Work ticket auditing method and device, equipment and medium

By using the pre-trained work ticket review model, and using word segmentation and feature extraction modules to automatically review work tickets, the traditional manual review speed and low accuracy are solved, and more efficient work ticket review is achieved.

CN118940753APending Publication Date: 2024-11-12GUANGDONG POWER GRID CO LTD +1
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202410987422.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-23
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

Traditional work ticket safety measures review methods rely on manual labor, resulting in slow review speed and low accuracy, affecting the accuracy of work ticket review.

Method used

The automatic audit method based on the pre-trained work ticket audit model is adopted, including word segmentation module, first feature extraction module, second feature extraction module and judgment module. The output vector and feature matrix are generated through word vector sequence and timing information, and used to determine the work ticket review results.

Benefits of technology

It improves the speed and accuracy of work ticket review, and solves the problems of slow manual review speed and low accuracy.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118940753A_ABST
    Figure CN118940753A_ABST
Patent Text Reader

Abstract

The embodiment of the invention discloses a work ticket auditing method, device and equipment and a medium, and is applied to the field of deep learning, and the method comprises the steps: inputting the text content of a to-be-audited work ticket into a pre-trained work ticket auditing model, and determining a word vector sequence according to the text content of the to-be-audited work ticket based on a word segmentation module; based on the first feature extraction module, generating output vectors corresponding to the word vectors according to the word vectors in the word vector sequence and time sequence information among the word vectors, and determining a first feature matrix according to the output vectors corresponding to the word vectors; based on the second feature extraction module, determining a second feature matrix according to the word vector sequence; and based on the judgment module, determining an audit result of the work ticket to be audited according to the first feature matrix and the second feature matrix. According to the technical scheme provided by the invention, the work ticket checking speed and checking accuracy can be improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of deep learning, and in particular to a work ticket review method, device, equipment and medium. Background Art

[0002] A work ticket is a written basis and safety license for overhaul, maintenance, installation, modification, commissioning, testing, etc. on power production sites, equipment, and systems. It is not only a directive document for overhaul or operation personnel to perform work tasks, but also one of the important management systems to ensure operational safety.

[0003] Reviewing the safety measures in work tickets is an important part of ensuring work safety and compliance. However, traditional methods of reviewing safety measures in work tickets are often manual. When processing a large number of work tickets, manual review is slow and has low accuracy, which affects the accuracy of work ticket review. Summary of the invention

[0004] The present invention provides a work ticket review method, device, equipment and medium. The technical solution provided by the present invention can solve the problems of low speed and low accuracy of manual work ticket review, and improve the speed and accuracy of work ticket review.

[0005] In a first aspect, an embodiment of the present invention discloses a work ticket review method, comprising:

[0006] Input the text content of the work ticket to be reviewed into a pre-trained work ticket review model, wherein the work ticket review model is generated based on historical work ticket sample training, and the work ticket review model includes a word segmentation module, a first feature extraction module, a second feature extraction module and a judgment module;

[0007] Based on the word segmentation module, determine the word vector sequence according to the text content of the work ticket to be reviewed;

[0008] Based on the first feature extraction module, an output vector corresponding to each word vector is generated according to each word vector in the word vector sequence and the time sequence information between the word vectors, and a first feature matrix is ​​determined according to the output vector corresponding to each word vector, wherein the output vector of any word vector is used to generate an output vector corresponding to the next word vector of the word vector;

[0009] Based on the second feature extraction module, determine a second feature matrix according to the word vector sequence;

[0010] Based on the judgment module, the audit result of the work ticket to be audited is determined according to the first characteristic matrix and the second characteristic matrix.

[0011] The second invention, the embodiment of the present invention discloses a work ticket review device, comprising:

[0012] An input module, used for inputting the text content of the work ticket to be reviewed into a pre-trained work ticket review model, wherein the work ticket review model is generated based on historical work ticket sample training, and the work ticket review model includes a word segmentation module, a first feature extraction module, a second feature extraction module and a judgment module;

[0013] A sequence generation module, used to determine a word vector sequence based on the word segmentation module and the text content of the work ticket to be reviewed;

[0014] A first extraction module, configured to generate an output vector corresponding to each word vector according to each word vector in the word vector sequence and the time sequence information between the word vectors based on the first feature extraction module, and determine a first feature matrix according to the output vector corresponding to each word vector, wherein the output vector of any word vector is used to generate an output vector corresponding to a subsequent word vector of the word vector;

[0015] A second extraction module, configured to determine a second feature matrix according to the word vector sequence based on the second feature extraction module;

[0016] An audit module is used to determine the audit result of the work ticket to be audited based on the judgment module and the first characteristic matrix and the second characteristic matrix.

[0017] In a third aspect, an embodiment of the present invention discloses an electronic device, the electronic device comprising:

[0018] at least one processor; and,

[0019] a memory communicatively connected to the at least one processor; wherein,

[0020] The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the work ticket review method described in any one of the embodiments of the present invention.

[0021] In a fourth aspect, an embodiment of the present invention discloses a computer-readable storage medium, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement any one of the work ticket review methods described in the embodiments of the present invention when executed.

[0022] An embodiment of the present invention provides a work ticket review method, device, equipment and medium, the method comprising: inputting the text content of the work ticket to be reviewed into a pre-trained work ticket review model, wherein the work ticket review model is generated based on historical work ticket sample training, and the work ticket review model comprises a word segmentation module, a first feature extraction module, a second feature extraction module and a judgment module; based on the word segmentation module, determining a word vector sequence according to the text content of the work ticket to be reviewed; based on the first feature extraction module, generating an output vector corresponding to each word vector according to each word vector in the word vector sequence and the time series information between the word vectors, and determining a first feature matrix according to the output vector corresponding to each word vector, wherein the output vector of any word vector is used to generate an output vector corresponding to the next word vector of the word vector; based on the second feature extraction module, determining a second feature matrix according to the word vector sequence; based on the judgment module, determining the review result of the work ticket to be reviewed according to the first feature matrix and the second feature matrix. Specifically, the output vectors corresponding to the word vectors in the word vector sequence and the time sequence information between the word vectors are generated, wherein the output vector of any word vector is used to generate the output vector corresponding to the next word vector of the word vector. In this way, the content features of the text content of the work ticket to be reviewed can be accurately extracted, thereby improving the accuracy of the review results of the work ticket. This method can also solve the problem of slow manual review and improve the speed of work ticket review. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0024] Figure 1 A flowchart of a work ticket review method provided by an embodiment of the present invention;

[0025] Figure 2 A schematic diagram of the structure of a SEFM provided by an embodiment of the present invention;

[0026] Figure 3 A schematic diagram of the structure of SemAT provided in an embodiment of the present invention;

[0027] Figure 4 A flowchart of a work ticket review method provided in Embodiment 2 of the present invention;

[0028] Figure 5 A schematic diagram of generating an output vector based on a word vector provided in an embodiment of the present invention;

[0029] Figure 6 A schematic diagram of a WCM provided by an embodiment of the present invention;

[0030] Figure 7 A schematic diagram of the overall work ticket review process provided by an embodiment of the present invention;

[0031] Figure 8 A schematic diagram of the structure of a work ticket review device provided in Embodiment 3 of the present invention;

[0032] Fig. 9 A schematic diagram of the structure of an electronic device provided in Embodiment 4 of the present invention. DETAILED DESCRIPTION

[0033] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0034] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0035] It should be noted that the collection, storage, use, processing, transmission, provision and disclosure of user personal information involved in the technical solution of the present disclosure are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0036] Embodiment 1

[0037] Figure 1 The flowchart of a work ticket review method provided by an embodiment of the present invention is applicable to the work ticket review of a power system. The method can be performed by a work ticket review device, which can be implemented in the form of software and / or hardware and configured in various computers or servers. Specifically, Figure 1As shown, including:

[0038] Step 110, input the text content of the work ticket to be reviewed into a pre-trained work ticket review model, wherein the work ticket review model is generated based on historical work ticket sample training, and the work ticket review model includes a word segmentation module, a first feature extraction module, a second feature extraction module and a judgment module.

[0039] Among them, the work ticket review model is a successfully trained model, which is used to determine the corresponding review results based on the text content of the work ticket to be reviewed. Exemplarily, the review results include qualified and unqualified.

[0040] Specifically, the work ticket review model includes a word segmentation module, a first feature extraction module, a second feature extraction module and a judgment module. For the specific use and function of each module, please refer to the subsequent content of the embodiments of the present invention.

[0041] Step 120: Based on the word segmentation module, determine the word vector sequence according to the text content of the work ticket to be reviewed.

[0042] Among them, the word segmentation module is used to convert the text content of the work ticket to be reviewed into word vector form and combine it into a word vector sequence.

[0043] Optionally, the text content of the work ticket to be reviewed is segmented to generate keywords for each text;

[0044] The text keywords are vectorized to generate word vectors corresponding to the text keywords, and the word vectors are combined into a word vector sequence.

[0045] Specifically, since the text content of the work ticket to be reviewed is natural language, which includes task keywords, modal particles or other non-important words, the text content can be segmented to determine the text keywords in the text content, and then the text keywords can be vectorized to obtain a word vector sequence that represents the key information of the text content of the work ticket to be reviewed.

[0046] It should be noted that, since natural language has a certain text order, each word vector in the word vector sequence corresponding to the text content also has a certain arrangement order, which is determined by the text order of the text content. The arrangement order of the word vector determined by the text order can fully consider the connection relationship between each word vector in the subsequent word vector feature extraction to improve the accuracy of feature extraction.

[0047] Step 130, based on the first feature extraction module, generate an output vector corresponding to each word vector according to each word vector in the word vector sequence and the timing information between the word vectors, and determine a first feature matrix according to the output vector corresponding to each word vector, wherein the output vector of any word vector is used to generate an output vector corresponding to the next word vector of the word vector.

[0048] Among them, the first feature extraction module is used to generate output vectors corresponding to word vectors according to the word vector sequence, and combine the output vectors into a first feature matrix. The timing information can represent the connection order between word vectors. The timing information contains the key features of the calculated word vectors, which are used to calculate the output vectors of subsequent word vectors. The output vector represents the feature extraction result of the corresponding word vector, and is also used to generate the output vector of the subsequent word vector.

[0049] Specifically, in an embodiment of the present invention, the generation of the output vector of any word vector requires the input of three vectors, including: the current word vector, the vector representing the time series information, and the output vector of the previous word vector. Therefore, the output vector of each word vector not only depends on the corresponding word vector, but also contains some features of each word vector in the past. Therefore, the output vector covers more text features of the work ticket to be reviewed. In this way, the accuracy of feature extraction of the work ticket to be reviewed can be improved, thereby improving the accuracy of the audit result.

[0050] Step 140: Based on the second feature extraction module, determine a second feature matrix according to the word vector sequence.

[0051] Optionally, this step includes:

[0052] splicing the word vectors in the word vector sequence to generate a splicing matrix;

[0053] Performing convolution operations on the splicing matrix and each preset convolution kernel to determine each convolution result;

[0054] A second feature matrix is ​​generated according to a preset activation function and the convolution results.

[0055] For example, the concatenation matrix F1 can be input into convolution kernels of sizes 2*2, 4*4 and 6*6, respectively, and then the corresponding feature matrices F2, F3 and F4 can be obtained, and then F2, F3 and F4 can be input into SFEM (Semantic feature extraction module) to generate corresponding convolution results F5, F6 and F7, and then the second feature matrix can be generated according to the convolution results F5, F6, F7 and the preset Mish activation function. Specifically, Figure 2A schematic diagram of the structure of the SEFM provided in an embodiment of the present invention, wherein T1 is a concatenation matrix, Conv1d is a 1*2 convolution kernel, SemAT (SemanticAttention, semantic attention module), BN is a normalization layer for matrix normalization, and SMH is a preset activation function, specifically Dropout is the hidden layer, and T2 is the convolution result corresponding to the convolution kernel. Figure 3 A structural diagram of SemAT provided for an embodiment of the present invention, specifically, Conv is a convolution kernel of different sizes, Softmax is normalization, and by inputting the splicing matrix into different convolution kernels for convolution operation, the correlation of the feature matrix can be improved, and the receptive field of the feature matrix can be expanded, thereby effectively extracting key semantic feature information and suppressing irrelevant noise features. At the same time, the activation function SMH avoids the gradient vanishing and gradient exploding phenomena during feature extraction.

[0056] Step 150: Based on the judgment module, determine the audit result of the work ticket to be audited according to the first characteristic matrix and the second characteristic matrix.

[0057] Optionally, the judgment module includes a fully connected layer, and the step includes:

[0058] Concatenate the first feature matrix and the second feature matrix to generate a discriminant matrix;

[0059] The discriminant matrix is ​​input into the fully connected layer, the discriminant matrix is ​​operated according to the weight and bias of the fully connected layer, and the audit result of the work ticket to be audited is obtained based on the operation result.

[0060] Specifically, the discriminant matrix can be input into the fully connected layer, and then the discriminant matrix is ​​operated according to the weight and bias of the fully connected layer, and the audit result of the work ticket to be audited is obtained based on the operation result. Among them, the structure and operation method of the fully connected layer are not limited here.

[0061] Optionally, the training process of the work ticket review model described in the embodiment of the present invention includes:

[0062] Inputting historical work ticket samples into the work ticket review model to be trained, wherein the historical work ticket samples include the text content of the historical work ticket and the actual review results of the historical work ticket;

[0063] Using a word segmentation module, determine the word vector sequence corresponding to the text content of the historical work ticket;

[0064] Using a first feature matrix extraction module to determine an output vector corresponding to each word vector in the word vector sequence, and determining a first feature matrix according to the output vector corresponding to each word vector;

[0065] Using a second feature matrix extraction module to determine the word vector sequence to determine a second feature matrix;

[0066] Inputting the first characteristic matrix and the second characteristic matrix into a judgment module to determine a prediction and review result;

[0067] Determine the loss value according to the predicted audit result, the actual audit result and the loss function of the work ticket audit model;

[0068] If the loss value does not meet the preset accuracy requirement, the model parameters of the work ticket review model are updated according to the loss value.

[0069] Specifically, in each training process, the word segmentation module is used to determine the word vector sequence corresponding to the text content of the historical work ticket; the first feature matrix extraction module is used to determine the output vector corresponding to each word vector in the word vector sequence, and the first feature matrix is ​​determined according to the output vector corresponding to each word vector; the second feature matrix extraction module is used to determine the second feature matrix of the word vector sequence; the first feature matrix and the second feature matrix are input into the judgment module to determine the predicted audit result. Furthermore, the loss value is determined according to the predicted audit result, the actual audit result and the loss function of the work ticket audit model; if the loss value does not meet the preset accuracy requirement, the model parameters of the work ticket audit model are updated according to the loss value.

[0070] An embodiment of the present invention discloses a work ticket review method, which includes: inputting the text content of the work ticket to be reviewed into a pre-trained work ticket review model, wherein the work ticket review model is generated based on historical work ticket sample training, and the work ticket review model includes a word segmentation module, a first feature extraction module, a second feature extraction module and a judgment module; based on the word segmentation module, determining a word vector sequence according to the text content of the work ticket to be reviewed; based on the first feature extraction module, generating an output vector corresponding to each word vector according to each word vector in the word vector sequence and the time series information between the word vectors, and determining a first feature matrix according to the output vector corresponding to each word vector, wherein the output vector of any word vector is used to generate an output vector corresponding to the next word vector of the word vector; based on the second feature extraction module, determining a second feature matrix according to the word vector sequence; based on the judgment module, determining the review result of the work ticket to be reviewed according to the first feature matrix and the second feature matrix. Specifically, the output vector corresponding to each word vector is generated according to each word vector in the word vector sequence and the timing information between the word vectors, wherein the output vector of any word vector is used to generate the output vector corresponding to the next word vector of the word vector, and the output vector of each word vector not only depends on the corresponding word vector, but also contains some features of the past word vectors, so the output vector covers more text features of the work ticket to be reviewed. In this way, the content features of the text content of the work ticket to be reviewed can be accurately extracted, thereby improving the accuracy of the review results of the work ticket. This method can also solve the problem of slow manual review and improve the speed of work ticket review.

[0071] Embodiment 2

[0072] Figure 4 This is a flowchart of a work ticket review method provided in Embodiment 2 of the present invention. This embodiment, based on the above embodiments, further defines the method of generating output vectors corresponding to each word vector, such as Figure 4 As shown, including:

[0073] Step 210: input the text content of the work ticket to be reviewed into a pre-trained work ticket review model, wherein the work ticket review model is generated based on historical work ticket sample training, and the work ticket review model includes a word segmentation module, a first feature extraction module, a second feature extraction module and a judgment module.

[0074] Step 220: Based on the word segmentation module, determine the word vector sequence according to the text content of the work ticket to be reviewed.

[0075] Step 230: Obtain a preset intermediate vector, a preset output vector, and the first word vector in the word vector sequence.

[0076] Step 240: The preset intermediate vector, the preset output vector and the first word vector form an input set.

[0077] Among them, for the generation of the output vector corresponding to any word vector, three vectors are required, including: an intermediate vector representing the temporal information between word vectors, an output vector corresponding to the previous word vector, and the current word vector. Therefore, for the generation of the first output vector, since there are no intermediate vectors and output vectors to refer to, a preset is required. For example, a unit vector or other standard vectors can be used as a preset intermediate vector and a preset output vector.

[0078] Step 250: perform vector operations on each vector in the input set, and output an output set corresponding to the input set, wherein the output set includes output vectors and intermediate vectors, wherein the intermediate vectors represent the temporal information between word vectors.

[0079] Specifically, the preset intermediate vector, the preset output vector and the first word vector may be combined into an input set, and then vector operations are performed on each vector in the input set to generate an output set corresponding to the first word vector, wherein the output set includes the output vector and the intermediate vector required to generate the output vector of the next word vector.

[0080] Optionally, performing a vector operation on each vector in the input set and outputting an output set corresponding to the input set includes:

[0081] Determine a first vector according to the vector sum of the word vectors in the input set and the output vector;

[0082] Determine a second vector based on a linear projection of the first vector on the word vectors in the input set;

[0083] Determine a third vector according to the vector product of the second vector and the first vector;

[0084] Determine a fourth vector according to a preset activation function and the third vector;

[0085] Determining a fifth vector according to a vector product of the fourth vector and an intermediate vector in the input set;

[0086] Determine an intermediate vector in the output set according to the vector product of the fifth vector and the fourth vector;

[0087] An output vector in the output set is determined according to a vector product of the fourth vector and an intermediate vector in the output set.

[0088] Optionally, the first vector L1, the second vector L2, the third vector L3, the fourth vector L4, and the fifth vector L5 can be calculated by the following formula to further calculate the output vector yt and the intermediate vector z t ;

[0089] L1=add(x t ,y t-1 )

[0090] L2=LNP(L1,x t )

[0091] L3=Hadamard(L1,L2)

[0092] L4=Mish(L3)

[0093] L5=Hadamard(Mish(L3),z t-1 )

[0094] z t =Hadamard(L4,L5)

[0095] y t =Hadamard(Relu(z t ),L4);

[0096] Where add represents matrix addition, LNP(L1,x t ) indicates that L1 is at x t Linear projection on . Hadamard represents vector element multiplication, Relu represents Relu function calculation, and Mish represents Mish activation function. t-1 ,y t-1 They are respectively the intermediate vector and output vector in the output set of the previous word vector.

[0097] For example, Figure 5 A schematic diagram of generating an output vector according to a word vector provided in an embodiment of the present invention, wherein z is an intermediate vector, y t is the word vector x t The corresponding output vector, WCM represents the vector operation module, (z0, y0, x1) constitutes the input set of the first word vector, (z1, y1) constitutes the output set of the first word vector, (y1, y2…y m ) constitutes the first characteristic matrix. Further, Figure 6 Schematic diagram of a WCM provided by an embodiment of the present invention, wherein arrows represent operation directions, circles represent operation nodes, l1, l2, l3, l4 and l5 represent the first vector, the second vector, the third vector, the fourth vector and the fifth vector, respectively, (x t ,z t-1 ,y t-1 ) represents the word vector x t The input set of zt-1 ,y t-1 are the intermediate vector and output vector in the output set of the previous word vector, respectively, t ,y t is the intermediate vector and output vector of the current word vector, z t ,y t Enough to form the word vector x t The output collection of .

[0098] Step 260: If the output times do not meet the preset termination condition, the input set is updated according to the next word vector in the word vector sequence and the output set, and vector operations are performed on each vector in the input set to output the output set corresponding to the input set.

[0099] Furthermore, the number of outputs is the number of times the output set is generated, and each output set corresponds to a word vector. Therefore, the preset end condition represents the quantity condition of traversing each word vector in the word vector sequence, that is, the number of outputs is used to judge whether the traversal of each word vector in the word vector sequence is completed. If the preset end condition is not met, the traversal continues until the output set corresponding to each word vector is output.

[0100] Specifically, the updating process can be understood as replacing the word vector in the input set with the current word vector, and replacing the intermediate vector and output vector in the output set of the previous word vector with the intermediate vector and output vector in the input set of the current word vector.

[0101] Step 270: If the number of outputs meets a preset termination condition, the output vectors in each output set are combined into a first feature matrix.

[0102] For example, the first feature matrix can be formed based on the Mish function and the word vectors.

[0103] Step 280: Based on the second feature extraction module, determine a second feature matrix according to the word vector sequence.

[0104] Step 290: Based on the judgment module, determine the audit result of the work ticket to be audited according to the first characteristic matrix and the second characteristic matrix.

[0105] For example, Figure 7 A schematic diagram of the overall process of work ticket review provided by an embodiment of the present invention, wherein the specific structure of WCM (Word embedding Chain module) is as follows Figure 5 As shown in the figure, Mish is the preset activation function, F1 is the concatenation matrix, Conv represents convolution kernels of different sizes, and SEFM is the convolution calculation module. The specific structure is as follows Figure 2As shown, Maxpooling is the maximum pooling module, Concat represents the concatenation operation, and FC is the fully connected layer. Specifically, after inputting the content text of the work ticket to be reviewed, the work ticket can be decomposed into different word vectors through the word segmentation algorithm, and then the word vector can be input into the WCM module to determine the output vector y corresponding to the word vector x. Further, each word vector set Mish activation function generates a first feature matrix representing the global features of the work ticket to be reviewed; at the same time, each word vector is spliced ​​into F1, and different convolution kernels are input to generate a convolution result, and then input into the SEFM module for another convolution operation, and the output sentence is subjected to the maximum pooling to generate the corresponding feature matrix. The feature matrix combined with the Mish activation function will generate a second feature matrix representing the local features of the word vector. Further, the first feature matrix and the second feature matrix are spliced, and the final audit result is determined through the fully connected layer.

[0106] An embodiment of the present invention discloses a work ticket review method. Specifically, an intermediate vector representing time series information, an output vector corresponding to a previous word vector, and a current word vector are formed into an input set, and vector operations are performed based on the vectors in the input set to generate an output vector corresponding to the current word vector. This method can retain the time series features between word vectors. At the same time, the later generated output vector is based on the previously generated output vector. This method can enable the generated output vector to retain the continuity features of the text content corresponding to the word vector. Through the above method, a first feature matrix containing more feature information can be generated, thereby improving the accuracy of word vector feature extraction, and thereby improving the accuracy of the review of the work ticket to be reviewed.

[0107] Embodiment 3

[0108] Figure 8 This is a schematic diagram of the structure of a work ticket review device provided in Embodiment 3 of the present invention. Figure 8 As shown, the device comprises:

[0109] An input module 310 is used to input the text content of the work ticket to be reviewed into a pre-trained work ticket review model, wherein the work ticket review model is generated based on historical work ticket sample training, and the work ticket review model includes a word segmentation module, a first feature extraction module, a second feature extraction module and a judgment module;

[0110] A sequence generation module 320, for determining a word vector sequence according to the text content of the work ticket to be reviewed based on the word segmentation module;

[0111] A first extraction module 330 is used to generate an output vector corresponding to each word vector according to each word vector in the word vector sequence and the time sequence information between the word vectors based on the first feature extraction module, and determine a first feature matrix according to the output vector corresponding to each word vector, wherein the output vector of any word vector is used to generate an output vector corresponding to the next word vector of the word vector;

[0112] A second extraction module 340, configured to determine a second feature matrix according to the word vector sequence based on the second feature extraction module;

[0113] The review module 350 is used to determine the review result of the work ticket to be reviewed based on the judgment module and the first characteristic matrix and the second characteristic matrix.

[0114] The device of this embodiment generates output vectors corresponding to each word vector according to each word vector in the word vector sequence and the timing information between the word vectors, wherein the output vector of any word vector is used to generate an output vector corresponding to the next word vector of the word vector. In this way, the content features of the text content of the work ticket to be reviewed can be accurately extracted, thereby improving the accuracy of the review results of the work ticket. This method can also solve the problem of slow manual review and improve the speed of work ticket review.

[0115] Optionally, the sequence generation module 320 includes:

[0116] The word segmentation unit is used to perform word segmentation on the text content of the work ticket to be reviewed and generate keywords for each text;

[0117] The vectorization unit is used to vectorize the text keywords, generate word vectors corresponding to the text keywords, and combine the word vectors into a word vector sequence.

[0118] Optionally, the first extraction module 330 includes:

[0119] An acquisition unit, used to acquire a preset intermediate vector, a preset output vector and the first word vector in the word vector sequence;

[0120] A set generating unit, configured to form an input set from the preset intermediate vector, the preset output vector and the first word vector;

[0121] An operation unit, configured to perform vector operation on each vector in the input set, and output an output set corresponding to the input set, wherein the output set includes an output vector and an intermediate vector, wherein the intermediate vector represents the temporal information between word vectors;

[0122] A judgment unit is used to update the input set according to the next word vector in the word vector sequence and the output set if the output number does not meet the preset end condition, and perform vector operations on each vector in the input set to output an output set corresponding to the input set; if the output number meets the preset end condition, the output vectors in each output set are combined into a first feature matrix.

[0123] The operation unit is specifically used to: determine the first vector according to the vector sum of the word vectors in the input set and the output vector;

[0124] Determine a second vector based on a linear projection of the first vector on the word vectors in the input set;

[0125] Determine a third vector according to the vector product of the second vector and the first vector;

[0126] Determine a fourth vector according to a preset activation function and the third vector;

[0127] Determining a fifth vector according to a vector product of the fourth vector and an intermediate vector in the input set;

[0128] Determine an intermediate vector in the output set according to the vector product of the fifth vector and the fourth vector;

[0129] An output vector in the output set is determined according to a vector product of the fourth vector and an intermediate vector in the output set.

[0130] Optionally, the second extraction module 340 includes:

[0131] A concatenation unit, used for concatenating the word vectors in the word vector sequence to generate a concatenation matrix;

[0132] A computing unit, used for performing convolution operations on the splicing matrix and each preset convolution kernel to determine each convolution result;

[0133] The activation unit is used to generate a second feature matrix according to a preset activation function and the convolution results.

[0134] Optionally, the judgment module in the audit module 350 includes a fully connected layer, specifically including: a splicing unit, used to splice the first feature matrix and the second feature matrix to generate a discriminant matrix; an audit unit, used to input the discriminant matrix into the fully connected layer, operate the discriminant matrix according to the weights and biases of the fully connected layer, and obtain the audit result of the work ticket to be audited based on the operation result.

[0135] Optionally, the device of the embodiment of the present invention further includes: a training module for training a work ticket review model, and the training module specifically includes:

[0136] An input unit, used to input historical work ticket samples into the work ticket review model to be trained, wherein the historical work ticket samples include the text content of the historical work ticket and the actual review results of the historical work ticket;

[0137] A word segmentation unit, used to determine a word vector sequence corresponding to the text content of the historical work ticket using a word segmentation module;

[0138] A first extraction unit, configured to use a first feature matrix extraction module to determine an output vector corresponding to each word vector in the word vector sequence, and determine a first feature matrix according to the output vector corresponding to each word vector;

[0139] A second extraction unit is used to use a second feature matrix extraction module to determine the word vector sequence to determine a second feature matrix;

[0140] A prediction unit, used for inputting the first characteristic matrix and the second characteristic matrix into a judgment module to determine a prediction review result;

[0141] A loss value determination unit, used to determine the loss value according to the predicted audit result, the actual audit result and the loss function of the work ticket audit model;

[0142] A loop unit is used to update the model parameters of the work ticket review model according to the loss value if the loss value does not meet the preset accuracy requirement.

[0143] The work ticket review device provided in the embodiment of the present invention can execute the work ticket review method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0144] Embodiment 4

[0145] Fig. 9 A schematic diagram of the structure of an electronic device 10 that can be used to implement an embodiment of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required herein.

[0146] like Fig. 9As shown, the electronic device 10 includes at least one processor 11, and a memory connected to the at least one processor 11, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc., wherein the memory stores a computer program that can be executed by at least one processor, and the processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 to the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the electronic device 10 can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other through a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0147] A number of components in the electronic device 10 are connected to the I / O interface 15, including: an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a disk, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0148] The processor 11 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 11 executes the various methods and processes described above, such as a work ticket review method.

[0149] In some embodiments, the work ticket audit method can be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed on the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the work ticket audit method described above can be performed. Alternatively, in other embodiments, the processor 11 can be configured to execute the work ticket audit method by any other appropriate means (for example, by means of firmware).

[0150] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0151] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when the computer program is executed by the processor, the functions / operations specified in the flow chart and / or block diagram are implemented. The computer program may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0152] In the context of the present invention, a computer-readable storage medium may be a tangible medium that may contain or store a computer program for use by or in combination with an instruction execution system, device or equipment. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, 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 disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0153] To provide interaction with a user, the systems and techniques described herein may be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form (including acoustic input, voice input, or tactile input).

[0154] The systems and techniques described herein may be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0155] A computing system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The client and server relationship is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services.

[0156] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of the present invention can be achieved, and this document does not limit this.

[0157] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A work ticket review method, characterized in that: include: Input the text content of the work ticket to be reviewed into a pre-trained work ticket review model, wherein the work ticket review model is generated based on historical work ticket sample training, and the work ticket review model includes a word segmentation module, a first feature extraction module, a second feature extraction module and a judgment module; Based on the word segmentation module, determine the word vector sequence according to the text content of the work ticket to be reviewed; Based on the first feature extraction module, an output vector corresponding to each word vector is generated according to each word vector in the word vector sequence and the time sequence information between the word vectors, and a first feature matrix is ​​determined according to the output vector corresponding to each word vector, wherein the output vector of any word vector is used to generate an output vector corresponding to the next word vector of the word vector; Based on the second feature extraction module, determine a second feature matrix according to the word vector sequence; Based on the judgment module, the audit result of the work ticket to be audited is determined according to the first characteristic matrix and the second characteristic matrix.

2. The method according to claim 1, characterized in that Determining the word vector sequence according to the text content of the work ticket to be reviewed includes: Perform word segmentation on the text content of the work ticket to be reviewed and generate keywords for each text; The text keywords are vectorized to generate word vectors corresponding to the text keywords, and the word vectors are combined into a word vector sequence.

3. The method according to claim 1, characterized in that: The step of generating output vectors corresponding to each word vector according to each word vector in the word vector sequence and the time sequence information between the word vectors, and determining a first feature matrix according to the output vectors corresponding to each word vector, includes: Get the preset intermediate vector, the preset output vector and the first word vector in the word vector sequence; The preset intermediate vector, the preset output vector and the first word vector form an input set; Performing vector operations on each vector in the input set, and outputting an output set corresponding to the input set, wherein the output set includes output vectors and intermediate vectors, wherein the intermediate vectors represent the temporal information between word vectors; If the output times do not meet the preset end condition, the input set is updated according to the next word vector in the word vector sequence and the output set, and vector operations are performed on each vector in the input set to output an output set corresponding to the input set; If the output times meet the preset termination condition, the output vectors in each output set are combined into a first feature matrix.

4. The method according to claim 3, characterized in that The performing vector operations on each vector in the input set and outputting an output set corresponding to the input set includes: Determine a first vector according to the vector sum of the word vectors in the input set and the output vector; Determine a second vector based on a linear projection of the first vector on the word vectors in the input set; Determine a third vector according to the vector product of the second vector and the first vector; Determine a fourth vector according to a preset activation function and the third vector; Determining a fifth vector according to a vector product of the fourth vector and an intermediate vector in the input set; Determine an intermediate vector in the output set according to the vector product of the fifth vector and the fourth vector; An output vector in the output set is determined according to a vector product of the fourth vector and an intermediate vector in the output set.

5. The method according to claim 1, characterized in that: The determining a second feature matrix according to the word vector sequence includes: splicing the word vectors in the word vector sequence to generate a splicing matrix; Performing convolution operations on the splicing matrix and each preset convolution kernel to determine each convolution result; A second feature matrix is ​​generated according to a preset activation function and the convolution results.

6. The method according to claim 1, characterized in that The judgment module includes a fully connected layer, and the determination of the audit result of the to-be-audited work ticket according to the first feature matrix and the second feature matrix includes: Concatenate the first feature matrix and the second feature matrix to generate a discriminant matrix; The discriminant matrix is ​​input into the fully connected layer, the discriminant matrix is ​​operated according to the weight and bias of the fully connected layer, and the audit result of the work ticket to be audited is obtained based on the operation result.

7. The method according to any one of claims 1 to 6, characterized in that: The training process of the work ticket review model includes: Inputting historical work ticket samples into the work ticket review model to be trained, wherein the historical work ticket samples include the text content of the historical work ticket and the actual review results of the historical work ticket; Using a word segmentation module, determine the word vector sequence corresponding to the text content of the historical work ticket; Using a first feature matrix extraction module to determine an output vector corresponding to each word vector in the word vector sequence, and determining a first feature matrix according to the output vector corresponding to each word vector; Using a second feature matrix extraction module to determine the word vector sequence to determine a second feature matrix; Inputting the first characteristic matrix and the second characteristic matrix into a judgment module to determine a prediction and review result; Determine the loss value according to the predicted audit result, the actual audit result and the loss function of the work ticket audit model; If the loss value does not meet the preset accuracy requirement, the model parameters of the work ticket review model are updated according to the loss value.

8. A work ticket review device, characterized in that: include: An input module, used for inputting the text content of the work ticket to be reviewed into a pre-trained work ticket review model, wherein the work ticket review model is generated based on historical work ticket sample training, and the work ticket review model includes a word segmentation module, a first feature extraction module, a second feature extraction module and a judgment module; A sequence generation module, used to determine a word vector sequence based on the word segmentation module and the text content of the work ticket to be reviewed; A first extraction module, configured to generate an output vector corresponding to each word vector according to each word vector in the word vector sequence and the time sequence information between the word vectors based on the first feature extraction module, and determine a first feature matrix according to the output vector corresponding to each word vector, wherein the output vector of any word vector is used to generate an output vector corresponding to a subsequent word vector of the word vector; A second extraction module, configured to determine a second feature matrix according to the word vector sequence based on the second feature extraction module; An audit module is used to determine the audit result of the work ticket to be audited based on the judgment module and the first characteristic matrix and the second characteristic matrix.

9. An electronic device, characterized in that: The electronic device comprises: at least one processor; and, a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the work ticket review method described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the work ticket review method described in any one of claims 1 to 7 when executed.

Citation Information

Cited By

  • Substation text auditing method, system and equipment based on large model and medium

    CN119761375A

  • A substation text auditing method, system, device and medium based on a large model

    CN119761375B