Work order processing method and device, nonvolatile storage medium and electronic equipment

Abstract

The invention discloses a work order processing method and device, a nonvolatile storage medium and electronic equipment. The method comprises the steps that a first work order error reporting information set of a target work order is extracted, and the first work order error reporting information set comprises keywords used for representing error reporting information of the target work order; determining the similarity between the first work order error reporting information set and the plurality of historical work order error reporting information sets, and determining a second work order error reporting information set from the plurality of historical work order error reporting information sets according to the similarity; and determining a preset solution associated with the second work order error reporting information, and processing the target work order according to the preset solution. The technical problem that the work order processing efficiency is low due to the fact that the error reporting work order cannot be automatically processed in the related technology is solved.

Description

Technical Field

[0001] This application relates to the field of electronic digital data processing, and more specifically, to a work order processing method, apparatus, non-volatile storage medium, and electronic device. Background Technology

[0002] In related technologies, when processing work orders, error-reported work orders cannot be processed automatically and require manual intervention. This results in low work order processing efficiency.

[0003] There is currently no effective solution to the above problems. Summary of the Invention

[0004] This application provides a work order processing method, apparatus, non-volatile storage medium, and electronic device to at least solve the technical problem of low work order processing efficiency caused by the inability to automatically process erroneous work orders in related technologies.

[0005] According to one aspect of the embodiments of this application, a work order processing method is provided, comprising: extracting a first work order error information set of a target work order, wherein the first work order error information set includes keywords for representing error information of the target work order; determining the similarity between the first work order error information set and multiple historical work order error information sets, and determining a second work order error information set from the multiple historical work order error information sets based on the similarity; determining a preset solution associated with the second work order error information, and processing the target work order according to the preset solution.

[0006] Optionally, determining the similarity between the first work order error information set and each historical work order error information set includes: determining the Jaccard similarity and cosine similarity between the first work order error information set and each historical work order error information set; determining the second work order error information set from each historical work order error information set based on the Jaccard similarity and cosine similarity includes: determining the comprehensive similarity index between the first work order error information set and each historical work order error information set based on the Jaccard similarity and cosine similarity between the first work order error information set and each historical work order error information set; determining the historical work order error information set with the largest corresponding comprehensive similarity index and whose corresponding comprehensive similarity index is greater than a preset comprehensive similarity threshold as the second work order error information set.

[0007] Optionally, the preset comprehensive similarity threshold is determined by: determining the confidence index corresponding to the first work order error information set; determining the benchmark threshold; and determining the preset comprehensive similarity threshold based on the benchmark threshold and the confidence index.

[0008] Optionally, before extracting the first set of error information for the target work order, the method further includes: determining training data based on historical work order error data; training the information extraction model based on the training data, wherein the trained information extraction model is used to extract the first set of error information for the target work order, wherein the information extraction model includes a bidirectional converter encoder representation module, a bidirectional long short-term memory network module, and a conditional random field module.

[0009] Optionally, the output data of the bidirectional transformer encoder representation module is the input data of the bidirectional long short-term memory network module, and the output data of the bidirectional long short-term memory network module is the input data of the conditional random field module. The bidirectional transformer encoder representation module is used to determine the context-dependent representation information of the input sequence of the information extraction model; the bidirectional long short-term memory network module is used to determine the bidirectional dependency information of the input sequence based on the context-dependent representation information; and the conditional random field module is used to determine the label sequence corresponding to the input sequence based on the bidirectional dependency information. The label sequence is used to indicate the word type of each word in the input sequence and to indicate the keywords in the input sequence that represent error information.

[0010] Optionally, determining training data based on historical work order error data includes: performing word segmentation on the historical work order error data to obtain multiple keywords; determining the annotation information corresponding to the keywords, wherein the annotation information is used to indicate the type of the keywords; and using the keywords with added annotation information as training data.

[0011] Optionally, after segmenting the historical work order error data to obtain multiple keywords, the method further includes: performing semantic integrity verification on the multiple keywords and identifying fragmented keywords that do not have complete semantics; merging the fragmented keywords to obtain keywords with complete semantics.

[0012] According to another aspect of the embodiments of this application, a work order processing apparatus is also provided, comprising: a first processing module, configured to extract a first work order error information set of a target work order, wherein the first work order error information set includes keywords for representing error information of the target work order; a second processing module, configured to determine the similarity between the first work order error information set and multiple historical work order error information sets, and determine a second work order error information set from the multiple historical work order error information sets based on the similarity; and a third processing module, configured to determine a preset solution associated with the second work order error information, and process the target work order according to the preset solution.

[0013] According to another aspect of the embodiments of this application, a non-volatile storage medium is also provided, wherein a program is stored in the non-volatile storage medium, and the program controls the device where the non-volatile storage medium is located to execute a work order processing method when it runs.

[0014] According to another aspect of the embodiments of this application, an electronic device is also provided, including a memory and a processor, wherein the processor is used to run a program stored in the memory, wherein the program executes a work order processing method when it runs.

[0015] According to another aspect of the embodiments of this application, a computer program product is also provided, including a computer program that implements a work order processing method when executed by a processor.

[0016] In this embodiment, a first set of error information for the target work order is extracted, which includes keywords representing the error information of the target work order. The similarity between the first set of error information and multiple historical sets of error information is determined, and a second set of error information is determined from these historical sets based on the similarity. A preset solution associated with the second set of error information is determined, and the target work order is processed according to the preset solution. By determining the second set of error information that matches the first set based on similarity and executing the preset solution associated with the second set of error information to process the target work order, the purpose of automatically determining the processing method for error-prone work orders is achieved. This realizes the technical effect of automatically processing error-prone work orders, thereby solving the technical problem of low work order processing efficiency caused by the inability to automatically process error-prone work orders in related technologies. Attached Figure Description

[0017] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0018] Figure 1 This is a schematic diagram of the structure of a computer terminal (or mobile device) according to an embodiment of this application;

[0019] Figure 2 This is a flowchart illustrating a work order processing method according to an embodiment of this application;

[0020] Figure 3 This is a schematic diagram of the workflow of an information extraction model provided according to an embodiment of this application;

[0021] Figure 4 This is a schematic diagram of a work order processing device provided according to an embodiment of this application. Detailed Implementation

[0022] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0023] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0024] To better understand the embodiments of this application, the technical terms involved in the embodiments of this application are explained below:

[0025] Named Entity Recognition (NER): Entity extraction is a fundamental task in Natural Language Processing (NLP) used to identify and extract specific types of named entities (such as people's names, place names, organization names, time, quantity, etc.) from unstructured text and classify them into predefined categories.

[0026] Intelligent Automatic Business Ticket Processing (AIAP): AIAP refers to the fully intelligent process of automatically receiving, analyzing, processing, and responding to business tickets using artificial intelligence, automation technologies, and other means. The system can autonomously complete ticket processing with little or no human intervention.

[0027] In related technologies, when operator service work orders encounter errors such as order cancellation or lost orders, manual processing is often required, resulting in low work order processing efficiency.

[0028] To address the aforementioned issues, this application provides a solution that utilizes natural language processing and deep learning algorithms to preprocess and annotate the raw work order data. It automatically identifies and extracts entities from error messages that occur during service activation and intelligently matches solutions for abnormal work orders. This accurately and efficiently resolves work order bottlenecks, reduces the daily workload of frontline staff, significantly saves manual operation time, lowers the work order bottleneck rate, improves customer service quality, and ultimately reduces costs and increases efficiency for enterprises. Detailed explanation follows.

[0029] According to an embodiment of this application, a method embodiment for processing work orders is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0030] The methods and embodiments provided in this application can be executed on mobile terminals, computer terminals, or similar computing devices. Figure 1 A hardware block diagram of a computer terminal (or mobile device 10) for implementing a work order processing method is shown. Figure 1 As shown, the computer terminal 10 (or mobile device 10) may include one or more processors 102 (shown as 102a, 102b, ..., 102n in the figure) 102 (processor 102 may include, but is not limited to, a microprocessor MCU or a programmable logic device FPGA, etc.), a memory 104 for storing data, and a transmission device 106 for communication functions. In addition, it may also include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of a BUS bus), a network interface, a power supply, and / or a camera. Those skilled in the art will understand that... Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the aforementioned electronic device. For example, computer terminal 10 may also include... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown.

[0031] It should be noted that the aforementioned one or more processors 102 and / or other data processing circuits are generally referred to herein as "data processing circuits". These data processing circuits may be embodied, in whole or in part, in software, hardware, firmware, or any other combination thereof. Furthermore, the data processing circuits may be a single, independent processing module, or may be wholly or partially integrated into any other element within the computer terminal 10 (or mobile device 10). As involved in the embodiments of this application, the data processing circuit serves as a processor control mechanism (e.g., selection of a variable resistor termination path connected to an interface).

[0032] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the work order processing method in this embodiment. The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, thereby realizing the aforementioned work order processing method. The memory 104 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor 102, and these remote memories can be connected to the computer terminal 10 (or mobile device 10) via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0033] The transmission device 106 is used to receive or send data via a network. Specific examples of the network described above may include a wireless network provided by the communication provider of the computer terminal 10. In one example, the transmission device 106 includes a Network Interface Controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission device 106 may be a Radio Frequency (RF) module, used for wireless communication with the Internet.

[0034] The display may be, for example, a touchscreen liquid crystal display (LCD) that allows the user to interact with the user interface of the computer terminal 10 (or mobile device 10).

[0035] Under the above operating environment, this application provides a work order processing method, such as... Figure 2 As shown, the method includes the following steps:

[0036] Step S202: Extract the first set of error messages for the target work order, wherein the first set of error messages includes keywords used to represent the error messages of the target work order;

[0037] In some embodiments of this application, the keywords used to represent error information of the target work order are also called entities.

[0038] In some embodiments of this application, before extracting the first set of error information of the target work order, the method further includes: determining training data based on historical work order error data; training the information extraction model based on the training data, wherein the trained information extraction model is used to extract the first set of error information of the target work order, wherein the information extraction model includes a bidirectional converter encoder representation module (BERT), a bidirectional long short-term memory network module (BiLSTIM), and a conditional random field module (CRF).

[0039] In some embodiments of this application, the step of determining training data based on historical work order error data includes: performing word segmentation on the historical work order error data to obtain multiple keywords; determining the annotation information corresponding to the keywords, wherein the annotation information is used to indicate the type of the keywords; and using the keywords with added annotation information as training data.

[0040] In some embodiments of this application, the raw data of SIM cards collected by the operator can be divided into training set, validation set and test set according to a preset ratio (such as 6:2:2).

[0041] In some embodiments of this application, regular expressions can be used to clean the error information field in the original work order data to obtain preprocessed error information, which is historical work order error data. Preprocessing can filter out sensitive user privacy data, including user IP addresses, home addresses, etc., and remove abnormal data caused by data recording, format conversion, import and export, etc.

[0042] As an optional implementation, the preprocessed error message fields can be handed over to business experts to link the standard solutions corresponding to the error messages to form a standard corpus. The error messages in the corpus can then undergo entity extraction operations to form standardized work order error message entities. This results in a set of historical work order error messages and the solutions corresponding to each set. Each historical work order error message set includes historical work order error data from the same work order, and this data includes keywords that reflect the error information of that work order—that is, the error message fields.

[0043] In some embodiments of this application, a domain-adaptive annotator can be constructed based on existing word segmentation algorithms such as the jieba word segmentation algorithm. This annotator is then used to dynamically segment and initially annotate historical work order error data, thereby obtaining keywords that reflect the error information. This allows for basic entity boundary identification of the preprocessed error information, such as identifying idle IPs as independent words, reducing the complexity of manual annotation.

[0044] As an optional implementation method, after segmenting historical work order error data to obtain multiple keywords, the method further includes: performing semantic integrity verification on multiple keywords and identifying fragmented keywords that do not have complete semantics; merging fragmented keywords to obtain keywords with complete semantics.

[0045] In some embodiments of this application, the aforementioned fragmented keywords refer to keywords whose business information cannot be determined based on the keyword alone, such as "idle" and "IP". Subsequently, adjacent fragmented keywords are merged into keywords that can reflect complete business semantic information based on semantic information, such as modifying "idle" + "IP" into the complete entity "idle IP".

[0046] In some embodiments of this application, predefined domain-specific keywords can be added to historical work order error data to perform domain knowledge enhancement processing. For example, communication device keywords can be added, and "mse pre-check" can be marked as a professional operation entity.

[0047] In some embodiments of this application, for historical work order error data that has undergone semantic integrity verification and merging processing, as well as domain knowledge enhancement processing, an entity-attribute mapping table containing the contextual associations of the historical work order error data can also be generated. This table includes keyword attributes corresponding to each entity (i.e., each keyword) in the data. The entity attributes stored in the mapping table can be set according to the application scenario, and may include, for example, the entity's word type, status information, etc.

[0048] In some embodiments of this application, when generating training data for the information extraction model based on historical work order error data, a BIO annotation converter can also be used to annotate the historical work order error data. BIO annotation is a commonly used sequence labeling method, mainly applied in natural language processing tasks such as named entity recognition, relation extraction, and text classification. It can label each element in a sequence as B, I, or O, representing the starting word of an entity, other words, and non-entity words, respectively. Specifically, if X represents a noun phrase (NP), then the three BIO labels are: B-NP (the beginning of the noun phrase), I-NP (the rest of the noun phrase), and O (not a noun phrase). Therefore, a passage can be divided into different segments and labeled using BIO. The specific operation is as follows:

[0049] The first step is to determine the automated annotation conversion algorithm: For each character in the entity annotation set generated by expert review (that is, the historical work order error data generated by expert review), determine its entity type and generate the corresponding BIO tag.

[0050] The second step is to establish a conflict detection mechanism: sort overlapping entities by weight and prioritize retaining entities with high business value.

[0051] In some embodiments of this application, historical work order error data can be vector-embedded before training the information extraction model. During the vector embedding stage, the semantic understanding model can be initially trained using publicly available text data lacking specific domain features, and then incrementally trained using labeled historical work order error data. This optimizes the vector space distribution of technical terms such as "IPWG".

[0052] In some embodiments of this application, the entities in the historical work order error data can also be classified, as shown in the table below:

[0053]

[0054] As an optional implementation method, such as Figure 3As shown, the output data of the bidirectional transformer encoder module is the input data of the bidirectional long short-term memory network module, and the output data of the bidirectional long short-term memory network module is the input data of the conditional random field module. The bidirectional transformer encoder module is used to determine the context-dependent representation information of the input sequence of the information extraction model; the bidirectional long short-term memory network module is used to determine the bidirectional dependency information of the input sequence based on the context-dependent representation information; and the conditional random field module is used to determine the label sequence corresponding to the input sequence based on the bidirectional dependency information. The label sequence indicates the word type of each word in the input sequence and the keywords in the input sequence that represent error messages.

[0055] In some embodiments of this application, after cleaning, labeling, and embedding historical work order error data to obtain training data, the BERT-BiLSTM-CRF joint model (i.e., the aforementioned information extraction model) can be trained using the training data. In this model, the BERT layer generates high-quality initial representations through pre-trained knowledge, the BiLSTM performs task-adaptive feature tuning based on these representations, and the CRF ultimately completes label decision through global probability modeling. Specific training parameters can be set independently; for example, a batch size of 20, a learning rate of 1e-3, and 100 training epochs can be used, resulting in an F1 score higher than 95.00% on both the validation and test sets.

[0056] In some embodiments of this application, before using the above-described information extraction model to perform entity extraction processing and determine the entities in the work order (i.e., the keywords that reflect the error information), it is necessary to preprocess the data input into the information extraction model, including the following steps:

[0057] The first step is text normalization:

[0058] Unicode standardization is implemented, control characters are filtered, and full-width / half-width symbols are standardized. BERT tokenizer is used for word segmentation (WordPiece), and special [CLS] and [SEP] markers are added to meet the maximum sequence length constraint.

[0059]

[0060]

[0061] Where Text is the original input text, [CLS] and [SEP] are special markers in BERT, adding the [CLS] marker to the beginning and the [SEP] marker to the end of the segmented subword list, max_len is the maximum sequence length, and Tokens[:max_len-2] indicates that the segmented subword list is truncated, keeping only the first max_len-2 subwords;

[0062] The second step is vectorization:

[0063] Generate three core tensors:

[0064] Token IDs: Index sequences mapped via a vocabulary

[0065] Attention Mask: A 0 / 1 matrix that distinguishes between valid tokens and padding.

[0066] Segment IDs: Segment identifiers where a single sentence task is set to zero.

[0067] In some embodiments of this application, such as Figure 3 As shown, the forward propagation process of the information extraction model includes the following steps:

[0068] Step 1, BERT encoding:

[0069] The input embedding matrix is ​​processed through 12 Transformer layers to output a context-sensitive representation:

[0070]

[0071] in The output of the BERT encoder represents the context-dependent representation of the input sequence, with each row corresponding to an embedding vector for an input position. Given the word embedding matrix of the input sequence, each row corresponds to the embedding vector of a word or subword. is the segment embedding matrix used to distinguish different sentences (in sentence pair tasks). For example, in BERT, sentence A has a token of 0, and sentence B has a token of 1. N represents the length of the input sequence (i.e., the number of words or subwords). This represents the hidden layer dimension of the BERT model (typically 768). This represents the dimension of the output matrix, which has n sequence positions, and the embedding vector at each position has a dimension of... .

[0072] The second step is BiLSTM feature extraction:

[0073] Bidirectional LSTM captures bidirectional dependencies in sequences and outputs the concatenated hidden state:

[0074]

[0075]

[0076]

[0077] in This represents the hidden state of the forward LSTM at time step t. This represents the hidden state of the backward LSTM at time step t. This is the embedding vector of the BERT output matrix at time step t. It is an LSTM unit used to capture bidirectional dependencies of sequences. The output of the BiLSTM is formed by concatenating the hidden states of the forward and backward LSTMs. The hidden layer dimension of a unidirectional LSTM. This represents the dimension of the BiLSTM output matrix, which has n sequence positions, and the dimension of the feature vector at each position is... (Because the hidden states of the forward and backward directions are spliced ​​together).

[0078] Step 3: CRF decoding:

[0079] Define emission matrix (k is the number of label categories) and transition matrix The optimal tag sequence is decoded using the Viterbi algorithm. :

[0080]

[0081]

[0082] in This represents the score of the sequence label path y. Let be the emission matrix, representing the matching score between the feature vector and the label at each position, where k is the number of label categories. Let be the transition matrix, representing the score for transitioning from one label to another. These are the labels for time steps t and t-1. It is the optimal label sequence obtained by the Viterbi algorithm. This means selecting the sequence with the highest score among all possible label sequences y.

[0083] In some embodiments of this application, the output of the information extraction model can be further processed by word alignment and label sequence parsing. Word alignment includes merging words segmented by WordPiece, which can employ a first-word strategy. For example, the word "unwanted" might be split into ["un", "##want", "##ed"]. However, in sequence labeling tasks, these words need to be aligned back to the original word labels to generate the correct label sequence.

[0084] Tag sequence parsing includes generating entity boundaries based on BIO annotation rules and filtering illegal combinations (such as "I-NP" without "B-NP").

[0085] In some embodiments of this application, such as Figure 3 As shown, in the information extraction model, BERT is used as the encoder to encode the input sequence into a context-sensitive embedding representation, which is then fed into a BiLSTM recurrent neural network. Specifically, forward LSTM and backward LSTM are used to encode the input text from front to back and from back to front, respectively, capturing the bidirectional dependencies of the sequence. Finally, the decoding model CRF is used to calculate the optimal label sequence using the emission matrix and transition matrix. Furthermore, before using the information extraction model to determine the work order error information set, BERT can be fine-tuned based on event data and solution data from the work order scenario.

[0086] Step S204: Determine the similarity between the first set of work order error messages and multiple sets of historical work order error messages, and determine the second set of work order error messages from the multiple sets of historical work order error messages based on the similarity.

[0087] In the technical solution provided in step S204, the step of determining the similarity between the first work order error information set and each historical work order error information set includes: determining the Jaccard similarity and cosine similarity between the first work order error information set and each historical work order error information set; determining the second work order error information set from each historical work order error information set based on the Jaccard similarity and cosine similarity includes: determining the comprehensive similarity index between the first work order error information set and each historical work order error information set based on the Jaccard similarity and cosine similarity between the first work order error information set and each historical work order error information set; determining the historical work order error information set with the largest corresponding comprehensive similarity index and whose corresponding comprehensive similarity index is greater than a preset comprehensive similarity threshold as the second work order error information set.

[0088] In some embodiments of this application, the first set of work order error messages can be defined as set A, and the set of historical work order error messages can be defined as set B. Then, a comprehensive similarity index can be constructed by combining Jaccard similarity and vector space cosine similarity to compare the similarity between different sets.

[0089]

[0090]

[0091] in Each weight coefficient needs to be determined through validation set tuning. A higher value indicates a higher similarity, and vice versa. and This represents the vectorized feature representation of two sets.

[0092] In some embodiments of this application, the aforementioned comprehensive similarity index S can be used to calculate the maximum S value between the extracted new card order error information entity set and the standardized card order error information entity set, and the set with the largest S value can be identified as the second work order error information set. The standard solution linked to the second work order error information set in the corpus can then be output. Optionally, when determining the second work order error information set, a composite index judgment threshold can be set based on expert experience. Error information-solutions with a composite index S value higher than this threshold are classified as reliable and output, while those with a value lower than this threshold are classified as unreliable and not output.

[0093] In some embodiments of this application, when the total similarity index S between set A and each set B is found to be lower than the preset comprehensive similarity threshold, abnormal events such as card orders corresponding to the target work order can be recorded, and the collected content can be sent to business experts for completion on a regular basis. The completed entity information and solution information are stored in a standardized card order entity list so that when system problems occur in the future, the results can be output quickly and accurately, thereby improving the efficiency of problem solving.

[0094] As an optional implementation, the preset comprehensive similarity threshold is determined as follows: a confidence index corresponding to the first work order error information set is determined; a baseline threshold is determined; and the preset comprehensive similarity threshold is determined based on the baseline threshold and the confidence index. Additionally, the conditional random field module in the information extraction model outputs a probability value indicating that the actual output path is the optimal path during the decoding phase. This probability value can be normalized to a confidence weight ω, which is the confidence index mentioned above. Here, ω∈[0,1]. When ω is close to 1, it indicates that the model is very confident that the prediction result is optimal. When ω is close to 1, it indicates that the model is less certain whether the prediction result is optimal.

[0095] Optionally, a confidence-based dynamic thresholding mechanism is a method that adjusts the threshold by combining the path probability from the CRF decoding stage. This dynamic thresholding mechanism allows for dynamic adjustment of the decision threshold based on the model's confidence level, thereby improving the model's adaptability to different confidence levels. The formula for dynamically determining the preset comprehensive similarity threshold based on confidence level is as follows:

[0096]

[0097] in It is a baseline threshold, usually set by experts based on experience. This is the adjustment coefficient, used to control the degree of influence of confidence on the threshold; an initial value of 0.2 is recommended. ω is the normalized confidence weight.

[0098] As can be seen from the above formula, when ω>0.5, it indicates that the model is relatively confident in the prediction results, and the dynamic threshold is appropriate at this time. This will increase, thereby increasing the threshold strictness.

[0099] When ω < 0.5, it indicates that the model is not very certain about the prediction results, and in this case, the dynamic threshold is used. This will lower the threshold, thereby relaxing the stringency.

[0100] By introducing confidence weights, the dynamic threshold mechanism can automatically adjust the threshold based on the model's prediction confidence, thereby better adapting to decision-making needs at different confidence levels. (Adjustment coefficient) The dynamic threshold mechanism can be adjusted according to specific application scenarios, making it more flexible. Furthermore, dynamically adjusting the threshold can optimize decisions based on the model's confidence level, thereby improving the overall accuracy of the model.

[0101] Step S206: Determine the preset solution associated with the error information of the second work order, and process the target work order according to the preset solution.

[0102] By extracting a first set of error information for the target work order, which includes keywords representing the error information of the target work order; determining the similarity between the first set of error information and multiple historical sets of error information, and determining a second set of error information for the target work order from the multiple historical sets of error information based on the similarity; determining a preset solution associated with the second set of error information, and processing the target work order according to the preset solution, the method of automatically determining the processing method for error-reported work orders is achieved. This solves the technical problem of low work order processing efficiency caused by the inability to automatically process error-reported work orders in related technologies.

[0103] Furthermore, the work order processing method provided in this application embodiment addresses work order errors such as card-related errors that occur during the operator's business work order activation process. It offers a complete work order processing flow, from card-related data preprocessing, annotation, and vector embedding to entity extraction, automatic matching of card-related solutions, and automatic corpus updates. The BIO entity annotation method is used to automatically annotate the original card-related error information, laying the foundation for subsequent accurate entity extraction and automatic solution matching. A BERT-BiLSTM-CRF-based entity extraction model for card-related error information is further deployed to efficiently and accurately extract key entity content from the error information. In addition, an innovative intelligent matching algorithm for card-related event solutions is designed. By cleverly calculating the composite index of new card-related error information entities and standardized card-related error information entities, the algorithm automatically selects the solution with the highest matching degree, significantly improving the overall efficiency of work order operations.

[0104] This application provides a work order processing device. Figure 4 This is a schematic diagram of the device. From Figure 4 As can be seen from the diagram, the device includes: a first processing module 40, used to extract a first set of error information for the target work order, wherein the first set of error information includes keywords used to represent the error information of the target work order; a second processing module 42, used to determine the similarity between the first set of error information and multiple sets of historical error information, and to determine a second set of error information for the target work order from the multiple sets of historical error information based on the similarity; and a third processing module 44, used to determine a preset solution associated with the second set of error information, and to process the target work order according to the preset solution.

[0105] In some embodiments of this application, before extracting the first set of error information of the target work order, the first processing module 40 is further configured to: determine training data based on historical work order error data; train the information extraction model based on the training data, wherein the trained information extraction model is used to extract the first set of error information of the target work order, wherein the information extraction model includes a bidirectional converter encoder representation module, a bidirectional long short-term memory network module, and a conditional random field module.

[0106] In some embodiments of this application, the output data of the bidirectional transformer encoder representation module is the input data of the bidirectional long short-term memory network module, and the output data of the bidirectional long short-term memory network module is the input data of the conditional random field module. The bidirectional transformer encoder representation module is used to determine the context-dependent representation information of the input sequence of the information extraction model; the bidirectional long short-term memory network module is used to determine the bidirectional dependency information of the input sequence based on the context-dependent representation information; and the conditional random field module is used to determine the label sequence corresponding to the input sequence based on the bidirectional dependency information. The label sequence is used to indicate the word type of each word in the input sequence and to indicate the keywords in the input sequence that represent error information.

[0107] In some embodiments of this application, the first processing module 40 determines training data based on historical work order error data by performing word segmentation on the historical work order error data to obtain multiple keywords; determining the annotation information corresponding to the keywords, wherein the annotation information is used to indicate the type of the keywords; and using the keywords with added annotation information as training data.

[0108] In some embodiments of this application, after segmenting historical work order error data to obtain multiple keywords, the first processing module 40 is further used to: perform semantic integrity verification on the multiple keywords and determine fragmented keywords that do not have complete semantics; merge the fragmented keywords to obtain keywords with complete semantics.

[0109] In some embodiments of this application, the step of the second processing module 42 in determining the similarity between the first work order error information set and each historical work order error information set includes: determining the Jaccard similarity and cosine similarity between the first work order error information set and each historical work order error information set.

[0110] The second processing module 42 determines the second work order error information set from each historical work order error information set based on Jaccard similarity and cosine similarity, including: determining the comprehensive similarity index between the first work order error information set and each historical work order error information set based on the Jaccard similarity and cosine similarity between the first work order error information set and each historical work order error information set; and determining the historical work order error information set with the largest corresponding comprehensive similarity index and the corresponding comprehensive similarity index greater than the preset comprehensive similarity threshold as the second work order error information set.

[0111] In some embodiments of this application, the preset comprehensive similarity threshold is determined by: determining the confidence index corresponding to the first work order error information set; determining the benchmark threshold; and determining the preset comprehensive similarity threshold based on the benchmark threshold and the confidence index.

[0112] It should be noted that each module in the above-mentioned work order processing device can be a program module (for example, a set of program instructions to implement a certain function) or a hardware module. For the latter, it can be manifested in the following forms, but is not limited to them: each of the above modules is manifested as a processor, or the functions of each of the above modules are implemented by a processor.

[0113] According to an embodiment of this application, a non-volatile storage medium is also provided, which stores a program. When the program runs, it controls the device where the non-volatile storage medium is located to execute the following work order processing method: extracting a first set of work order error information for the target work order, wherein the first set of work order error information includes keywords used to represent error information of the target work order; determining the similarity between the first set of work order error information and multiple historical sets of work order error information, and determining a second set of work order error information from the multiple historical sets of work order error information based on the similarity; determining a preset solution associated with the second set of work order error information, and processing the target work order according to the preset solution.

[0114] According to an embodiment of this application, an electronic device is also provided, including a memory and a processor. The processor is used to run a program stored in the memory, wherein the program executes the following work order processing method: extracting a first work order error information set of the target work order, wherein the first work order error information set includes keywords used to represent error information of the target work order; determining the similarity between the first work order error information set and multiple historical work order error information sets, and determining a second work order error information set from the multiple historical work order error information sets based on the similarity; determining a preset solution associated with the second work order error information, and processing the target work order according to the preset solution.

[0115] According to an embodiment of this application, a computer program product is also provided, including a computer program that, when executed by a processor, implements the following work order processing method: extracting a first work order error information set of a target work order, wherein the first work order error information set includes keywords used to represent error information of the target work order; determining the similarity between the first work order error information set and multiple historical work order error information sets, and determining a second work order error information set from the multiple historical work order error information sets based on the similarity; determining a preset solution associated with the second work order error information, and processing the target work order according to the preset solution.

[0116] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0117] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0118] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0119] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0120] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to related technologies, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0121] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A work order processing method, characterized in that, include: Extract a first set of error messages for the target work order, wherein the first set of error messages includes keywords used to represent the error messages of the target work order; Determine the similarity between the first set of work order error information and multiple sets of historical work order error information, and determine the second set of work order error information from the multiple sets of historical work order error information based on the similarity. Determine a preset solution associated with the error message of the second work order, and process the target work order according to the preset solution.

2. The work order processing method according to claim 1, characterized in that, Determining the similarity between the first set of error messages for work orders and the sets of error messages for each historical work order includes: Determine the Jaccard similarity and cosine similarity between the first set of error messages for work orders and each set of error messages for historical work orders; The second work order error information set is determined from the various historical work order error information sets based on the Jaccard similarity and the cosine similarity, including: Based on the Jaccard similarity and cosine similarity between the first work order error information set and each historical work order error information set, a comprehensive similarity index between the first work order error information set and each historical work order error information set is determined. The set of historical work order error information that has the largest corresponding comprehensive similarity index and whose corresponding comprehensive similarity index is greater than a preset comprehensive similarity threshold is determined as the second work order error information set.

3. The work order processing method according to claim 2, characterized in that, The preset comprehensive similarity threshold is determined in the following way: Determine the confidence index corresponding to the first work order error information set; Determine the baseline threshold; The preset comprehensive similarity threshold is determined based on the benchmark threshold and the confidence index.

4. The work order processing method according to claim 1, characterized in that, Before extracting the first set of error messages for the target work order, the method further includes: Training data is determined based on historical work order error data; The information extraction model is trained based on the training data. After training, the information extraction model is used to extract the first work order error information set of the target work order. The information extraction model includes a bidirectional converter encoder representation module, a bidirectional long short-term memory network module, and a conditional random field module.

5. The work order processing method according to claim 4, characterized in that, The output data of the bidirectional converter encoder module is the input data of the bidirectional long short-term memory network module, and the output data of the bidirectional long short-term memory network module is the input data of the conditional random field module. The bidirectional converter encoder representation module is used to determine the context-related representation information of the input sequence of the information extraction model; The bidirectional long short-term memory network module is used to determine the bidirectional dependency information of the input sequence based on the context-related representation information; The conditional random field module is used to determine the label sequence corresponding to the input sequence based on the bidirectional dependency information, wherein the label sequence is used to indicate the word type of each word in the input sequence, and to indicate the keywords in the input sequence that reflect error information.

6. The work order processing method according to claim 4, characterized in that, The training data, determined based on historical work order error data, includes: The historical work order error data was segmented to obtain multiple keywords; Determine the annotation information corresponding to the keyword, wherein the annotation information is used to indicate the type of the keyword; The keywords with the added annotation information are used as the training data.

7. The work order processing method according to claim 6, characterized in that, After performing word segmentation on the historical work order error data to obtain multiple keywords, the method further includes: Perform semantic integrity verification on the multiple keywords and identify fragmented keywords that do not have complete semantics; By merging the fragmented keywords, keywords with complete semantic meaning are obtained.

8. A work order processing device, characterized in that, include: The first processing module is used to extract a first set of error information for the target work order, wherein the first set of error information includes keywords used to represent the error information of the target work order; The second processing module is used to determine the similarity between the first work order error information set and multiple historical work order error information sets, and to determine the second work order error information set from the multiple historical work order error information sets based on the similarity. The third processing module is used to determine a preset solution associated with the error information of the second work order, and to process the target work order according to the preset solution.

9. A non-volatile storage medium, characterized in that, The non-volatile storage medium stores a program, wherein when the program is executed, it controls the device where the non-volatile storage medium is located to execute the work order processing method according to any one of claims 1 to 7.

10. An electronic device, characterized in that, include: A memory and a processor, the processor being configured to run a program stored in the memory, wherein the program, when running, executes the work order processing method according to any one of claims 1 to 7.

11. A computer program product, characterized in that, It includes a computer program that, when executed by a processor, implements the work order processing method according to any one of claims 1 to 7.

Images