merged, so as to improve the entity matching precision in the subsequent query process.

In an exemplary embodiment, a worksheet knowledge graph may also be constructed based on the extracted entities in the existing worksheets.

Illustratively, fig. 6 shows a flow diagram of a method of generating a worksheet knowledge graph in an exemplary embodiment of the present disclosure. Referring to fig. 6, the method may include steps S610 to S630. Wherein:

in step S610, entity extraction is performed on the second target field of the processed work order to obtain an entity corresponding to the second target field of the processed work order.

In an exemplary embodiment, the second target field may include all fields in the processed work order, or may include a part of fields in the processed work order, which may determine, according to a user requirement, which fields in the processed work order need to be physically extracted, so as to construct a work order knowledge graph, which is not particularly limited in this exemplary embodiment.

It should be noted that the second target field may include all of the first target fields. In other words, the second target field may be identical to the first target field, and the second target field may also include other work order fields in addition to all of the first target fields.

The manner of extracting the entity from the work order may refer to the related content in step S210, which is not described herein.

In step S620, an entity relationship triplet is generated according to one or more of the affiliation of the processed work order and the second target field, the causal relationship between the entities corresponding to the second target field, and the time series relationship between the entities corresponding to the second target field.

For example, a first entity relationship triplet may be generated based on the membership of the work order identification entity (e.g., work order 1, work order 2) of the processed work order and the fields in the processed work order. If the fault field in the work order 1 includes the fault entity 1, and the fault reason field in the work order 1 includes the fault reason entity 1, two first entity relationship triples (work order 1, affiliation, fault entity 1), (work order 1, affiliation, fault reason entity 1) may be generated.

In other words, triples (head node, relationship, tail node) may be generated based on the affiliations of worksheets and fields, where the head node is a worksheet-identifying entity, the relationship is an affiliation, and the tail node is an entity included in a field.

The second entity relationship triplet may also be generated according to a causal relationship or a time sequence relationship between the entities of each field, for example, for a fault reason field, if the fault reason 1 indicated by the fault reason entity 1 may cause the occurrence of the fault reason 2 indicated by the fault reason entity 2, that is, the fault reason entity 1 and the fault reason entity 2 have a causal relationship, the second entity relationship triplet (the fault reason entity relationship 1, the causal relationship, and the fault reason entity relationship 2) may be generated. For another example, for the fault alarm field, alarm 1 is first performed, and then alarm 2 is performed, that is, a time sequence relationship exists between fault alarm entity 1 and fault alarm entity 2, then a second entity relationship triplet (fault alarm entity 1, time sequence relationship, fault alarm entity 2) may be generated.

In step S630, a worksheet knowledge graph is generated according to the entity relationship triplet.

For example, a worksheet knowledge graph can be constructed in a graphic database based on the entity relationship triples, so that all entity information corresponding to the worksheet can be obtained in the worksheet knowledge graph by a worksheet identification entity (such as a worksheet number), and the worksheet number to which the entity belongs can be obtained in the worksheet knowledge graph through the entity information of a specific field.

Illustratively, fig. 7 shows a schematic diagram of a generated worksheet knowledge graph in an exemplary embodiment of the present disclosure. Referring to fig. 7, the information of the entity included in each work order may be displayed in the work order knowledge graph, and association information between different entities may also be displayed, for example, an association exists between an alarm 1 entity and an alarm 2 entity, an association exists between an alarm 2 entity and a network element 2 entity, an association exists between an alarm 1 entity and a network element 1 entity in the work order 1, and an association exists between an alarm 1 entity and a network element 3 entity in the work order 2.

In an exemplary embodiment, the existing entity cluster may include an entity cluster obtained by clustering the entities of the fields in the knowledge graph.

In an exemplary embodiment, the first target field may be customized according to a user requirement, for example, the first target field may include a field having a greater influence on the work order solution, for example, a fault major field, a fault minor field, a fault name, a fault reason, and the like. The number of the first target fields may be customized according to the requirement, and the present exemplary embodiment is not particularly limited thereto.

For example, the entity extraction method may be used to extract the entity from the first target field in the work order to be processed, and obtain the entity feature vector of the entity extracted from each first target field in the work order to be processed based on the language coding model coding. And then, for each first target field, calculating the cosine distance between each entity characteristic vector in the first target field and the cluster center characteristic vector of each existing entity cluster, and selecting the existing entity cluster with the minimum cosine distance as the matching entity cluster of the entity of the first target field. For a first target field in which a plurality of entities exist, the union of the matched clusters of the plurality of entities is the matched entity cluster of the first target field. That is, for a first target field where multiple entities exist, there are multiple corresponding matches to the entity cluster.

Illustratively, fig. 8 shows a schematic diagram of a matched entity cluster in an exemplary embodiment of the present disclosure. Referring to fig. 8, for the entity characteristics 81 of the work order to be processed currently, it is determined that the corresponding matching entity cluster is the existing entity cluster 82.

With continued reference to fig. 1, in step S120, a list of matching worksheets for the first target field is determined based on the processed worksheets to which the entities included in the matching entity cluster belong.

For example, a specific embodiment of step S120 may include searching the worksheet knowledge map for a processed worksheet to which an entity included in the matched entity cluster belongs; and determining a matching work order list of the first target field according to the searched processed work orders.

For example, after determining the matched entity cluster of each first target field, the entity set in the matched entity cluster may be obtainedThen inquiring the entity set in the worksheet knowledge graph>The processed worksheet of any of the entities in question, thereby obtaining a list of matching worksheets for the first target field.

For a first target field with a plurality of entities, a matching work order list corresponding to each entity can be determined according to the matching cluster corresponding to each entity, and the matching work order list of the first target field is determined according to the union of the matching work order lists corresponding to each entity.

In step S130, a candidate list of the work order to be processed is determined according to the matching list of the first target field.

By way of example, one embodiment of step S130 may include; performing multiple intersection operations on the matched worksheet list of each first target field according to a first preset intersection operation sequence, wherein the first preset intersection operation sequence is used for indicating the sequence of intersection operations between the first target fields; and determining a candidate work order list of the work orders to be processed according to the intersection operation result.

For example, the first preset intersection operation sequence corresponding to the first target field may be determined according to the importance degree of the first preset target field to the work order solution. And performing intersection operation on the matched work order list of the two fields with the highest importance degree, performing intersection operation on the intersection operation result and the matched work order list corresponding to the field with the rank 3 of the importance degree, and the like, so that multiple intersection operations are performed according to a preset intersection operation sequence, and a candidate work order list of the work order to be processed is determined according to the multiple intersection operation result.

Illustratively, when the result of the current intersection operation is empty, determining the candidate work order list of the work order to be processed according to the result of the previous intersection operation of the current intersection operation.

In an exemplary embodiment, the matching worksheet list of each first target field may be sequentially generated according to the first preset intersection operation sequence, that is, the matching worksheet list of each first target field is not generated at first, then intersection operation is directly performed first according to the current matching worksheet list of the first target field, when the current intersection operation result is not empty, the matching worksheet list of the next first target field is determined, if the current intersection operation result is empty, the matching worksheet list of the next first target field is not calculated, and the candidate worksheet list of the to-be-processed worksheets can be determined directly according to the previous intersection operation result of the current time. In this way, computing resources may be saved and efficiency may be improved.

If there isA first target field with a first predetermined intersection operation order of 1,2,3, …, < >>The corresponding matching worksheet lists are +.>Then the first can be calculatedMatching worksheet list of target field 1 and matching worksheet list of first target field 2, e.g., respectively +.>And->Then first +.>Andperforming intersection operation to obtain a first intersection operation result, and if the first intersection operation result is not null, determining a matching worksheet list of the first target field 3, for example +.>Then the first intersection operation result is added with +.>And performing intersection operation to obtain a second intersection operation result, if the second intersection element result is not null, continuously calculating a matching work order list of the first target field 4, performing intersection operation with the second intersection operation result, if the second intersection operation result is null, determining a candidate work order list of the work order to be processed directly according to the second intersection operation result without determining the matching work order list of the first target field 4 to the first target field, and so on.

For example, another embodiment of step S130 may be shown with reference to fig. 9. FIG. 9 illustrates a flow chart of a method of determining a list of candidate worksheets for the worksheets to be processed in an exemplary embodiment of the present disclosure. As shown in fig. 9, the method may include steps S910 to S920.

In step S910, an initial candidate list of the work order to be processed is determined from the work order knowledge graph according to a preset rule.

In an exemplary embodiment, the preset rule is determined according to a third target field. Wherein the third target field may comprise a number of fields that have a leading impact on the work order solution. The third target field may also include a field where there are no multiple entity co-fingers.

The preset rules may include that the entities of the fault major and minor fields are completely matched with the fault major and minor in the worksheet knowledge map. In this way, the work orders of which the fault large-class field entity and the fault small-class field entity are completely matched with the fault large-class field entity and the fault small-class field entity in the work order to be processed can be searched from the knowledge graph according to the preset rule, so that an initial candidate work order list is obtained。

For example, the fault large field entity in the to-be-processed worksheet and the fault large field entity in the knowledge graph may be subjected to character matching, a first initial candidate worksheet list may be determined according to the worksheet to which the fault large field entity successfully matched in the worksheet knowledge graph belongs, then the fault small field entity in the to-be-processed worksheet and the fault small field entity in the worksheet knowledge graph may be subjected to character matching, a second initial candidate worksheet list may be determined according to the worksheet to which the fault small field entity successfully matched in the worksheet knowledge graph belongs, and an initial candidate worksheet list may be determined according to the intersection of the first initial candidate worksheet list and the second initial candidate worksheet list.

Because the entities of the fault major and minor fields have consistent professional terms expression, the problem that a plurality of entities refer together does not exist, so that the fault major and minor fields can be directly subjected to character matching with the entities in the knowledge graph without calculating corresponding matched entity cluster clusters, and then a matched work order list is determined according to the matched entity cluster clusters, thereby improving the processing efficiency.

Of course, in the scenario where the fault major class and the fault minor class exist in the plurality of entity co-fingers, for the purpose of matching accuracy, the corresponding matching worksheet list may be determined through the above steps S110 to S120.

In step S920, a candidate work order list of the work order to be processed is determined according to the initial candidate work order list and the matching work order list of each of the first target fields.

For example, one specific embodiment of step S920 may include: performing multiple intersection operations on the initial candidate list and the matched list of each first target field according to a second preset intersection operation sequence, wherein the second preset intersection operation sequence is used for indicating the sequence of the intersection operations between the matched list of each first target field and the initial candidate list; and determining a candidate work order list of the work orders to be processed according to the intersection operation result.

For example, the intersection operation sequence of the matching work order list corresponding to the first target field and the initial candidate work order list may be determined according to the importance degree of the first target field to the work order solution, so as to obtain a second preset intersection operation sequence.

If the intersection sequence of the second preset intersection operation is field 1, field 2, field 3 and field 4, performing intersection operation on the matching work order list corresponding to field 1 and the initial candidate work order list to obtain set 1, performing intersection operation on the matching work order list corresponding to field 2 and set 1 to obtain set 2, performing intersection operation on the matching work order list corresponding to field 3 and set 2 to obtain set 3, and performing intersection operation on the matching work order list corresponding to field 4 and set 3 to obtain set 4.

For example, when the result of the current intersection operation is null, the candidate work order list of the work order to be processed is determined according to the result of the previous intersection operation of the current intersection operation. If set 4 is empty, set 3 is a list of candidate worksheets.

Similarly, the matching worksheet list of each first target field may be calculated according to the second preset intersection operation sequence, for example, the matching worksheet list of field 1 is calculated first, and the intersection operation is performed on the matching worksheet list and the initial candidate worksheet list to obtain set 1, if set 1 is empty, the initial candidate worksheet list is the candidate worksheet list of the worksheet to be processed, and the matching worksheet lists of field 2, field 3 and field 4 do not need to be calculated.

For example, the matching work order list of each first target field and the initial candidate work order list can be intersected to obtain the candidate work order list of the work order to be processedThe following formula (1)

（1）

In the case of the formula (1),for the initial list of candidate worksheets,/for example>In order to be a list of candidate worksheets,is->Matching worksheet list of the first target fields, < >>，/>Can be determined according to the number of work order fields for entity extraction, e.g. +.>May be less than or equal to the number of work order fields in which the entity extraction is performed. And according to the second preset intersection sequence, in the process of executing the intersection calculation for a plurality of times, if a certain intersection calculation result is empty, the intersection result of the last step can be used as a candidate work order list of the work orders to be processed.

With continued reference to fig. 1, in step S140, a similarity between the to-be-processed work order and the candidate work orders in the candidate work order list is calculated, and according to the similarity, a reference recommended work order of the to-be-processed work order is determined from the candidate work orders.

Illustratively, fig. 10 shows a flow-chart entity diagram of a method of computing similarity in an exemplary embodiment of the present disclosure. Referring to fig. 10, the method may include steps S1010 to S1020. Wherein:

In step S1010, a first similarity between an entity of the fourth target field of the work order to be processed and an entity of the fourth target field of the candidate work order in the candidate work order list is calculated.

In an exemplary embodiment, the fourth target field may be the same as the first target field, or may be a union of the first target field and the third target field, or may be a part of the fields in the first target field, which is not particularly limited in this exemplary embodiment.

For example, one specific embodiment of step S1010 may include: and aiming at a special entity consisting of a plurality of information entities in a fourth target field, determining a first similarity between the special entity in the work order to be processed and the corresponding special entity in the candidate work order according to the matching condition between each information entity in the special entity in the work order to be processed and the corresponding information entity in the candidate work order.

For example, for the network element field, it is in the form of: the Provice-City-machine Room-Type, namely a network element field entity comprises a plurality of entity information, and the entity information comprises a plurality of domains ranging from large to small, so that the value domain can be normalized to be between 0 and 1 according to the complete matching condition of different network element domains among worksheets, for example, one domain is matched to obtain 0.25 score, and each domain is successfully matched to be 1 score.

In an exemplary embodiment, the cosine similarity between the entity of the fourth target field of the work order to be processed and the entity of the fourth target field of the candidate work order in the candidate work order list may also be calculated and normalized to be within the [0,1] interval, so as to determine the first similarity of each entity.

In step S1020, according to the first similarity, a similarity between the work order to be processed and the candidate work order in the candidate work order list is determined.

For example, for a field where multiple entities exist, the average of the first similarities of multiple fonts may be taken as the similarity of the field. For a field that has only one entity, then the first similarity for that entity is the first similarity for that field. And determining the similarity between the two work orders to be processed and the candidate work orders according to the average value of the first similarity of each field.

By way of example, fig. 11 shows a flow diagram of another method of calculating similarity in an exemplary embodiment of the present disclosure. Referring to fig. 11, the method may include steps S1110 to S1130. Wherein:

in step S1110, for each candidate work order in the candidate work order list, a work order pair sub-graph is determined according to the first sub-knowledge graph of the candidate work order and the second sub-knowledge graph of the work order to be processed.

For example, the work order pair sub-graph may be constructed two by two based on the graph structure information of the first sub-knowledge graph of the work order to be processed and the second sub-knowledge graph of each candidate work order.

In step S1120, the work order sub-graph is input into a pre-training graph neural network model, and node features and network topology information in the work order sub-graph are encoded through the pre-training graph neural network model, so as to obtain a second similarity between a first sub-knowledge graph and a second sub-knowledge graph in the work order sub-graph.

For example, a graph neural network model, such as GCN (Graph Convolutional Network, graph convolution network), GIN (Graph Isomorphism Network, graph isomorphic neural network), and the like, may be used to encode each entity node feature and network topology information of each sub-graph in the work order pair sub-graph, thereby outputting the similarity of the two sub-knowledge maps in the work order pair sub-graph.

The characteristics of each entity node in the work order pair sub-graph can be obtained through the language coding model Bert, and the network topology information can be obtained through the triples corresponding to the relations among the nodes in the sub-graph.

In an exemplary embodiment, the worksheets in the worksheet knowledge graph may be combined two by two in advance to generate a worksheet sub-graph, and then a label whether the worksheets match or not is added to the worksheet sub-graph according to the matching relationship of the two worksheets to generate the training data set. And then training an initial graph neural network model through the training data set to obtain a pre-training graph neural network model. The graph neural network is trained by adopting a two-class data set constructed by matching worksheets in pairs in the knowledge graph, so that the value range of the score obtained by reasoning is [0,1].

In step S1130, a similarity between the work order to be processed and the candidate work order is determined according to the second similarity.

For example, the second similarity may be determined as a similarity between the work order to be processed and the candidate work order indicated by the work order sub-graph.

For example, another embodiment for calculating the similarity between the work order to be processed and the candidate work orders in the candidate work order list may include: determining a first target similarity between the work order to be processed and the candidate work order in the candidate work order list according to the similarity between the entity of the fourth target field of the work order to be processed and the entity of the fourth target field of the candidate work order in the candidate work order list; determining a work order pair sub-graph according to a first sub-knowledge graph of the candidate work order and a second sub-knowledge graph of the work order to be processed aiming at each candidate work order in the candidate work order list; inputting the work order sub-graph into a pre-training graph neural network model, and encoding node characteristics and network topology information in the work order sub-graph through the pre-training graph neural network model to obtain a second target similarity between the work order to be processed and the candidate work order; determining fusion weights of the first target similarity and the second target similarity according to an ensemble learning algorithm; and fusing the first target similarity and the second target similarity based on the fusion weight so as to determine the similarity between the work order to be processed and the candidate work order in the candidate work order list according to the fusion result.

The first target similarity may be understood as an average value of the first similarities of the fourth target fields. The second target similarity can be understood as the similarity based on the pre-trained graph neural network model according to the method shown in fig. 11 described above. Therefore, the determination of the first target similarity may refer to the determination of the similarity in fig. 10, and the determination of the second target similarity may refer to the determination of the similarity in fig. 11, which will not be described herein.

For example, as shown in fig. 12, an algorithm of ensemble learning, such as stacking, may be used, and the learning adjustment may determine the fusion weight of the similarities calculated in fig. 10 and 11, and then the similarities calculated in fig. 10 and 11 are weighted and fused, so as to determine the similarity between the work order to be processed and the candidate work order.

The determining, according to the similarity, the reference recommended worksheet of the to-be-processed worksheets from the candidate worksheets includes: and sorting the similarity in a descending order, and determining the reference recommended work orders of the work orders to be processed according to the candidate work orders indicated by the similarity of the N sorted work orders.

For example, based on the similarity sorting of the candidate worksheets, a certain number of candidate worksheets are selected as recommended results, and a worker is assisted in analyzing fault information aiming at the current worksheets to be processed, so that a more reasonable solution is quickly generated.

By way of example, fig. 13 shows a flow diagram of another worksheet recommendation method in an exemplary embodiment of the present disclosure. Referring to fig. 13, the method may include steps S1310 to S1340. Wherein: in step S1310, case work order entity information is extracted, and a work order knowledge graph is constructed; in step S1320, clustering the entity features of each field in the knowledge graph; in step S1330, the current worksheet entity information is extracted and feature matched; in step S1340, the multi-field feature similarity is calculated, and a reference recommended work order is determined according to the similarity.

For example, in the present disclosure, entity extraction may be performed based on a case work order to build a knowledge graph. And (3) coding and clustering field entity information in the map to realize common-finger entity merging, calculating the distance from the entity characteristics of the new work order to each cluster, and selecting the entity set of the nearest cluster. And calculating the similarity of the worksheets based on the multi-field matching result and calculating the score of the subgraph formed by the current worksheets and each case worksheet based on the graph neural network, sorting the case worksheets according to the scores, and selecting an optimal case for assisting in generating a solution of the current worksheet.

Illustratively, fig. 14 shows a flow diagram of yet another worksheet recommendation method in an exemplary embodiment of the present disclosure. Referring to fig. 14, the method may include steps S1401 to S1410. Wherein: in step S1401, a simple field of the work order data is extracted, in step S1402, a complex field of the work order data is extracted, and in step S1403, a case work order knowledge graph is constructed; in step S1404, multi-field entity feature encoding; in step S1405, multi-field entity feature clustering, and in step S1406, current work order information is extracted; in step S1407, the work order multi-field entity feature matches the cluster, in step S1408, the multi-field entity feature matches the score, in step S1409, the graph network evaluates the work order sub-graph feature matches the score, in step S1410, the multi-class match scores are combined to sort the case work order to obtain the recommended match case work order.

In the method, similar entities are merged through text feature coding and feature clustering methods, so that the method is used for subsequent entity matching, the matching precision is improved, the feature matching similarity is convenient to calculate, and the convenience in case work order selection is realized.

In addition, processing by using a text coding model to obtain characteristics of each node, and adopting a graph neural network model to infer a subgraph formed by the current work order and each case work order to obtain a global score, wherein the global score is used for evaluating the matching degree of the work order pair and realizing robust work order association.

It is noted that the above-described figures are merely schematic illustrations of processes involved in a method according to exemplary embodiments of the present disclosure, and are not intended to be limiting. It will be readily appreciated that the processes shown in the above figures do not indicate or limit the temporal order of these processes. In addition, it is also readily understood that these processes may be performed synchronously or asynchronously, for example, among a plurality of modules.

Further, referring to fig. 15, in this exemplary embodiment, a worksheet recommendation device 1500 is further provided, which may include a matching entity cluster determining module 1510, a matching worksheet list determining module 1520, a candidate worksheet list determining module 1530, and a reference recommended worksheet determining module 1540. Wherein: a matched entity cluster determining module 1510 configured to determine, for a first target field of a work order to be processed, a first distance between an entity feature of the first target field and a cluster center feature of an existing entity cluster, and determine a matched entity cluster of the first target field according to the first distance; a matching worksheet list determining module 1520 configured to determine a matching worksheet list of the first target field based on the processed worksheets to which the entities included in the matching entity cluster belong; the candidate list determining module 1530 is configured to determine a candidate list of the to-be-processed worksheets according to the matched worksheets of the first target field; the reference recommended work order determining module 1540 is configured to calculate the similarity between the work order to be processed and the candidate work orders in the candidate work order list, and determine the reference recommended work order of the work order to be processed from the candidate work orders according to the similarity; the existing entity clustering cluster is obtained by clustering the entities in the processed worksheet in advance.

In an exemplary implementation manner, based on the foregoing embodiment, the clustering of the existing entity clusters by clustering the entities in the processed worksheet in advance includes: extracting the entity from the processed work order, and carrying out text coding on the extracted entity through a language coding model to obtain an entity characteristic vector; based on the distance between the entity feature vectors, the entity features are clustered according to a clustering algorithm to obtain the existing entity clustering clusters.

In an exemplary implementation manner, based on the foregoing embodiment, the apparatus may further include a work order knowledge graph construction module, where the module may be configured to: extracting the entity from the second target field of the processed work order to obtain the entity corresponding to the second target field of the processed work order; generating an entity relation triplet according to one or more of the subordinate relation between the processed work order and the second target field, the causal relation between the entities corresponding to the second target field and the time sequence relation between the entities corresponding to the second target field; and generating a work order knowledge graph according to the entity relation triplet.

In an exemplary implementation, based on the foregoing embodiment, the matching worksheet list determining module 1520 may be specifically configured to: searching a processed work order of the entity included in the matched entity cluster in the work order knowledge graph; and determining a matching work order list of the first target field according to the searched processed work orders.

In an exemplary implementation, based on the foregoing embodiments, the candidate worksheet list determination module 1530 may be specifically configured to: performing multiple intersection operations on the matched worksheet list of each first target field according to a first preset intersection operation sequence, wherein the first preset intersection operation sequence is used for indicating the sequence of intersection operations between the first target fields; and determining a candidate work order list of the work orders to be processed according to the intersection operation result.

In an exemplary implementation, based on the foregoing embodiments, the candidate worksheet list determination module 1530 may be specifically configured to: determining an initial candidate work order list of the work order to be processed from the work order knowledge graph according to a preset rule, wherein the preset rule is determined according to a third target field; and determining a candidate work order list of the work order to be processed according to the initial candidate work order list and the matched work order list of each first target field.

In an exemplary implementation manner, based on the foregoing embodiment, the determining, according to the initial candidate list and the matching list of the first target field, the candidate list of the to-be-processed work order includes: performing multiple intersection operations on the initial candidate list and the matched list of each first target field according to a second preset intersection operation sequence, wherein the second preset intersection operation sequence is used for indicating the sequence of the intersection operations between the matched list of each first target field and the initial candidate list; and determining a candidate work order list of the work orders to be processed according to the intersection operation result.

In an exemplary embodiment, based on the foregoing embodiment, the determining, according to the intersection operation result, the candidate list of the work order to be processed includes:

and when the result of the current intersection operation is empty, determining a candidate work order list of the work order to be processed according to the result of the previous intersection operation of the current intersection operation.

In an exemplary embodiment, based on the foregoing embodiment, the determining, according to the similarity, the reference recommended worksheet of the to-be-processed worksheet from the candidate worksheets includes: and sorting the similarity in a descending order, and determining the reference recommended work orders of the work orders to be processed according to the candidate work orders indicated by the similarity of the N sorted work orders.

The specific details of each module in the above apparatus are already described in the method section embodiments, and the details that are not disclosed can be found in the method section embodiments, so that they will not be described in detail.

The exemplary embodiment of the disclosure also provides an electronic device for executing the worksheet recommending method. In general, the electronic device may include a processor and a memory for storing executable instructions of the processor, the processor being configured to perform the work order recommendation method described above via execution of the executable instructions.

The configuration of the electronic device will be exemplarily described below with reference to the mobile terminal 1600 in fig. 16. It will be appreciated by those skilled in the art that the configuration of fig. 16 can be applied to stationary type devices in addition to components specifically for mobile purposes.

As shown in fig. 16, the mobile terminal 1600 may specifically include: processor 1601, memory 1602, bus 1603, mobile communication module 1604, antenna 1, wireless communication module 1605, antenna 2, display 1606, camera module 1607, audio module 1608, power module 1609, and sensor module 1610.

The processor 1601 may include one or more processing units, for example: the processor 1601 may include an AP (Application Processor ), modem processor, GPU (Graphics Processing Unit, graphics processor), ISP (Image Signal Processor ), controller, encoder, decoder, DSP (Digital Signal Processor ), baseband processor and/or NPU (Neural-Network Processing Unit, neural network processor), and the like.

The processor 1601 may be coupled with a memory 1602 or other component by a bus 1603.

Memory 1602 may be used to store computer-executable program code that includes instructions. The processor 1601 performs various functional applications and data processing of the mobile terminal 1600 by executing instructions stored in the memory 1602. The memory 1602 may also store application data, such as files storing images, video, audio, etc., and the memory 1602 may also store existing entity clusters and cluster centers of existing entity clusters, as well as worksheet knowledge maps.

The communication functions of the mobile terminal 1600 may be implemented by the mobile communication module 1604, the antenna 1, the wireless communication module 1605, the antenna 2, a modem processor, a baseband processor, and the like. The antennas 1 and 2 are used for transmitting and receiving electromagnetic wave signals. The mobile communication module 1604 may provide a 2G, 3G, 4G, 5G, etc. mobile communication solution for application on the mobile terminal 1600. The wireless communication module 1605 may provide wireless communication solutions for wireless local area networks, bluetooth, near field communications, etc., that are employed on the mobile terminal 1600.

The display 1606 is used to implement display functions such as displaying user interfaces, images, displaying reference recommended worksheets, and the like. The image capturing module 1607 is used to implement a capturing function, such as capturing an image, video, and the like. The audio module 1608 is used to implement audio functions, such as playing audio, etc. The power module 1609 is used to implement power management functions such as charging a battery, powering a device, monitoring battery status, etc. The sensor module 1610 may include a depth sensor 16101, a speed sensor 16102, a gyro sensor 16103, a barometric sensor 16104, etc. to implement a corresponding sensing function.

Those skilled in the art will appreciate that the various aspects of the present disclosure may be implemented as a system, method, or program product. Accordingly, various aspects of the disclosure may be embodied in the following forms, namely: an entirely hardware embodiment, an entirely software embodiment (including firmware, micro-code, etc.) or an embodiment combining hardware and software aspects may be referred to herein as a "circuit," module "or" system.

In particular, according to embodiments of the present disclosure, the processes described above with reference to flowcharts may be implemented as computer software programs. For example, embodiments of the present disclosure include a computer program product comprising a computer program embodied on a computer readable storage medium, the computer program comprising program code for performing the method shown in the flowcharts. In such embodiments, the computer program may be downloaded and installed from a network, and/or installed based on removable media. The computer program, when executed by a Central Processing Unit (CPU), performs the various functions defined in the method and apparatus of the present application.

Exemplary embodiments of the present disclosure also provide a computer-readable storage medium having stored thereon a program product capable of implementing the method described above in the present specification. In some possible implementations, various aspects of the disclosure may also be implemented in the form of a program product comprising program code for causing a terminal device to carry out the steps according to the various exemplary embodiments of the disclosure as described in the "exemplary methods" section of this specification, e.g. any one or more of the steps of fig. 1, when the program product is run on the terminal device.

The computer readable medium shown in the present disclosure may be a computer readable signal medium or a computer readable storage medium or any combination of the two. The computer readable storage medium can be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or a combination of any of the foregoing. More specific examples of the computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

In the context of this disclosure, a computer-readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. In the present disclosure, however, the computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, with the computer-readable program code embodied therein. Such a propagated data signal may take any of a variety of forms, including, but not limited to, electro-magnetic, optical, or any suitable combination of the foregoing. A computer readable signal medium may also be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, fiber optic cable, RF, etc., or any suitable combination of the foregoing.

Furthermore, the program code for carrying out operations of the present disclosure may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, C++ or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computing device, partly on the user's device, as a stand-alone software package, partly on the user's computing device, partly on a remote computing device, or entirely on the remote computing device or server. In the case of remote computing devices, the remote computing device may be connected to the user computing device through any kind of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or may be connected to an external computing device (e.g., connected via the Internet using an Internet service provider).

Other embodiments of the disclosure will be apparent to those skilled in the art from consideration of the specification and practice of the disclosure disclosed herein. This application is intended to cover any adaptations, uses, or adaptations of the disclosure following, in general, the principles of the disclosure and including such departures from the present disclosure as come within known or customary practice within the art to which the disclosure pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the disclosure being indicated by the following claims.

It is to be understood that the present disclosure is not limited to the precise arrangements and instrumentalities shown in the drawings, and that various modifications and changes may be effected without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. A work order recommendation method, comprising:

determining a first distance between an entity characteristic of a first target field and a clustering center characteristic of an existing entity cluster aiming at the first target field of a work order to be processed, and determining a matching entity cluster of the first target field according to the first distance;

determining a matching work order list of the first target field based on the processed work orders to which the entities included in the matching entity cluster belong; extracting the entity from the second target field of the processed work order to obtain the entity corresponding to the second target field of the processed work order; generating an entity relation triplet according to one or more of the subordinate relation between the processed work order and the second target field, the causal relation between the entities corresponding to the second target field and the time sequence relation between the entities corresponding to the second target field; generating a work order knowledge graph according to the entity relation triplet; the work order knowledge graph is used for determining a matched work order list of the first target field;

Determining a candidate work order list of the work order to be processed according to the matched work order list of the first target field;

calculating the similarity between the work order to be processed and the candidate work orders in the candidate work order list, and determining a reference recommended work order of the work order to be processed from the candidate work orders according to the similarity;

2. The worksheet recommendation method of claim 1, wherein the clustering of the existing entity clusters by pre-clustering the entities in the processed worksheets comprises:

extracting the entity from the processed work order, and carrying out text coding on the extracted entity through a language coding model to obtain an entity characteristic vector;

based on the distance between the entity feature vectors, the entity features are clustered according to a clustering algorithm to obtain the existing entity clustering clusters.

3. The worksheet recommendation method of claim 1, wherein the determining the matching worksheet list of the first target field based on the processed worksheets to which the entities included in the matching entity cluster belong comprises:

Searching a processed work order of the entity included in the matched entity cluster in the work order knowledge graph;

and determining a matching work order list of the first target field according to the searched processed work orders.

4. The worksheet recommendation method of claim 1, wherein determining the candidate worksheet list of the to-be-processed worksheets according to the matching worksheet list of the first target field comprises:

performing multiple intersection operations on the matched worksheet list of each first target field according to a first preset intersection operation sequence, wherein the first preset intersection operation sequence is used for indicating the sequence of intersection operations between the first target fields;

and determining a candidate work order list of the work orders to be processed according to the intersection operation result.

5. The worksheet recommendation method of claim 4, wherein determining the candidate worksheet list of the to-be-processed worksheets according to the matching worksheet list of the first target field comprises:

determining an initial candidate work order list of the work order to be processed from the work order knowledge graph according to a preset rule, wherein the preset rule is determined according to a third target field;

And determining a candidate work order list of the work order to be processed according to the initial candidate work order list and the matched work order list of each first target field.

6. The method of claim 5, wherein determining the candidate list of the work order to be processed according to the initial candidate list of the work order and the matching list of the work order in the first target field comprises:

performing multiple intersection operations on the initial candidate list and the matched list of each first target field according to a second preset intersection operation sequence, wherein the second preset intersection operation sequence is used for indicating the sequence of the intersection operations between the matched list of each first target field and the initial candidate list;

7. The worksheet recommendation method according to claim 4 or 5, wherein determining the candidate worksheet list of the worksheets to be processed according to the result of the intersection operation comprises:

8. The work order recommendation method according to claim 1, wherein the calculating the similarity between the work order to be processed and the candidate work orders in the candidate work order list includes:

calculating a first similarity between an entity of a fourth target field of the work order to be processed and an entity of a fourth target field of a candidate work order in the candidate work order list;

and determining the similarity between the work order to be processed and the candidate work orders in the candidate work order list according to the first similarity.

9. The work order recommendation method according to claim 1, wherein the calculating the similarity between the work order to be processed and the candidate work orders in the candidate work order list includes:

determining a work order pair sub-graph according to a first sub-knowledge graph of the candidate work order and a second sub-knowledge graph of the work order to be processed aiming at each candidate work order in the candidate work order list;

inputting the work order sub-graph into a pre-training graph neural network model, and encoding node characteristics and network topology information in the work order sub-graph through the pre-training graph neural network model to obtain a second similarity between a first sub-knowledge graph and a second sub-knowledge graph in the work order sub-graph;

And determining the similarity between the work order to be processed and the candidate work order according to the second similarity.

10. The work order recommendation method according to claim 1, wherein the calculating the similarity between the work order to be processed and the candidate work orders in the candidate work order list includes:

determining a first target similarity between the work order to be processed and the candidate work order in the candidate work order list according to the similarity between the entity of the fourth target field of the work order to be processed and the entity of the fourth target field of the candidate work order in the candidate work order list;

inputting the work order sub-graph into a pre-training graph neural network model, and encoding node characteristics and network topology information in the work order sub-graph through the pre-training graph neural network model to obtain a second target similarity between the work order to be processed and the candidate work order;

determining fusion weights of the first target similarity and the second target similarity according to an ensemble learning algorithm;

And fusing the first target similarity and the second target similarity based on the fusion weight so as to determine the similarity between the work order to be processed and the candidate work order in the candidate work order list according to the fusion result.

11. The worksheet recommendation method according to claim 1, wherein the determining the reference recommended worksheet of the to-be-processed worksheets from the candidate worksheets according to the similarity comprises:

and sorting the similarity in a descending order, and determining the reference recommended work orders of the work orders to be processed according to the candidate work orders indicated by the similarity of the N sorted work orders.

12. A work order recommendation device, comprising:

the matching entity cluster determining module is configured to determine, for a first target field of a work order to be processed, a first distance between an entity feature of the first target field and a cluster center feature of an existing entity cluster, and determine a matching entity cluster of the first target field according to the first distance;

the matching work order list determining module is configured to determine a matching work order list of the first target field based on the processed work orders of the entities included in the matching entity cluster; extracting the entity from the second target field of the processed work order to obtain the entity corresponding to the second target field of the processed work order; generating an entity relation triplet according to one or more of the subordinate relation between the processed work order and the second target field, the causal relation between the entities corresponding to the second target field and the time sequence relation between the entities corresponding to the second target field; generating a work order knowledge graph according to the entity relation triplet; the work order knowledge graph is used for determining a matched work order list of the first target field;

The candidate work order list determining module is configured to determine a candidate work order list of the work order to be processed according to the matched work order list of the first target field;

the reference recommended work order determining module is configured to calculate the similarity between the work order to be processed and the candidate work orders in the candidate work order list, and determine the reference recommended work order of the work order to be processed from the candidate work orders according to the similarity;

13. A computer readable storage medium, on which a computer program is stored, which computer program, when being executed by a processor, implements the method according to any one of claims 1 to 11.

14. An electronic device, comprising:

one or more processors; and

a memory for storing one or more programs that, when executed by the one or more processors, cause the one or more processors to implement the method of any of claims 1-11.