Repair order intelligent identification and classification method, electronic device, and storage medium
By using pre-defined regular expressions and large language models in a collaborative process, the issues of convenience and accuracy in multi-level classification of equipment faults are resolved, achieving efficient and accurate automatic classification.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- ZHEJIANG ZHIPU XINPIAN TECHNOLOGY CO LTD
- Filing Date
- 2026-03-12
- Publication Date
- 2026-06-19
AI Technical Summary
In existing technologies, multi-level classification methods for equipment faults require cumbersome manual operation by users, and the automatic classification of models consumes a lot of computational resources and has low accuracy.
The system uses a preset regular expression to match the N-level classification results of the repair request text. If the match fails, it obtains the classification results from level 1 to level N-1 through the first language model and generates prompt words. The system then uses the second language model to determine the N-level classification result based on the known results of the previous levels.
It achieves classification efficiency and accuracy without requiring manual user operation, reduces computational resource consumption, and ensures fast and accurate multi-level classification results.
Smart Images
Figure CN122240837A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of artificial intelligence technology, specifically providing a method for intelligent identification and classification of repair orders, an electronic device, and a storage medium. Background Technology
[0002] When equipment malfunctions, users can report the problem using the repair system. The system generates a repair order based on the user's input. After receiving the repair order, it needs to be categorized into multiple levels based on the repair details. These multiple levels of categorization include at least classifying the equipment type, fault location, and fault details in that order.
[0003] Currently, methods for multi-level classification of equipment faults mainly include two approaches: manual classification by users and automatic classification by models. Manual classification requires users to manually confirm the results of each level of classification when inputting the fault report into the system. This method requires multiple manual operations and is cumbersome. Automatic classification uses an artificial intelligence model to automatically classify the fault report. For example, if the fault report is "air conditioner not cooling," the model can automatically classify it as follows: equipment category: air conditioner; fault location: air conditioner compressor; fault content: not cooling. However, if there are many candidate categories for fault classification (e.g., 20 equipment categories, 400 fault locations, 800 fault causes), it not only increases the model's consumption of computing resources (or computational cost) but also affects the model's classification accuracy.
[0004] Accordingly, a new technical solution is needed in this field to solve the above problems. Summary of the Invention
[0005] This application aims to solve the above-mentioned technical problems, namely, to solve or at least partially solve the following technical problems: how to conveniently and accurately classify the repair content in the repair order into multiple levels.
[0006] In a first aspect, this application provides an intelligent identification and classification method for repair orders, comprising:
[0007] Retrieve the repair request text from the repair order;
[0008] The repair request text is matched against the N-level classification results based on a preset regular expression.
[0009] If a match is successful, obtain the N-level classification result obtained from the match, where N > 1;
[0010] If the match fails, the N-level classification result of the repair request text is obtained in the following way:
[0011] The repair request text is classified using a first-largest language model to obtain classification results from level 1 to level N-1. Based on the classification results from level 1 to level N-1, prompt words are generated. Based on the prompt words and using a second-largest language model, the repair request text is classified to obtain a level N classification result. The prompt words are used to indicate that, given the classification results from level 1 to level N-1, one of the classification options for level N is selected as the level N classification result.
[0012] In one technical solution of the above-mentioned intelligent identification and classification method for repair orders, the repair text is used to describe the fault that occurred in the equipment, and the N-level classification result is a three-level classification result. The first-level classification result represents the equipment category, the second-level classification result represents the fault location, and the third-level classification result represents the fault content.
[0013] In one technical solution of the above-mentioned intelligent identification and classification method for repair orders, the step of generating prompt words based on the classification results of levels 1 to N-1 includes:
[0014] From the classification options of the Nth level, select the classification options that are associated with the classification results of the (N-1)th level as candidate classification options;
[0015] Based on the candidate classification options and the classification results of levels 1 to N-1, a prompt word is generated. The prompt word is used to indicate that, given the classification results of levels 1 to N-1, one of the candidate classification options should be selected as the classification result of level N.
[0016] In one technical solution of the above-mentioned intelligent identification and classification method for repair orders, the (N-1)th level classification result represents the fault location, the Nth level classification result represents the fault content, and the classification option associated with the (N-1)th level classification result is the fault that occurred at the fault location.
[0017] In one technical solution of the above-mentioned intelligent identification and classification method for repair orders, before matching the N-level classification results of the repair text based on a preset regular expression, the method further includes:
[0018] The repair request text is preprocessed, and the final repair request text is obtained based on the preprocessing result. The preprocessing includes entity recognition.
[0019] In one technical solution of the above-mentioned intelligent identification and classification method for repair orders, the repair text is used to describe the fault that occurred in the equipment, and the entity identification includes identifying the equipment name and fault description information.
[0020] In one technical solution of the above-mentioned intelligent recognition and classification method for repair orders, the repair text in the repair order is the text obtained by speech recognition of the repair voice input by the user.
[0021] In one technical solution of the above-mentioned intelligent identification and classification method for repair orders, the number of classification options for the first to Nth levels increases sequentially, the first large language model adopts a lightweight large language model, and the second large language model adopts an enhanced large language model.
[0022] In a second aspect, an electronic device is provided, comprising at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores a computer program that, when executed by the at least one processor, implements the method described in any of the technical solutions provided in the first aspect.
[0023] In a third aspect, a computer-readable storage medium is provided, wherein a plurality of program codes are stored therein, the program codes being adapted to be loaded and executed by a processor to perform the method described in any of the technical solutions provided in the first aspect above.
[0024] The above-described technical solutions of this application have at least one or more of the following beneficial effects:
[0025] In one technical solution of the intelligent identification and classification method for repair orders provided in this application, the method can obtain the repair text in the repair order; match the N-level classification results of the repair text based on a preset regular expression; if the match is successful, obtain the N-level classification result obtained by the match, where N>1; if the match fails, obtain the N-level classification result of the repair text in the following way: classify the repair text using the first major language model to obtain the classification results from level 1 to level N-1, generate prompt words based on the classification results from level 1 to level N-1, classify the repair text based on the prompt words and using the second major language model to obtain the N-level classification result, and the prompt words are used to indicate that, under the condition that the classification results from level 1 to level N-1 are known, one of the classification options of the N-level classification is selected as the N-level classification result.
[0026] The above implementation scheme can automatically classify various repair request texts without requiring users to confirm the classification results at each level or manually input the classification results at each level, thus improving the convenience of classification operations.
[0027] In the above implementation scheme, matching the N-level classification results based on a preset regular expression can significantly improve classification efficiency. If the matching fails, the large language model is used again for classification to ensure that an N-level classification result can be obtained. In addition, when using the large language model for classification, the dual-model (first and second large language models) collaborative classification can reduce the consumption of computing resources by the model and also reduce the impact on the model's classification accuracy, so as to obtain an accurate N-level classification result as quickly as possible even if the matching fails. Attached Figure Description
[0028] The disclosure of this application will become more readily understood with reference to the accompanying drawings. It will be readily understood by those skilled in the art that these drawings are for illustrative purposes only and are not intended to limit the scope of protection of this application. Wherein:
[0029] Figure 1 This is a flowchart illustrating the main steps of a repair order intelligent identification and classification method according to an embodiment of this application. Figure 1 ;
[0030] Figure 2 This is a flowchart illustrating the main steps of a repair order intelligent identification and classification method according to an embodiment of this application. Figure 2 ;
[0031] Figure 3 This is a schematic diagram of the main structure of an electronic device according to an embodiment of this application. Detailed Implementation
[0032] Some embodiments of this application are described below with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of this application and are not intended to limit the scope of protection of this application.
[0033] First, an embodiment of the intelligent identification and classification method for repair orders provided in this application will be described.
[0034] See appendix Figure 1 , Figure 1 This is a schematic flowchart illustrating the main steps of a repair order intelligent identification and classification method according to an embodiment of this application. Figure 1 As shown, the intelligent identification and classification method for repair orders in this application embodiment mainly includes the following steps S101 to S105.
[0035] Step S101: Obtain the repair report text from the repair order.
[0036] A repair request document is used to describe the malfunction of the object being repaired. The object can be equipment, which may include, but is not limited to, household appliances, vehicles, etc. For example, the content of a repair request document for a vehicle could be, "My brake air compressor is constantly leaking air."
[0037] In some implementations, the repair request text can be text obtained by speech recognition of the user's input voice request. This implementation allows users to describe the repair request via voice, eliminating the need for manual input, simplifying the repair process and improving the user experience. Alternatively, conventional speech recognition methods can be used to recognize the repair request voice in this implementation, and this implementation is not specifically limited to this method.
[0038] Step S102: Match the N-level classification results of the repair report text based on the preset regular expression (N>1); if the match is successful, proceed to step S103; if the match fails, proceed to step S104. The number of preset regular expressions can be multiple.
[0039] The N-level classification result consists of the classification results from the 1st to the Nth level. Adjacent classification levels are hierarchical, and the later level classification is a sub-classification of the earlier level classification. In some implementations, the repair report text is used to describe the fault that occurred in the equipment. N=3, the first-level classification result indicates the equipment category, the second-level classification result indicates the fault location, and the third-level classification result indicates the fault content.
[0040] Matching the N-level classification results of repair request texts based on preset regular expressions can be understood as follows: The repair request text is sequentially matched with each level of the N-level classification system using regular expressions. If the repair request text matches successfully with each level of classification, then the repair request text matches the regular expression; otherwise, the match fails. For any level of classification, if the repair request text matches a classification option of that level, then the repair request text matches that level of classification successfully, and that matched classification option is used as the matching option for that level of classification. After the repair request text matches the regular expression successfully, the matching options from each level of classification can be combined to obtain the final N-level classification result. Specifically, when matching the repair request text with each level of classification, if the repair request text contains a classification option of the current level of classification, then the match is successful, and that contained classification option is used as the matching option for the current level of classification. If the repair request text does not contain a classification option of the current level of classification, but a matching option from another level of classification is associated with a classification option of the current level of classification, then the match is successful, and the associated classification option of the current level of classification is used as the matching option for the current level of classification.
[0041] For example, when N=3, the first, second, and third level classification results represent equipment category, fault location, and fault content, respectively. One regular expression is: In this regular expression, the first-level category options include "air conditioner" and "cooling," while the second-level category options include "not cooling," "no airflow," and "abnormal noise." If the repair request text is "air conditioner not cooling," the matching results for the first and third-level categories are "air conditioner" and "not cooling," respectively. "Not cooling" is associated with the second-level category option "air conditioner compressor," therefore, "air conditioner compressor" is selected as the second-level category matching option. Finally, the N-level category result for the repair request text is "Equipment Category: Air Conditioner, Fault Location: Air Conditioner Compressor, Fault Content: Not Cooling."
[0042] Matching N-level classification results using preset regular expressions can effectively improve the classification efficiency of repair request texts. For example, in some implementations, N-level classification results can be obtained within 10ms. Furthermore, in some implementations, when setting the preset regular expressions, frequently occurring repair request text content (i.e., high-frequency text) can be statistically analyzed. Regular expressions can then be set based on these high-frequency texts, so that when the repair request text in the repair order is high-frequency text, the N-level classification result of the repair request text can be quickly obtained through regular expression matching.
[0043] In some implementations, the repair request text can be preprocessed before executing step S102, and the final repair request text can be obtained based on the preprocessing result before executing step S102. That is, in step S102, the final repair request text is matched based on a preset regular expression. In addition, when the matching fails, when executing steps S104 to S105, the final repair request text is also classified using the first and second language models.
[0044] When obtaining the final repair report text based on the preprocessing results, the preprocessing results can be used as the final repair report text. Preprocessing may include entity recognition, and the preprocessing result consists of multiple entities identified through entity recognition. These entities are combined to form the final repair report text. In some embodiments, the repair report text describes a fault in the equipment, and entity recognition may include identifying the equipment name and fault description information in the repair report text. For example, if the repair report text is "Air conditioner is not cooling," the identified equipment name is "Air conditioner," and the fault description information is "not cooling." Another example is "My brake air compressor is constantly leaking air," where the identified equipment name is "Brake air compressor," and the fault description information is "leaking air."
[0045] Step S103: Obtain the N-level classification result obtained from the matching.
[0046] Step S104: Use the Large Language Model (LLM) to classify the repair request text to obtain classification results from level 1 to level N-1.
[0047] The first large language model can be obtained by fine-tuning a pre-trained large language model using a dataset from the repair report classification domain. This dataset includes multiple repair report texts and their annotation information, which can include the classification results of the repair report texts from level 1 to level N-1. For example, the dataset can include more than 3,000 repair report texts and their annotation information.
[0048] For any level of classification from level 1 to level N-1, the repair request text is input into the first language model for processing. The model can obtain the probability (i.e., classification probability) of each classification option in the current level classification, and the classification option with the highest classification probability is taken as the classification result of the current level classification. Taking N=3 as an example, the classification probability can be obtained by the following formula (1):
[0049] (1)
[0050] The meanings of the parameters in formula (1) are as follows:
[0051] Indicates the first The various category options for each level of classification, This indicates the repair request text. Indicates the first The classification probability of each category option in the hierarchical classification. express function, Represents the largest language model The Middle The weights corresponding to the level classification, Represents the largest language model The Middle The bias terms corresponding to the level classification, and All of these are model parameters.
[0052] When fine-tuning the first language model, the objective function shown in equation (2) can be used:
[0053] (2)
[0054] The meanings of the parameters in formula (2) are as follows:
[0055] This represents the Cross-Entropy Loss Function. Represents a dataset, This represents the repair request text in the dataset. Indicates repair request text The annotation information, Indicates that the repair request text will be submitted. The classification results from level 1 to level N-1 are obtained by inputting the data into the first large language model. This represents the model parameters of the largest language model. Indicate to make Minimize the model parameters as the objective Optimize, This represents the optimized model parameters. The model parameters are returned after training is complete. As the final model parameters.
[0056] Step S105: Generate prompt words based on the classification results of levels 1 to N-1. Classify the repair request text using the prompt words and the second largest language model (LLM) to obtain the Nth level classification result. The prompt words indicate which classification option to choose as the Nth level classification result given the known classification results of levels 1 to N-1. Finally, combine the classification results of levels 1 to N-1 obtained in step S104 with this Nth level classification result to obtain the Nth level classification result for the repair request text.
[0057] The second language model processes the prompt words and repair request text to obtain the probability of each classification option in the Nth level category, and takes the classification option with the highest probability as the classification result of the Nth level category. Taking N=3 as an example, the classification probability can be obtained by the following formula (3):
[0058] (3)
[0059] The meanings of the parameters in formula (3) are as follows: This represents the category options for the third-level category. This indicates the repair request text. and These represent the classification results for levels 1 and 2, respectively. This represents the classification probability of each option in the third-level category. Indicates a prompt word, This indicates a splicing operation. This represents the second largest language model.
[0060] In the methods described in steps S101 to S105 above, matching the N-level classification results based on a preset regular expression can significantly improve classification efficiency. Regular expressions cannot cover all classification cases (i.e., have low generalization ability), so there are cases of matching failure. However, when a match fails, the method can continue to use a large language model for classification, ensuring that an N-level classification result is obtained. This compensates for the low generalization ability of regular expression matching. Furthermore, using a dual-model collaborative classification approach when using a large language model reduces the model's consumption of computational resources and minimizes the impact on the model's classification accuracy, thus enabling the rapid acquisition of accurate N-level classification results even in the event of a matching failure.
[0061] The following describes an embodiment of the intelligent identification and classification method for repair orders provided in this application, specifically the method for generating prompt words in step S105.
[0062] In some embodiments according to this application, prompt words can be generated through the following steps S1051 to S1052.
[0063] Step S1051: From the classification options of the Nth level classification, select the classification option that is associated with the classification result of the (N-1)th level as the candidate classification option.
[0064] Step S1052: Generate prompt words based on the candidate classification options and the classification results of levels 1 to N-1. The prompt words are used to indicate that, given the classification results of levels 1 to N-1, one of the candidate classification options should be selected as the classification result of level N.
[0065] In some implementations, the (N-1)th level classification result represents the fault location, the Nth level classification result represents the fault content, and the classification option associated with the (N-1)th level classification result is the fault occurring at the fault location. For example, N=3, the first, second, and third level classification results represent the equipment category, fault location, and fault content, respectively, and the repair text is "My brake air compressor is constantly leaking air." Inputting this repair text into the first language model yields a first-level classification result of "brake" and a second-level classification result of "air compressor." The third-level classification options for faults occurring at the air compressor location include loose cylinder head mounting bolts, air filter failure, air leak, water leak, and oil leak. Based on this, the generated prompt could be: "Given that the first-level classification result is 'brake' and the second-level classification result is 'air compressor,' please select the most suitable option from the following third-level classification options: loose cylinder head mounting bolts, air filter failure, air leak, water leak, oil leak." The second language model uses this prompt to obtain a third-level classification result of "air leak."
[0066] In some implementations, the same fault may occur at different fault locations; for example, both tires and air compressors may leak air. In this implementation, a separate fault set can be set for each fault location, containing all possible faults at that location. When obtaining candidate classification options, the fault set representing the fault location according to the (N-1)th level classification result can be obtained, and the faults within that set can be used as candidate classification options. By setting a separate fault set for each fault location, the occurrence of the same fault at different locations can be isolated, avoiding confusion and further improving classification accuracy.
[0067] Based on the methods described in steps S1051 to S1052 above, the number of classification options for the Nth level category that the second language model needs to process can be reduced, thus compressing the number of classification options. This helps to improve the processing speed (or inference speed) of the second language model, thereby quickly obtaining the third-level classification result and improving the classification efficiency of repair request texts. For example, in some implementations, the second language model may only need to process 3% of the classification options in the Nth level category, which compresses 97% of the classification options.
[0068] The following describes an embodiment of the intelligent identification and classification method for repair orders.
[0069] In some embodiments according to this application, the number of classification options for the first to Nth levels increases sequentially. For example, N=3. The first, second, and third level classification results represent the equipment category, fault location, and fault content, respectively. The first level classification has more than 20 classification options, the second level classification has more than 400 classification options, and the third level classification has more than 800 classification options.
[0070] In this embodiment, a lightweight large language model can be used as the first large language model. Lightweight large language models have lower computational requirements, which helps improve the processing speed of the first large language model, thereby quickly obtaining the first and second-level classification results and improving the classification efficiency of repair request texts. Furthermore, compared to the number of classification options in the Nth level classification, the number of classification options in the first to Nth levels is relatively small. Therefore, even with a lightweight large language model, the classification accuracy of the first large language model will not be affected due to the smaller number of classification options to be processed.
[0071] For the Nth level classification, which has a large number of options, an enhanced large language model can be used as the second large language model to ensure its classification accuracy at the Nth level. The enhanced large language model has a larger parameter size than the lightweight large language model.
[0072] The embodiments of this application do not specifically limit the types of lightweight large language models and enhanced large language models. For example, in some implementations, the lightweight large language model may adopt the GLM-9B-0414 model, and the enhanced large language model may adopt the GLM-32B-0414 model.
[0073] The following is in conjunction with the appendix Figure 2 This paper describes the main steps and processes of the intelligent identification and classification method for repair orders provided in this application in some application scenarios. For example... Figure 2 As shown, the intelligent identification and classification method for repair orders may include the following steps S201 to S209.
[0074] Step S201: The user inputs a voice message requesting repair. Step S202: The voice message is processed using speech recognition to obtain the repair text. The voice message can describe a device malfunction; therefore, the repair text describes the malfunction. Step S203: The repair text is matched against the three-level classification results based on a preset regular expression. The first-level classification result indicates the device category, the second-level classification result indicates the fault location, and the third-level classification result indicates the fault content. Step S204: Determine if the match is successful; if successful, proceed to step S205 and output the three-level classification result based on the matching result; if unsuccessful, proceed to step S206.
[0075] Step S206: Classify the repair report text using the first major language model to obtain first and second-level classification results. Step S207: Generate prompt words, using the same method as described in steps S1051 to S1052 of the aforementioned embodiments. Step S208: Classify the repair report text using the prompt words and the second major language model to obtain third-level classification results. Step S209: Output the third-level classification result based on the first, second, and third-level classification results. Specifically, the first, second, and third-level classification results are combined to form the third-level classification result. For example, if the first major language model classifies the repair report text as "vehicle" and the fault location as "exterior compartment - window," the candidate classification options for the third-level classification when generating prompt words include "broken," "detached," and "not lit," and the second major language model classifies the repair report text based on the prompt words to obtain the fault content as "broken."
[0076] It should be noted that although the steps in the above embodiments are described in a specific order, those skilled in the art will understand that in order to achieve the effect of this application, different steps do not necessarily have to be executed in such an order. They can be executed simultaneously (in parallel) or in other orders. These adjusted solutions are equivalent to the technical solutions described in this application and therefore will also fall within the protection scope of this application.
[0077] Those skilled in the art will understand that all or part of the processes in the method of the above-described embodiment can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable file, or some intermediate form. The computer-readable storage medium can include any entity or device capable of carrying the computer program code, a medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory, a random access memory, an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.
[0078] Another aspect of this application provides a computer-readable storage medium.
[0079] In one embodiment of a computer-readable storage medium according to this application, the computer-readable storage medium can be configured to store a program that performs the intelligent identification and classification method for repair orders according to the above-described method embodiments. This program can be loaded and run by a processor to implement the above-described method. For ease of explanation, only the parts related to the embodiments of this application are shown; for specific technical details not disclosed, please refer to the method section of the embodiments of this application. The computer-readable storage medium can be a storage device comprising various electronic devices. Optionally, in the embodiments of this application, the computer-readable storage medium is a non-transitory computer-readable storage medium.
[0080] Another aspect of this application provides an electronic device.
[0081] In one embodiment of an electronic device according to this application, the electronic device may include at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores a computer program that, when executed by the at least one processor, implements the methods described in any of the above embodiments. See Appendix Figure 3 , Figure 3 The example illustrates a memory and processor connected via a bus communication connection.
[0082] The electronic devices described in this application may be, but are not limited to, mobile phones, tablets, desktops, laptops, handheld computers, notebook computers, in-vehicle devices, ultra-mobile personal computers (UMPCs), netbooks, personal digital assistants (PDAs), etc., and the embodiments of this application do not limit them.
[0083] In the description of this application, "processor" can include hardware, software, or a combination of both. A processor can be a central processing unit, microprocessor, graphics processor, digital signal processor, or any other suitable processor. A processor has data and / or signal processing capabilities. A processor can be implemented in software, in hardware, or a combination of both. Computer-readable storage media includes any suitable medium capable of storing program code, such as magnetic disks, hard disks, optical disks, flash memory, read-only memory, random access memory, etc. The term "A and / or B" means all possible combinations of A and B, such as only A, only B, or A and B.
[0084] The technical solutions of this application have been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of this application is obviously not limited to these specific embodiments. Without departing from the principles of this application, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of this application.
Claims
1. A method for intelligent identification and classification of repair orders, characterized in that, The method includes: Retrieve the repair request text from the repair order; The repair request text is matched against the N-level classification results based on a preset regular expression; If a match is successful, obtain the N-level classification result obtained from the match, where N > 1; If the match fails, the N-level classification result of the repair request text is obtained in the following way: The repair request text is classified using a first-largest language model to obtain classification results from level 1 to level N-1. Based on the classification results from level 1 to level N-1, prompt words are generated. Based on the prompt words and using a second-largest language model, the repair request text is classified to obtain a level N classification result. The prompt words are used to indicate that, given the classification results from level 1 to level N-1, one of the classification options for level N is selected as the level N classification result.
2. The method according to claim 1, characterized in that, The repair report text is used to describe the fault that occurred in the equipment. The N-level classification result is a three-level classification result. The first-level classification result indicates the equipment category, the second-level classification result indicates the fault location, and the third-level classification result indicates the fault content.
3. The method according to claim 1 or 2, characterized in that, The generation of prompt words based on the classification results of levels 1 to N-1 includes: From the classification options of the Nth level, select the classification options that are associated with the classification results of the (N-1)th level as candidate classification options; Based on the candidate classification options and the classification results of levels 1 to N-1, a prompt word is generated. The prompt word is used to indicate that, given the classification results of levels 1 to N-1, one of the candidate classification options should be selected as the classification result of level N.
4. The method according to claim 3, characterized in that, The (N-1)th level classification result represents the fault location, the Nth level classification result represents the fault content, and the classification option associated with the (N-1)th level classification result is the fault that occurred at the fault location.
5. The method according to claim 1, characterized in that, Before matching the N-level classification results of the repair request text based on the preset regular expression, the method further includes: The repair request text is preprocessed, and the final repair request text is obtained based on the preprocessing result. The preprocessing includes entity recognition.
6. The method according to claim 5, characterized in that, The repair report text is used to describe the fault that occurred in the equipment, and the entity identification includes identifying the equipment name and fault description information.
7. The method according to claim 1, characterized in that, The repair request text in the repair order is the text obtained by speech recognition of the user's input repair request voice.
8. The method according to claim 1, characterized in that, The number of classification options increases sequentially from level 1 to level N. The first large language model adopts a lightweight large language model, while the second large language model adopts an enhanced large language model.
9. An electronic device, characterized in that, include: At least one processor; And, a memory communicatively connected to the at least one processor; The memory stores a computer program, which, when executed by the at least one processor, implements the intelligent identification and classification method for repair orders as described in any one of claims 1 to 8.
10. A computer-readable storage medium storing a plurality of program codes, characterized in that, The program code is adapted to be loaded and run by a processor to perform the intelligent identification and classification method for repair orders as described in any one of claims 1 to 8.