Power marketing work order reply generation method and system based on generative large model

Abstract

The application relates to a power marketing work order reply generation method and system based on a generative large model, which comprises the following steps: acquiring a power marketing work order under any business scenario, searching and matching a pre-constructed work order reply template database to obtain recalled work order reply templates; inputting the recalled work order reply templates, the power marketing work order and a second prompt word into a large language model to generate a standardized work order reply; wherein the construction steps of the work order reply template database comprise the following steps: acquiring power marketing work order reply data under multiple business scenarios, performing high-frequency template extraction to obtain candidate work order reply templates under each business scenario; splicing the candidate work order reply templates, pre-defined work order reply templates and a first prompt word, inputting the spliced templates into a large language model to generate standardized work order reply templates, and constructing a work order reply template database. Compared with the prior art, the application has the advantages of improving the flexibility and efficiency of work order reply automation.

Description

Technical Field

[0001] This invention relates to the field of electricity marketing work order processing technology, and in particular to a method and system for generating electricity marketing work order responses based on a generative large model. Background Technology

[0002] In the fields of power marketing and customer service, work order processing systems face complex challenges across multiple business scenarios. Traditional work order processing methods primarily rely on manual experience to create response templates and rule-based matching methods, which struggle to handle diverse business needs such as repair request acceptance, customer service response, work order scheduling, and equipment inspection. This manually-driven model is not only inefficient but also often results in inconsistent work order response quality due to the complexity of business scenarios and the professionalism of industry terminology, making it difficult to guarantee service consistency and accuracy. With the rapid growth of business volume, the limitations of traditional methods in terms of parsing efficiency, generalization ability, and adaptability are becoming increasingly apparent.

[0003] In recent years, the rapid development of generative artificial intelligence technology has brought about a revolution in power marketing work order processing. Generative AI based on large language models can understand the unique business scenarios and terminology of the power industry, significantly improving the accuracy and adaptability of work order responses through semantic analysis and contextual modeling. This technology can dynamically generate professional and standardized response content based on specific business scenarios, effectively solving the pain point that traditional rule-based matching methods struggle to handle complex business needs. The introduction of generative AI has endowed work order processing systems with stronger semantic understanding and contextual reasoning capabilities, significantly reducing the need for manual intervention and providing key technical support for the intelligent upgrade of power marketing services. For example, patent application CN119599683A discloses an intelligent work order management system and method based on prompt word optimization.

[0004] However, current generative AI-based power marketing work order processing suffers from poor automation flexibility, slow generation speed, and poor text quality. Therefore, it is necessary to improve existing processing technologies to solve these problems. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for generating power marketing work order responses based on a generative large model, which improves the automation level and processing efficiency of work order responses.

[0006] The objective of this invention can be achieved through the following technical solutions:

[0007] A method for generating electricity marketing work order responses based on a generative large model, comprising:

[0008] Obtain electricity marketing work orders from any business scenario, search and match them with a pre-built work order reply template database, and obtain the recalled work order reply templates;

[0009] The recalled work order reply template, the power marketing work order, and the second prompt word are concatenated and input into the large language model to generate a standardized work order reply;

[0010] The steps for constructing the work order response template database include:

[0011] Acquire power marketing work order response data from multiple business scenarios, extract high-frequency templates, and obtain candidate work order response templates for each business scenario;

[0012] The candidate work order response template, the predefined work order response template, and the first prompt word are concatenated, input into a large language model, a standardized work order response template is generated, and a work order response template database is constructed.

[0013] Furthermore, the business scenarios include repair request acceptance, customer service response, work order scheduling, and equipment inspection.

[0014] Furthermore, the step of obtaining the recall work order response template includes:

[0015] Based on the aforementioned electricity marketing work order, multiple keywords are extracted using a pre-trained model to form a keyword set;

[0016] Based on the keyword set and the work order reply template database, the BM25 algorithm is used for template matching to obtain multiple nearest neighbor work order reply templates, which are then used as the recalled work order reply templates.

[0017] Furthermore, the step of forming the keyword set includes:

[0018] For the aforementioned electricity marketing work order Word segmentation and part-of-speech tagging are performed to identify nouns and obtain a candidate word set. ;

[0019] The segmented electricity marketing work order is input into the pre-trained model to obtain the vector representation of each token in the sentence, expressed as:

[0020] ,

[0021] In the formula, A vector representation of the token. It is the number of tokens. It is the vector dimension. for d 3D column vector, For the real number field;

[0022] Based on the vector representation of each token, the average of the rows is used to obtain the global vector representation of the sentence, expressed as:

[0023] ,

[0024] In the formula, This is the global vector representation of the sentence. This is an averaging operation;

[0025] For the candidate word set Candidate words in The vector average of its occurrences across all positions in the sentence is taken as the candidate word. The vector representation of is expressed as:

[0026] ,

[0027] In the formula, Candidate words The vector representation of , Candidate words Vector representations of the first and second tokens in the sentence;

[0028] Based on the global vector representation and candidate words of the sentence Vector representation of candidate words The relevance to the sentence, wherein the expression for calculating the relevance is:

[0029] ,

[0030] In the formula, Correlation;

[0031] The relevance was sorted from high to low, and the top-k candidate words were selected as the keyword set. .

[0032] Furthermore, the step of obtaining multiple nearest neighbor work order reply templates includes:

[0033] Define the work order reply template database as Based on the keyword set, the BM25 weight of the templates in the power marketing work order and work order reply template database is calculated, wherein the calculation expression for the BM25 weight is:

[0034] ,

[0035] in:

[0036] ,

[0037] In the formula, BM25 weight, This is a work order reply template. Length, For a set of keywords, For terms Inverse document frequency, It is a term In the work order reply template In the frequency range, avgdl represents the average length of templates in the work order response template database. and It's about adjusting parameters. The number of templates in the work order reply template database. This represents the number of templates containing the term;

[0038] Sort the BM25 weights from high to low, and select the Top-K nearest neighbor work order response templates. The expression is:

[0039] ,

[0040] In the formula, This is the template for replying to the nearest neighbor work order.

[0041] Furthermore, the standardized work order response is expressed as follows:

[0042] ,

[0043] In the formula, For standardized work order responses, For large language models, This is a template for responding to recall work orders. For electricity marketing work orders, The second cue word. This indicates a splicing operation.

[0044] Furthermore, the high-frequency template is extracted using the LCS algorithm, and the specific steps include:

[0045] 1) Set the power marketing work order response data for the aforementioned multi-business scenario as follows: And perform preprocessing, in which Reply data for electricity marketing work orders in the k-th business scenario;

[0046] 2) For the preprocessed electricity marketing work order response data under each business scenario, calculate the longest common subsequence (LCS) between each pair of texts. Simultaneously, use dynamic programming to record sub-problems to obtain the final LCS. The expression for calculating the LCS is:

[0047] ,

[0048] In the formula, In the k-th business scenario, the first... i A text reply to an electricity marketing work order. For the k-th business scenario, the first j A text reply to an electricity marketing work order. Representing a subsequence Length;

[0049] The operational expression for the dynamic programming method is:

[0050] ,

[0051] In the formula, Indicates the first i Reply text for a power marketing work order The first m-length substring and the first m-th substring j Reply text for a power marketing work order The longest common subsequence of the first n length substrings, express The m-th character, express The nth character;

[0052] 3) Iterate through step 2) to obtain the set of common subsequences for each business scenario, represented as:

[0053] ,

[0054] In the formula, This is the set of common subsequences under the k-th business scenario;

[0055] 4) Set threshold The low-frequency subsequences in the common subsequence set for each business scenario are filtered out to obtain the high-frequency common subsequence set for each business scenario. This high-frequency template is then used as the candidate work order response template. The high-frequency common subsequence set for each business scenario is represented as follows:

[0056] ,

[0057] formula, This is the set of high-frequency common subsequences under the k-th business scenario. for The common subsequence in for exist The number of times it appears in for The number.

[0058] Furthermore, the preprocessing steps include:

[0059] The power marketing work order response data under the aforementioned multi-business scenarios is subjected to text cleaning, data anonymization, and tag replacement processing. Specifically, the text cleaning process removes special characters and garbled text from the text, the data anonymization process marks sensitive information in the text as specific placeholders, and the tag replacement process uses regular expressions to identify phrases containing time, location, and numbers in the text and replaces the phrases with specific placeholders.

[0060] Furthermore, the standardized work order response template is represented as follows:

[0061] ,

[0062] In the formula, This is a standard chemical bill reply template for the k-th business scenario. For large language models, This is a candidate work order response template for the k-th business scenario. This is a predefined work order response template for the k-th business scenario. This is the first prompt word in the k-th business scenario. This indicates a splicing operation.

[0063] This invention also provides a power marketing work order response generation system based on a generative large model, comprising:

[0064] Search and Matching Module: Used to obtain power marketing work orders in any business scenario, and perform search and matching with a pre-built work order reply template database to obtain the recalled work order reply templates;

[0065] Work order response module: used to concatenate the recalled work order response template, the power marketing work order, and the second prompt word, input them into the large language model, and generate a standardized work order response;

[0066] The retrieval and matching module includes:

[0067] High-frequency template extraction unit: used to acquire power marketing work order response data under multiple business scenarios, perform high-frequency template extraction, and obtain candidate work order response templates for each business scenario;

[0068] Database construction unit: used to concatenate the candidate work order reply template, the predefined work order reply template and the first prompt word, input them into the large language model, generate standardized work order reply templates, and build a work order reply template database.

[0069] Compared with the prior art, the present invention has the following beneficial effects:

[0070] (1) This invention integrates knowledge from multiple business scenarios and guides the optimization and generation mechanism of reply templates through prompt words and large language models, which significantly improves the automation level and processing efficiency of power marketing work order replies.

[0071] (2) This invention uses the BM25 algorithm for template matching. By independently calculating the BM25 weight, it is not only applicable to templates containing multiple keywords, but also allows adjustment of the text length to influence the weight by adjusting the parameters. This avoids the high similarity of long texts due to the large number of words. It is also friendly to long text processing and has the characteristics of strong interpretability and sensitivity to keywords, thereby improving the accuracy and efficiency of template matching.

[0072] (3) The BM25 algorithm of this invention can efficiently process sparse data. It can highlight the importance of keywords through inverse document frequency (IDF). Even if the keywords appear less frequently in the text, as long as the IDF is high enough, a high matching score can still be obtained, thereby further improving the accuracy of matching.

[0073] (4) This invention uses the LCS algorithm for high-frequency template extraction. When the LCS algorithm solves the longest common subsequence, a dynamic programming method is introduced to handle the case where the longest common subsequence problem has overlapping subproblems, which avoids repeated calculations, reduces time complexity, and thus improves computational efficiency. Attached Figure Description

[0074] Figure 1 This is a schematic diagram of the overall process of the method of the present invention;

[0075] Figure 2 This is a schematic diagram of the work order reply template generation process of the present invention;

[0076] Figure 3 This is a schematic diagram of the work order response generation process of the present invention. Detailed Implementation

[0077] Example 1

[0078] This example provides a method for generating electricity marketing work order responses based on a generative large model, such as... Figure 1 As shown, it includes the following steps:

[0079] S1: Collect power marketing work order response data from multiple business scenarios, extract high-frequency templates, and obtain candidate work order response templates.

[0080] This step is the template construction stage, which employs high-frequency template extraction technology and utilizes the LCS algorithm to generate candidate work order response templates from multi-service work order data. Specifically:

[0081] In this embodiment, the power marketing work order response data for multiple business scenarios includes response data for business scenarios such as repair acceptance, customer service response, work order scheduling, and equipment inspection.

[0082] Combination Figure 2 As shown, the dataset for power marketing work order responses across multiple business scenarios is defined as follows: Each of them Let this represent the set of work order responses for the k-th business scenario. For each business scenario, the following processing is performed:

[0083] 1. Text cleaning: Remove special characters and garbled text.

[0084] 2. Data anonymization: Sensitive information such as customer information and employee information is marked with specific placeholders.

[0085] 3. Tag Replacement: Use regular expressions to identify phrases such as time, location, and numbers, and replace them with specific placeholders.

[0086] The set of work order responses for each type of business scenario (in terms of...) (For example) Perform the following steps respectively: Calculate the longest common subsequence (LCS) between each pair of texts:

[0087] ,

[0088] in, Representing a subsequence Length, This represents the response text for a specific electricity marketing work order. Since the longest common subsequence problem has overlapping subproblems, dynamic programming can be used.

[0089] ,

[0090] in, express The m-th character, express The nth character, express The first m-length substring and The longest common subsequence of the first n lengths of the text. By iteratively calculating the longest common subsequence between each pair of texts, a set of common subsequences is obtained:

[0091] ,

[0092] Set threshold Filter low-frequency segments and retain high-frequency LCS components:

[0093] ,

[0094] in, It is the first The number, This is the set of high-frequency common subsequences under the k-th business scenario. for The common subsequence in for exist The number of times it appears in the text. From Extract from the pool to obtain candidate work order response templates. .

[0095] S2: Concatenate the candidate work order response template, the predefined work order response template, and the first prompt word, input them into the large language model, obtain the standardized work order response template, and construct the work order response template database.

[0096] This step is the template optimization stage. Candidate work order response templates, predefined work order response templates, and manually written first prompt words are concatenated and input into the Guangming Big Language Model to generate standardized work order response templates and build a work order response template database. Specifically:

[0097] Reply template for the extracted candidate work orders With the predefined work order reply template And manually written first prompt words Input the Guangming Big Language Model to generate standardized templates :

[0098] ,

[0099] in, This indicates a splicing operation. This represents the Guangming Big Language Model.

[0100] Receive standardized work order reply template Then, it is stored in the database DB to obtain the work order response template database.

[0101] S3: For the input power marketing work order, extract keywords, search the work order reply template database, and obtain the recalled work order reply template.

[0102] This step is the retrieval and matching stage. Part-of-speech tagging is performed on the input electricity marketing work orders to identify noun phrases as candidate words. These candidate words are then processed through the RoBERTa-wwm model, with scores calculated using attention to obtain a keyword set. A sparse search is then performed on this keyword set using the BM25 algorithm to retrieve the most relevant work order response templates. Specifically:

[0103] combine Figure 3 As shown, the electricity marketing work order in the newly input business scenario is defined as follows: The words are then input into a pre-trained model (such as RoBERTa-wwm), and the word segmentation results are part-of-speech tagged to identify nouns and obtain a candidate word set. This yields the vector representation of the token in the sentence:

[0104] ,

[0105] in It is the number of tokens. It is the vector dimension.

[0106] right The average of the rows yields the global vector representation of the sentence:

[0107] ,

[0108] For candidate words (As after word segmentation), the average of the vectors of all occurrences of a word in the sentence is taken as its vector representation:

[0109] ,

[0110] Calculate the relevance of candidate word c to the sentence:

[0111] ,

[0112] according to Sort the keywords from highest to lowest quality and select the top-k candidate words as the keyword set. .

[0113] Define the work order reply template database as Calculate the BM25 weight of each work order response template and the power marketing work order:

[0114] ,

[0115] in It is a term In the template The frequency in It's a template. The length of avgdl is the average length of the text in the database. and It's about adjusting parameters. This is a set of keywords. The IDF formula is as follows:

[0116] ,

[0117] in, This represents the total number of texts (templates) in the work order response template database. Represents included terms The number of templates.

[0118] Based on the calculated BM25 weights, sort them from high to low, and return the Top-K nearest neighbor work order response templates as the recall work order response templates:

[0119] ,

[0120] In the formula, This is a template for responding to a recall work order.

[0121] S4: Combine the recalled work order response template, the input power marketing work order, and the second prompt word, input them into the large language model, and obtain the standardized work order response.

[0122] This step is the response generation stage. It involves concatenating the recalled work order response template, the input electricity marketing work order, and manually written second prompt words, then inputting this into the Guangming Big Language Model. The Model then uses a Decoder-Only architecture Transformer model to generate a standardized work order response through autoregression. Specifically:

[0123] Recall work order reply template New electricity marketing work orders Inputting the manually written second prompt word into the Guangming Big Language Model generates a standardized work order response:

[0124] ,

[0125] In the formula, For standardized work order responses, This is a manually written second prompt word.

[0126] The above method integrates knowledge from multiple business scenarios such as repair request acceptance, customer service response, and work order scheduling. Through the optimization and generation mechanism of prompt-guided templates, it significantly improves the automation level and processing efficiency of power marketing work order responses.

[0127] Example 2

[0128] This example provides a power marketing work order response generation system based on a generative large model, including:

[0129] Search and Matching Module: Used to obtain power marketing work orders in any business scenario, and perform search and matching with a pre-built work order reply template database to obtain the recalled work order reply templates;

[0130] Work order response module: used to input the recalled work order response template, power marketing work order and second prompt words into the big language model to generate standardized work order responses;

[0131] The retrieval and matching module includes:

[0132] High-frequency template extraction unit: used to acquire power marketing work order response data under multiple business scenarios, perform high-frequency template extraction, and obtain candidate work order response templates for each business scenario;

[0133] Database construction unit: used to concatenate the candidate work order reply template, the predefined work order reply template and the first prompt word, input them into the large language model, generate standardized work order reply templates, and build a work order reply template database.

[0134] The rest are as in Example 1.

[0135] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a 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 invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0136] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of the present invention can be implemented using various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0137] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0138] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0139] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0140] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0141] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for generating electricity marketing work order responses based on a generative large model, characterized in that, include: Obtain electricity marketing work orders from any business scenario, search and match them with a pre-built work order reply template database, and obtain the recalled work order reply templates; The recalled work order response template, the power marketing work order, and the second prompt word are concatenated and input into the large language model to generate a standardized work order response; The steps for constructing the work order response template database include: Acquire power marketing work order response data from multiple business scenarios, extract high-frequency templates, and obtain candidate work order response templates for each business scenario; The candidate work order response template, the predefined work order response template, and the first prompt word are concatenated, input into a large language model, a standardized work order response template is generated, and a work order response template database is constructed. The steps for obtaining the recall work order response template include: Based on the aforementioned electricity marketing work order, multiple keywords are extracted using a pre-trained model to form a keyword set; Based on the keyword set and the work order reply template database, the BM25 algorithm is used for template matching to obtain multiple nearest neighbor work order reply templates, which are then used as the recalled work order reply templates. The steps for forming the keyword set include: For the aforementioned electricity marketing work order Word segmentation and part-of-speech tagging are performed to identify nouns and obtain a candidate word set. ; The segmented electricity marketing work order is input into the pre-trained model to obtain the vector representation of each token in the sentence. The expression is as follows: ， In the formula, A vector representation of the token. It is the number of tokens. It is the vector dimension. for d 3D column vector, For the real number field; Based on the vector representation of each token, the average of the rows is used to obtain the global vector representation of the sentence, expressed as: ， In the formula, This is the global vector representation of the sentence. This is an averaging operation; For the candidate word set Candidate words in The vector average of its occurrences across all positions in the sentence is taken as the candidate word. The vector representation of is expressed as: ， In the formula, Candidate words The vector representation of , Candidate words Vector representations of the first and second tokens in the sentence; Based on the global vector representation and candidate words of the sentence Vector representation of candidate words The relevance to the sentence, wherein the expression for calculating the relevance is: ， In the formula, Correlation; The relevance was sorted from high to low, and the top-k candidate words were selected as the keyword set. .

2. The method for generating electricity marketing work order responses based on a generative large model according to claim 1, characterized in that, The business scenarios include repair request acceptance, customer service response, work order scheduling, and equipment inspection.

3. The method for generating electricity marketing work order responses based on a generative large model according to claim 1, characterized in that, The steps for obtaining multiple nearest neighbor work order reply templates include: Define the work order reply template database as Based on the keyword set, the BM25 weight of the templates in the power marketing work order and work order reply template database is calculated, wherein the calculation expression for the BM25 weight is: ， in: ， In the formula, BM25 weight, This is a work order reply template. Length, For a set of keywords, For terms Inverse document frequency, It is a term In the work order reply template In the frequency range, avgdl represents the average length of templates in the work order response template database. and It's about adjusting parameters. The number of templates in the work order reply template database. This represents the number of templates containing the term; Sort the BM25 weights from high to low, and select the Top-K nearest neighbor work order response templates. The expression is: ， In the formula, This is the template for replying to the nearest neighbor work order.

4. The method for generating electricity marketing work order responses based on a generative large model according to claim 1, characterized in that, The standardized work order response is expressed as follows: ， In the formula, For standardized work order responses, For large language models, This is a template for responding to recall work orders. For electricity marketing work orders, The second cue word. This indicates a splicing operation.

5. The method for generating electricity marketing work order responses based on a generative large model according to claim 1, characterized in that, The high-frequency template is extracted using the LCS algorithm, and the specific steps include: 1) Set the power marketing work order response data for the aforementioned multi-business scenario as follows: And perform preprocessing, among which Reply data for electricity marketing work orders in the k-th business scenario; 2) For the preprocessed electricity marketing work order response data under each business scenario, calculate the longest common subsequence (LCS) between each pair of texts. Simultaneously, use dynamic programming to record sub-problems to obtain the final LCS. The expression for calculating the LCS is: ， In the formula, In the k-th business scenario, the first... i A text reply to an electricity marketing work order. For the k-th business scenario, the first j A text reply to an electricity marketing work order. Representing a subsequence Length; The operational expression for the dynamic programming method is: ， In the formula, Indicates the first i Reply text for a power marketing work order The first m-length substring and the first m-th substring j Reply text for a power marketing work order The longest common subsequence of the first n length substrings, express The m-th character, express The nth character; 3) Iterate through step 2) to obtain the set of common subsequences for each business scenario, represented as: ， In the formula, This is the set of common subsequences under the k-th business scenario; 4) Set threshold The low-frequency subsequences in the common subsequence set for each business scenario are filtered out to obtain the high-frequency common subsequence set for each business scenario. This high-frequency template is then used as the candidate work order response template. The high-frequency common subsequence set for each business scenario is represented as follows: ， formula, This is the set of high-frequency common subsequences under the k-th business scenario. for The common subsequence in for exist The number of times it appears in for The number.

6. The method for generating electricity marketing work order responses based on a generative large model according to claim 5, characterized in that, The preprocessing steps include: The power marketing work order response data under the aforementioned multi-business scenarios is subjected to text cleaning, data anonymization, and tag replacement processing. Specifically, the text cleaning process removes special characters and garbled text from the text, the data anonymization process marks sensitive information in the text as specific placeholders, and the tag replacement process uses regular expressions to identify phrases containing time, location, and numbers in the text and replaces the phrases with specific placeholders.

7. The method for generating electricity marketing work order responses based on a generative large model according to claim 1, characterized in that, The standardized work order response template is as follows: ， In the formula, This is a standard chemical bill reply template for the k-th business scenario. For large language models, This is a candidate work order response template for the k-th business scenario. This is a predefined work order response template for the k-th business scenario. This is the first prompt word in the k-th business scenario. This indicates a splicing operation.

8. A generation system for a power marketing work order response generation method based on a generative large model according to any one of claims 1-7, characterized in that, include: Search and Matching Module: Used to obtain power marketing work orders in any business scenario, and perform search and matching with a pre-built work order reply template database to obtain the recalled work order reply templates; Work order response module: used to concatenate the recalled work order response template, the power marketing work order, and the second prompt word, input them into the large language model, and generate a standardized work order response; The retrieval and matching module includes: High-frequency template extraction unit: used to acquire power marketing work order response data under multiple business scenarios, perform high-frequency template extraction, and obtain candidate work order response templates for each business scenario; Database construction unit: used to concatenate the candidate work order reply template, the predefined work order reply template and the first prompt word, input them into the large language model, generate standardized work order reply templates, and build a work order reply template database.

Images