CN114117029A - Solution recommendation method and system based on multi-level information enhancement - Google Patents
Solution recommendation method and system based on multi-level information enhancement Download PDFInfo
- Publication number
- CN114117029A CN114117029A CN202111407122.9A CN202111407122A CN114117029A CN 114117029 A CN114117029 A CN 114117029A CN 202111407122 A CN202111407122 A CN 202111407122A CN 114117029 A CN114117029 A CN 114117029A
- Authority
- CN
- China
- Prior art keywords
- work order
- vector
- representation
- information
- title
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 49
- 239000013598 vector Substances 0.000 claims abstract description 110
- 239000013604 expression vector Substances 0.000 claims abstract description 48
- 238000000605 extraction Methods 0.000 claims abstract description 16
- 230000006870 function Effects 0.000 claims abstract description 14
- 230000007246 mechanism Effects 0.000 claims abstract description 13
- 238000003860 storage Methods 0.000 claims description 9
- 230000004927 fusion Effects 0.000 claims description 5
- 230000003044 adaptive effect Effects 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 230000008569 process Effects 0.000 abstract description 8
- 238000010586 diagram Methods 0.000 description 10
- 238000004590 computer program Methods 0.000 description 8
- 238000012545 processing Methods 0.000 description 6
- 238000004519 manufacturing process Methods 0.000 description 3
- 239000000126 substance Substances 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000005065 mining Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000007500 overflow downdraw method Methods 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004140 cleaning Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/335—Filtering based on additional data, e.g. user or group profiles
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/258—Heading extraction; Automatic titling; Numbering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Data Mining & Analysis (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Databases & Information Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides a solution recommendation method and system based on multi-level information enhancement, which respectively extract the characteristics of the obtained work order title, work order description and customer information and perform embedded representation learning on the characteristic extraction result; according to the interval dot product attention mechanism and the embedded representation learning result, a work order title representation vector and a work order description representation vector are obtained; obtaining a local information representation vector according to the work order title representation vector, the work order description representation vector and the one-dimensional convolution model; obtaining a global information representation vector according to the work order title representation vector, the work order description representation vector and the autoregressive model; obtaining a comprehensive expression vector according to the local information expression vector and the global information expression vector; obtaining a solution recommendation result according to the comprehensive expression vector and a preset recommendation function model; the method and the system can automatically analyze and process the work order data, dig out the potential appeal of the client and realize accurate recommendation of the solution corresponding to the potential appeal.
Description
Technical Field
The invention relates to the technical field of electric power worksheet data processing, in particular to a solution recommendation method and system based on multi-level information enhancement.
Background
The statements in this section merely provide background information related to the present disclosure and may not constitute prior art.
The power grid enterprise business is complex, a large amount of semi-structured and unstructured text data exist in enterprise production and operation, a large amount of text data exist in the power grid enterprise production and operation process, and the text data relate to various professional fields such as electronics, chemistry, machinery and information, for example, a maintenance report in electric power contains a plurality of professional fields of professional equipment related to machinery, chemistry, physics, electronics and the like, and relate to various professional knowledge. The text data belongs to low-density value data, has the characteristics of large data volume, complex structure, lack of specification and the like, and is one of the difficult areas of data analysis and mining at present.
The work orders in the power 186 customer service system belong to the typical data, the text data mainly adopts a spoken description form and records a large amount of power service characteristics, but the text also contains a large amount of power professional terms, the format of the text data is not uniform, the content difference is large, the content of the work orders is mainly processed and classified by judgment of seat personnel at present, and due to the fact that manual experience is relied on, the processing timeliness is low, the classification rules are inconsistent, and the real appeal of a client cannot be effectively found.
Artificial intelligence and text mining technologies are gradually applied to various scenes of electric power, the way usually stays in representation learning of text shallow features, but lacks fine-grained and deep semantic understanding, and the shallow representation method is difficult to comprehensively capture hidden semantic information contained in work order description, so that the recommendation performance of a solution is reduced.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention provides a solution recommending method and system based on multi-level information enhancement, which can automatically analyze and process the work order data, extract the potential appeal of the client and realize the accurate recommendation of the solution corresponding to the potential appeal.
In order to achieve the purpose, the invention adopts the following technical scheme:
the invention provides a solution recommendation method based on multi-level information enhancement.
A solution recommendation method based on multi-level information enhancement comprises the following processes:
acquiring a work order title, a work order description and customer information in the work order data;
respectively extracting the characteristics of the obtained work order title, the work order description and the client information, and performing embedded representation learning on the characteristic extraction result;
according to the interval dot product attention mechanism and the embedded representation learning result, a work order title representation vector and a work order description representation vector are obtained;
obtaining a local information representation vector according to the work order title representation vector, the work order description representation vector and the one-dimensional convolution model;
obtaining a global information representation vector according to the work order title representation vector, the work order description representation vector and the autoregressive model;
obtaining a comprehensive expression vector according to the local information expression vector and the global information expression vector;
and obtaining a solution recommendation result according to the comprehensive expression vector and a preset recommendation function model.
Further, a characteristic word extraction algorithm based on a graph model is adopted to respectively extract the keyword characteristics of the work order title, the work order description and the customer information.
Further, according to the feature extraction result, a work order title sequence, a work order description sequence and a customer information sequence are respectively obtained, and the work order title sequence, the work order description sequence and the customer information sequence are subjected to embedding representation learning through the embedded vectors to obtain a work order title vector, a work order description vector and a customer information vector.
Furthermore, according to the interval dot product attention mechanism, the incidence relation between the customer information vector and the work order title vector and the incidence relation between the customer information and the work order description vector are learned to obtain the work order title expression vector and the work order description expression vector.
Further, the autoregressive model is a Transformer model.
Further, the local information expression vector and the global information expression vector are subjected to self-adaptive fusion to obtain a comprehensive expression vector of the work order data.
Further, the adaptive fusion includes:
Out=Norm(Concat(ACT-Skip([CL,OL]),ACT-Skip([CD,OD])))
ACT-Skip([CL,OL])=OL*σ(CL+OL)+(1-σ(CL+OL))*CL
ACT-Skip([CD,OD])=OD*σ(CD+OD)+(1-σ(CD+OD))*CD
wherein σ (-) denotes a sigmoid function, Norm (-) denotes a layer normalization operation, CLRepresenting vectors for local information of work order headers, CDDescribing local information representation vectors for work orders, OLRepresenting vectors for work order header global information, ODThe work order describes a global information representation vector.
The invention provides a solution recommendation system based on multi-level information enhancement.
A solution recommendation system based on multi-level information enhancement, comprising:
a data acquisition module configured to: acquiring a work order title, a work order description and customer information in the work order data;
a vector representation module configured to: respectively extracting the characteristics of the obtained work order title, the work order description and the client information, and performing embedded representation learning on the characteristic extraction result;
an attention mechanics learning module configured to: according to the interval dot product attention mechanism and the embedded representation learning result, a work order title representation vector and a work order description representation vector are obtained;
a local representation vector acquisition module configured to: obtaining a local information representation vector according to the work order title representation vector, the work order description representation vector and the one-dimensional convolution model;
a global representation vector acquisition module configured to: obtaining a global information representation vector according to the work order title representation vector, the work order description representation vector and the autoregressive model;
a comprehensive representation vector acquisition module configured to: obtaining a comprehensive expression vector according to the local information expression vector and the global information expression vector;
a solution recommendation module configured to: and obtaining a solution recommendation result according to the comprehensive expression vector and a preset recommendation function model.
A third aspect of the present invention provides a computer-readable storage medium having stored thereon a program that, when being executed by a processor, performs the steps of the above-mentioned solution recommendation method based on multi-level information enhancement.
A fourth aspect of the present invention provides an electronic device, which includes a memory, a processor, and a program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the solution recommendation method based on multi-level information enhancement.
Compared with the prior art, the invention has the beneficial effects that:
1. according to the method, the system, the medium or the electronic equipment, the keyword extraction is respectively carried out from three levels of the work order title, the work order description, the customer information and the like, and the interval click attention mechanism is introduced to respectively learn the incidence relation between the customer information and the work order title vector and the incidence relation between the customer information and the work order description vector, so that the expression vector of the work order data is obtained, and the solution recommendation performance is improved.
2. According to the method, the system, the medium or the electronic equipment, local dependency information mining is respectively carried out on the work order title expression vector and the work order description expression vector by utilizing one-dimensional convolution operation, and the local information expression vector is generated; meanwhile, capturing global information of the work order title representation vector and the work order description representation vector by using an autoregressive model (Transformer), and generating a global information representation vector; and the local information vector and the global information vector of the work order data are combined by using a self-adaptive information fusion method to generate a comprehensive expression vector of the work order data, so that the solution recommendation performance is further improved.
Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, are included to provide a further understanding of the invention, and are incorporated in and constitute a part of this specification, illustrate exemplary embodiments of the invention and together with the description serve to explain the invention and not to limit the invention.
Fig. 1 is a schematic diagram of an overall training flow of a solution recommendation method provided in embodiment 1 of the present invention.
Fig. 2 is a flowchart of a solution recommendation embodiment provided in embodiment 1 of the present invention.
Fig. 3 is a schematic structural diagram of a solution recommendation system provided in embodiment 1 of the present invention.
Detailed Description
The invention is further described with reference to the following figures and examples.
It is to be understood that the following detailed description is exemplary and is intended to provide further explanation of the invention as claimed. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of exemplary embodiments according to the invention. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, and it should be understood that when the terms "comprises" and/or "comprising" are used in this specification, they specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof, unless the context clearly indicates otherwise.
The embodiments and features of the embodiments of the present invention may be combined with each other without conflict.
Example 1:
embodiment 1 of the present invention provides a solution recommendation method based on multi-level information enhancement, including the following processes:
acquiring a work order title, a work order description and customer information in the work order data;
respectively extracting the characteristics of the obtained work order title, the work order description and the client information, and performing embedded representation learning on the characteristic extraction result;
according to the interval dot product attention mechanism and the embedded representation learning result, a work order title representation vector and a work order description representation vector are obtained;
obtaining a local information representation vector according to the work order title representation vector, the work order description representation vector and the one-dimensional convolution model;
obtaining a global information representation vector according to the work order title representation vector, the work order description representation vector and the autoregressive model;
obtaining a comprehensive expression vector according to the local information expression vector and the global information expression vector;
and obtaining a solution recommendation result according to the comprehensive expression vector and a preset recommendation function model.
When model training is carried out, firstly, based on historical electric power customer service data of a user, keyword feature extraction is carried out from three levels of work order titles, work order descriptions, customer information and the like by adopting a Text Rank algorithm; then, a click attention mechanism is introduced to enhance the representation of the title information and the description information in the work order data, and the client information is utilized to inquire and reconstruct the title information and the description information in the work order data, so that a work order title and a description representation vector are generated; based on the obtained work order header and the expression vector of the work order description, capturing local time sequence dependency and global time sequence dependency information in the work order data by using one-dimensional convolution and a Transformer; and finally, fusing the obtained global information and the local information by adopting a self-adaptive fusion method for solution recommendation, thereby improving the recommendation performance. The specific steps are shown in fig. 1 and fig. 2, and include:
s1: the method comprises the steps of collecting relevant power customer service data of mass power information, wherein the relevant power customer service data comprise historical work order data, customer information, corresponding solutions and the like, and preprocessing the obtained customer service data, including data cleaning, missing data completion, data definition and storage.
Specifically, 16259 pieces of work order data are totally counted by 3638 clients from 5/1/2021 to 5/31/2021 based on data acquired from a customer service system of a certain company in the electric power 186. The data includes data information such as work order title, work order description, customer information and corresponding solutions.
S2: based on historical customer service data, firstly, a TextRank algorithm is adopted for extracting keywords, and then, an embedded vector is adopted for information embedding expression learning.
S2.1: the implementation adopts a characteristic word extraction method (TextRank algorithm) based on a graph model to extract key characteristic words. The text is regarded as being composed of a plurality of words, a corresponding graph model is established, then important words in the text are sequenced by using a voting mechanism, and therefore the extraction of the characteristic words can be realized only by depending on the structural relationship of the text, and the method is simple, effective and quite wide in application.
The weight iteration formula of the algorithm is shown as the formula:
wherein d is an adjustment coefficient, typically having a value of 0.85; in (V)i) Indicates a point ViSet of all nodes of a node, Out (V)j) Refers to VjA set of nodes to which the node points; | Out (V)j) And | refers to the number of nodes in the set.
S2.2: defining a work order title sequence based on the S2.1 keyword feature extraction resultWork order description sequenceAnd a customer information sequenceWherein M, N, S respectively represent the word number of the longest sentence in the title sequence, the customer information sequence and the work order description sequence, and T represents the work order number.
The information is embedded and represented by the embedded vector, and the method comprises the following steps:
EL=LWL
EP=PWP
ED=DWD
s3: based on the work order title, the work order description and the embedded vector of the customer information obtained in S2, a space Dot Product Attention mechanism (Scaled Dot-Product Attention) is introduced to learn the association relationship between the customer information and the work order title and the association relationship between the customer information and the work order description respectively, so as to generate a representation vector R of the work order title information and the description informationLAnd RD:
Where Norm (·) denotes the layer normalization operation.
S4: generating a local information representation vector by adopting one-dimensional convolution operation based on the work order title representation vector and the work order description representation vector obtained in the step S3; and simultaneously adopting a Transformer to generate a global information representation vector.
On one hand, one-dimensional convolution operation is adopted to respectively mine local time sequence dependence information in the work order title and the work order description, and a local context information representation vector C is generatedLAnd CD。
In this embodiment, the inventors convolveThe size of the kernel is set to [2, d ]embedding]:
CL=Conv1D(RC)
DD=Conv1D(RD)
On the other hand, global information of the work order title expression vector and the work order description expression vector is captured by adopting a Transformer, and the work order title and the work order description global information expression vector O are respectively generatedLAnd OD。
Global information representation vector O for work order headerLThe method is realized as follows:
Q′L=K′L=V′L=EL
OL=Norm(FFN(Norm(MultiHead(Q′L,K′L,V′L)+EL))+Norm(MultiHead(Q′L,K′L,V′L)+EL))
wherein the content of the first and second substances,all represent a weight parameter matrix, FFN (-) represents a feed-forward neural network.
Global information representation vector O described by work orderD:
Q′D=K′D=V′D=ED
OD=Norm(FFN(Norm(MultiHead(Q′D,K′D,V′D)+ED))+Norm(MultiHeas(Q′D,K′D,V′D)+ED))
Wherein the content of the first and second substances,all represent a weight parameter matrix, FFN (-) represents a feed-forward neural network.
S5: in order to comprehensively combine the obtained global context information representation with the information representation of the local data, the embodiment provides an adaptive information fusion mode to generate a comprehensive representation vector Out of the work order data:
Out=Norm(Concat(ACT-Skip([CL,OL]),ACT-Skip([CD,OD])))
ACT-Skip([CL,OL])=OL*σ(CL+OL)+(1-σ(CL+OL))*CL
ACT-Skip([CD,OD])=OD*σ(CD+OD)+(1-σ(CD+OD))*CD
where σ (·) denotes a sigmoid function.
S6: based on the comprehensive expression vector Out obtained in S5, solution recommendation is performed on the sample to be detected, and a softmax layer is input for result recommendation:
y′=Softmax(MLP(Out))
wherein y' represents the final recommendation; MLP (-) denotes multi-layer perceptron operation.
The method comprises the steps of recommending a power customer service solution to a sample to be tested, pushing a recommendation result, and comparing the result with an actual use condition, wherein accuracy, Micro F1 and AUROC are used as evaluation indexes of a recommendation method, and the comparison result is shown in Table 1.
TABLE 1 recommended method comparison
Model | Accuracy of | Micro F1 | AUROC |
GRU | 0.2833 | 0.1335 | 0.5853 |
GRU-Att | 0.3164 | 0.1726 | 0.6166 |
BiGRU | 0.2786 | 0.1441 | 0.5932 |
Conv-GRU | 0.3733 | 0.2132 | 0.6489 |
Conv-BiGRU | 0.3705 | 0.2096 | 0.6449 |
Transformer-E | 0.3662 | 0.2189 | 0.6557 |
Ours | 0.3994 | 0.2434 | 0.6739 |
Based on the results in table 1, the proposed recommendation method of this example is superior in performance to the other methods.
Example 2:
as shown in fig. 3, embodiment 2 of the present invention provides a solution recommendation system based on multi-level information enhancement, including:
a data acquisition module configured to: acquiring a work order title, a work order description and customer information in the work order data;
a vector representation module configured to: respectively extracting the characteristics of the obtained work order title, the work order description and the client information, and performing embedded representation learning on the characteristic extraction result;
an attention mechanics learning module configured to: according to the interval dot product attention mechanism and the embedded representation learning result, a work order title representation vector and a work order description representation vector are obtained;
a local representation vector acquisition module configured to: obtaining a local information representation vector according to the work order title representation vector, the work order description representation vector and the one-dimensional convolution model;
a global representation vector acquisition module configured to: obtaining a global information representation vector according to the work order title representation vector, the work order description representation vector and the autoregressive model;
a comprehensive representation vector acquisition module configured to: obtaining a comprehensive expression vector according to the local information expression vector and the global information expression vector;
a solution recommendation module configured to: and obtaining a solution recommendation result according to the comprehensive expression vector and a preset recommendation function model.
Example 3:
embodiment 3 of the present invention provides a computer-readable storage medium, on which a program is stored, which, when executed by a processor, implements the steps in the solution recommendation method based on multi-level information enhancement described in embodiment 1.
Example 4:
embodiment 4 of the present invention provides an electronic device, which includes a memory, a processor, and a program stored in the memory and executable on the processor, where the processor implements the steps of the solution recommendation method based on multi-level information enhancement described in embodiment 1 when executing the program.
As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of a hardware embodiment, a software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention 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, optical storage, and the like) having computer-usable program code embodied therein.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. 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 data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing 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 data processing apparatus 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 data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by a computer program, which can be stored in a computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. The storage medium may be a magnetic disk, an optical disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), or the like.
The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (10)
1. A solution recommendation method based on multi-level information enhancement is characterized by comprising the following steps:
the method comprises the following steps:
acquiring a work order title, a work order description and customer information in the work order data;
respectively extracting the characteristics of the obtained work order title, the work order description and the client information, and performing embedded representation learning on the characteristic extraction result;
according to the interval dot product attention mechanism and the embedded representation learning result, a work order title representation vector and a work order description representation vector are obtained;
obtaining a local information representation vector according to the work order title representation vector, the work order description representation vector and the one-dimensional convolution model;
obtaining a global information representation vector according to the work order title representation vector, the work order description representation vector and the autoregressive model;
obtaining a comprehensive expression vector according to the local information expression vector and the global information expression vector;
and obtaining a solution recommendation result according to the comprehensive expression vector and a preset recommendation function model.
2. The multi-level information enhancement based solution recommendation method of claim 1, wherein:
and respectively extracting the work order title, the work order description and the keyword feature of the customer information by adopting a feature word extraction algorithm based on a graph model.
3. The multi-level information enhancement based solution recommendation method of claim 1, wherein:
and respectively obtaining a work order title sequence, a work order description sequence and a customer information sequence according to the feature extraction result, and performing embedded representation learning on the work order title sequence, the work order description sequence and the customer information sequence through the embedded vector to obtain a work order title vector, a work order description vector and a customer information vector.
4. The multi-level information enhancement based solution recommendation method of claim 3, wherein:
and according to the interval dot product attention mechanism, learning the incidence relation between the customer information vector and the work order title vector and the incidence relation between the customer information and the work order description vector to obtain a work order title expression vector and a work order description expression vector.
5. The multi-level information enhancement based solution recommendation method of claim 1, wherein:
the autoregressive model is a Transformer model.
6. The multi-level information enhancement based solution recommendation method of claim 1, wherein:
and carrying out self-adaptive fusion on the local information expression vector and the global information expression vector to obtain a comprehensive expression vector of the work order data.
7. The multi-level information enhancement based solution recommendation method of claim 6, wherein:
the adaptive fusion comprises the following steps:
Out=Norm(Concat(ACT-Skip([CL,OL]),ACT-Skip([CD,OD])))
ACT-Skip([CL,OL])=OL*σ(CL+OL)+(1-σ(CL+OL))*CL
ACT-Skip([CD,OD])=OD*σ(CD+OD)+(1-σ(CD+OD))*CD
wherein σ (-) denotes a sigmoid function, Norm (-) denotes a layer normalization operation, CLRepresenting vectors for local information of work order headers, CDDescribing local information representation vectors for work orders, OLRepresenting vectors for work order header global information, ODThe work order describes a global information representation vector.
8. A solution recommendation system based on multi-level information enhancement is characterized in that:
the method comprises the following steps:
a data acquisition module configured to: acquiring a work order title, a work order description and customer information in the work order data;
a vector representation module configured to: respectively extracting the characteristics of the obtained work order title, the work order description and the client information, and performing embedded representation learning on the characteristic extraction result;
an attention mechanics learning module configured to: according to the interval dot product attention mechanism and the embedded representation learning result, a work order title representation vector and a work order description representation vector are obtained;
a local representation vector acquisition module configured to: obtaining a local information representation vector according to the work order title representation vector, the work order description representation vector and the one-dimensional convolution model;
a global representation vector acquisition module configured to: obtaining a global information representation vector according to the work order title representation vector, the work order description representation vector and the autoregressive model;
a comprehensive representation vector acquisition module configured to: obtaining a comprehensive expression vector according to the local information expression vector and the global information expression vector;
a solution recommendation module configured to: and obtaining a solution recommendation result according to the comprehensive expression vector and a preset recommendation function model.
9. A computer-readable storage medium having a program stored thereon, wherein the program, when executed by a processor, implements the steps of the multi-level information enhancement based solution recommendation method according to any one of claims 1-7.
10. An electronic device comprising a memory, a processor and a program stored on the memory and executable on the processor, wherein the processor executes the program to implement the steps of the solution recommendation method based on multi-level information enhancement according to any one of claims 1-7.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111407122.9A CN114117029B (en) | 2021-11-24 | 2021-11-24 | Solution recommendation method and system based on multi-level information enhancement |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111407122.9A CN114117029B (en) | 2021-11-24 | 2021-11-24 | Solution recommendation method and system based on multi-level information enhancement |
Publications (2)
Publication Number | Publication Date |
---|---|
CN114117029A true CN114117029A (en) | 2022-03-01 |
CN114117029B CN114117029B (en) | 2023-11-24 |
Family
ID=80372706
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111407122.9A Active CN114117029B (en) | 2021-11-24 | 2021-11-24 | Solution recommendation method and system based on multi-level information enhancement |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114117029B (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114493377A (en) * | 2022-04-06 | 2022-05-13 | 广州平云小匠科技有限公司 | Work order dispatching method and system |
CN115757528A (en) * | 2023-01-06 | 2023-03-07 | 卡斯柯信号(北京)有限公司 | Information recommendation method and device |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111507550A (en) * | 2019-01-30 | 2020-08-07 | 广州泰迪智能科技有限公司 | Automatic recommendation method for optimal solution of work order problem |
US20210027178A1 (en) * | 2019-07-26 | 2021-01-28 | Ricoh Company, Ltd. | Recommendation method and recommendation apparatus based on deep reinforcement learning, and non-transitory computer-readable recording medium |
CN112925904A (en) * | 2021-01-27 | 2021-06-08 | 天津大学 | Lightweight text classification method based on Tucker decomposition |
-
2021
- 2021-11-24 CN CN202111407122.9A patent/CN114117029B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111507550A (en) * | 2019-01-30 | 2020-08-07 | 广州泰迪智能科技有限公司 | Automatic recommendation method for optimal solution of work order problem |
US20210027178A1 (en) * | 2019-07-26 | 2021-01-28 | Ricoh Company, Ltd. | Recommendation method and recommendation apparatus based on deep reinforcement learning, and non-transitory computer-readable recording medium |
CN112925904A (en) * | 2021-01-27 | 2021-06-08 | 天津大学 | Lightweight text classification method based on Tucker decomposition |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114493377A (en) * | 2022-04-06 | 2022-05-13 | 广州平云小匠科技有限公司 | Work order dispatching method and system |
CN114493377B (en) * | 2022-04-06 | 2022-07-12 | 广州平云小匠科技有限公司 | Work order dispatching method and system |
CN115757528A (en) * | 2023-01-06 | 2023-03-07 | 卡斯柯信号(北京)有限公司 | Information recommendation method and device |
Also Published As
Publication number | Publication date |
---|---|
CN114117029B (en) | 2023-11-24 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111797321B (en) | Personalized knowledge recommendation method and system for different scenes | |
Yang | The case for being automatic: introducing the automatic linear modeling (LINEAR) procedure in SPSS statistics | |
Zhao et al. | Learning hierarchical features from generative models | |
CN108647226B (en) | Hybrid recommendation method based on variational automatic encoder | |
CN109740655B (en) | Article scoring prediction method based on matrix decomposition and neural collaborative filtering | |
CN114117029B (en) | Solution recommendation method and system based on multi-level information enhancement | |
Pandey et al. | An analysis of machine learning techniques (J48 & AdaBoost)-for classification | |
Tran et al. | Mixed-variate restricted Boltzmann machines | |
CN113255842B (en) | Vehicle replacement prediction method, device, equipment and storage medium | |
Estevez-Velarde et al. | AutoML strategy based on grammatical evolution: A case study about knowledge discovery from text | |
Irissappane et al. | Leveraging GPT-2 for classifying spam reviews with limited labeled data via adversarial training | |
CN115270797A (en) | Text entity extraction method and system based on self-training semi-supervised learning | |
Hossain et al. | Reducing the effect of imbalance in text classification using SVD and GloVe with ensemble and deep learning | |
CN111709225A (en) | Event cause and effect relationship judging method and device and computer readable storage medium | |
Nandhini et al. | Association rule mining using swarm intelligence and domain ontology | |
CN114881173A (en) | Resume classification method and device based on self-attention mechanism | |
CN113239143B (en) | Power transmission and transformation equipment fault processing method and system fusing power grid fault case base | |
Farooq | Genetic algorithm technique in hybrid intelligent systems for pattern recognition | |
JP2017174004A (en) | Sentence meaning classification calculation device, model learning device, method, and program | |
CN117093849A (en) | Digital matrix feature analysis method based on automatic generation model | |
CN111275109A (en) | Power equipment state data characteristic optimization method and system based on self-encoder | |
CN114036319A (en) | Power knowledge extraction method, system, device and storage medium | |
CN116485084A (en) | Intelligent power material demand decision method and system based on data space | |
Mohanty et al. | Software reliability prediction using group method of data handling | |
Zhu et al. | Software defect prediction model based on stacked denoising auto-encoder |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |