Power report generation method based on large model and multi-document course learning framework

Through the collaborative approach of large models and multi-document course learning framework, the problem that traditional power report generation methods are difficult to deal with multi-element composite problems is solved, efficient and accurate report generation is achieved, and the quality and efficiency of power reports are improved.

CN119088911AActive Publication Date: 2024-12-06CHINA SOUTHERN POWER GRID ARTIFICIAL INTELLIGENCE TECHNOLOGY CO LTD

Patent Information

Application Number
CN202411036274.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-31
Publication Date
2024-12-06
Estimated Expiration
2044-07-31

AI Technical Summary

Technical Problem

Traditional power special report generation methods are difficult to effectively deal with multi-element composite problems, resulting in poor quality of report content and complex database information, resulting in low search efficiency.

Method used

Using a method based on large-scale and multi-document course learning framework, the power designated project query text is semanticly understood through preset large language models, key information is extracted, and associated files are integrated with the multi-document course learning framework to generate streamlined file information, and according to Report type generates high-quality special reports.

Benefits of technology

It improves the efficiency and quality of power special reports, can meet the multiple needs more quickly and accurately, and improves the overall quality and efficiency of power reports.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119088911A_ABST
    Figure CN119088911A_ABST
Patent Text Reader

Abstract

The invention relates to a power report generation method based on a large model and a multi-document course learning framework. The method comprises the following steps: in response to a question and answer operation on an electric power customer service platform, obtaining an electric power specified item query text of the question and answer operation; performing semantic understanding on the power specified project query text by adopting a preset large language model, and extracting to obtain query key information; the query key information is used for indicating a target query range, a target query object and a target report type; performing retrieval according to a target query range and a target query object in the query key information to obtain a plurality of power specified project associated files; integrating a plurality of power specified project associated files by adopting a multi-document curriculum learning framework to generate simplified file information; and according to the simplified file information and the target report type, generating a power specified project report as a query result of the power specified project query text. By adopting the method, a high-quality and multi-demand power special report generation service can be provided.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of electric power technology, and in particular to an electric power report generation method, apparatus, computer equipment, computer-readable storage medium and computer program product based on a large model and multi-document course learning framework. Background Art

[0002] At present, when the customer service platform of the power industry processes the demand for specific special report generation, it faces various problems and demands. For example, users can ask questions and inquire about power equipment failures, power load forecasts, power supply plans, etc.

[0003] In the related technology, the traditional special report generation method can support answering users' single-element question needs, but it is not effective in handling multi-element complex questions. In addition, due to the complexity of the information stored in the database, the traditional retrieval method is inefficient, resulting in poor quality of the content of the final generated report. Summary of the invention

[0004] Based on this, it is necessary to provide an electricity report generation method, device, computer equipment, computer-readable storage medium and computer program product based on a large model and multi-document course learning framework, which can improve the efficiency and effect of electricity special report generation in response to the above technical problems.

[0005] In a first aspect, the present application provides a method for generating an electric power report based on a large model and a multi-document course learning framework, comprising:

[0006] In response to a question-and-answer operation on the electric power customer service platform, obtaining an electric power designated item query text of the question-and-answer operation;

[0007] A preset large language model is used to perform semantic understanding on the query text of the power designated project, and key query information is extracted; the key query information is used to indicate the target query scope, target query object, and target report type;

[0008] Retrieve according to the target query scope and the target query object in the query key information to obtain multiple power designated project associated files;

[0009] A multi-document course learning framework is used to integrate the multiple power-specified project-related files to generate streamlined file information;

[0010] According to the simplified file information and the target report type, an electricity designated item report is generated as a query result of the electricity designated item query text.

[0011] In one embodiment, the method further comprises:

[0012] Construct training data and training prompt words based on power specified project query tasks;

[0013] A basic large language model for model training is obtained, and model parameters of the basic large language model are adjusted using the training data and the training prompt words until a training end condition is met, thereby obtaining the preset large language model.

[0014] In one embodiment, the step of constructing training data and training prompt words based on the power specified project query task includes:

[0015] Based on the power designated project query task, sample power designated project query texts of different project types and sample query key information corresponding to each of the sample power designated project query texts are obtained as the training data; the sample query key information has a marked label;

[0016] Based on the power designated project query task, execution instructions, power project related information, and input text indication information for the basic large language model are designed as the training prompt words.

[0017] In one embodiment, the method further comprises:

[0018] The multi-document course learning framework is constructed by applying the course learning strategy to the multi-document summarization task and combining it with the multi-document difficulty scoring function to perform model training on the preset backbone model.

[0019] In one embodiment, the method of applying the curriculum learning strategy to the multi-document summarization task and combining the multi-document difficulty scoring function to perform model training on the preset backbone model includes:

[0020] Acquire multiple sample document sets, and use the multi-document difficulty scoring function to calculate the difficulty score of each of the sample document sets;

[0021] Sorting the multiple sample document sets according to the difficulty score of each of the sample document sets;

[0022] Based on the sorted multiple sample document sets, the preset backbone model is trained by means of a rolling window curriculum.

[0023] In one embodiment, the method further comprises:

[0024] The multi-document difficulty scoring function is determined based on the number of documents in the document collection, the total length of the documents, and the distribution of the document lengths.

[0025] In a second aspect, the present application also provides a power report generation device based on a large model and a multi-document course learning framework, comprising:

[0026] A query text acquisition module, for obtaining a query text of a power designated item of the question and answer operation in response to a question and answer operation on the power customer service platform;

[0027] A key information extraction module, used to use a preset large language model to perform semantic understanding on the power designated project query text, and extract key query information; the key query information is used to indicate the target query scope, target query object, and target report type;

[0028] A retrieval file obtaining module is used to perform a search according to the target query scope and the target query object in the query key information to obtain multiple power designated project related files;

[0029] A simplified information integration module, used to integrate the plurality of power-specified project-related files using a multi-document course learning framework to generate simplified file information;

[0030] The power report generation module is used to generate a power designated project report according to the simplified file information and the target report type as a query result of the power designated project query text.

[0031] In a third aspect, the present application further provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0032] In response to a question-and-answer operation on the electric power customer service platform, obtaining an electric power designated item query text of the question-and-answer operation;

[0033] A preset large language model is used to perform semantic understanding on the query text of the power designated project, and key query information is extracted; the key query information is used to indicate the target query scope, target query object, and target report type;

[0034] Retrieve according to the target query scope and the target query object in the query key information to obtain multiple power designated project associated files;

[0035] A multi-document course learning framework is used to integrate the multiple power-specified project-related files to generate streamlined file information;

[0036] According to the simplified file information and the target report type, an electricity designated item report is generated as a query result of the electricity designated item query text.

[0037] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the following steps are implemented:

[0038] In response to a question-and-answer operation on the electric power customer service platform, obtaining an electric power designated item query text of the question-and-answer operation;

[0039] A preset large language model is used to perform semantic understanding on the query text of the power designated project, and key query information is extracted; the key query information is used to indicate the target query scope, target query object, and target report type;

[0040] Retrieve according to the target query scope and the target query object in the query key information to obtain multiple power designated project associated files;

[0041] A multi-document course learning framework is used to integrate the multiple power-specified project-related files to generate streamlined file information;

[0042] According to the simplified file information and the target report type, an electricity designated item report is generated as a query result of the electricity designated item query text.

[0043] In a fifth aspect, the present application further provides a computer program product, including a computer program, which implements the following steps when executed by a processor:

[0044] In response to a question-and-answer operation on the electric power customer service platform, obtaining an electric power designated item query text of the question-and-answer operation;

[0045] A preset large language model is used to perform semantic understanding on the query text of the power designated project, and key query information is extracted; the key query information is used to indicate the target query scope, target query object, and target report type;

[0046] Retrieve according to the target query scope and the target query object in the query key information to obtain multiple power designated project associated files;

[0047] A multi-document course learning framework is used to integrate the multiple power-specified project-related files to generate streamlined file information;

[0048] According to the simplified file information and the target report type, an electricity designated item report is generated as a query result of the electricity designated item query text.

[0049] The above-mentioned power report generation method, device, computer equipment, computer-readable storage medium and computer program product based on a large model and a multi-document course learning framework obtains the power designated project query text of the question and answer operation by responding to the question and answer operation on the power customer service platform, and then uses a preset large language model to perform semantic understanding on the power designated project query text, extracts the query key information, and the query key information is used to indicate the target query range, target query object, and target report type. According to the target query range and target query object in the query key information, a search is performed to obtain multiple power designated project related files, and then the multi-document course learning framework is used to integrate multiple The system can generate simplified file information from multiple power designated project-related files, and generate power designated project reports based on the simplified file information and target report types as the query results of the power designated project query text, thereby realizing the automatic generation optimization of power special reports. By adopting the collaborative mode of large model and multi-document course learning framework, the query text can be understood and processed based on the large model to extract key information, and the multi-document course learning framework can be used to quickly integrate and generate simplified information from multiple retrieved files, so that special report generation can be faster and more accurate, which is helpful to provide high-quality and multi-demand special report generation services, and effectively improve the quality and efficiency of power reports. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0051] Figure 1 A flowchart of a method for generating an electric power report based on a large model and a multi-document course learning framework in one embodiment;

[0052] Figure 2 A schematic diagram of a power report generation process based on a large model and a multi-document course learning framework in one embodiment;

[0053] Figure 3 A flowchart of a method for generating an electric power report based on a large model and a multi-document course learning framework in another embodiment;

[0054] Figure 4 A structural block diagram of an electric power report generating device based on a large model and a multi-document course learning framework in one embodiment;

[0055] Figure 5 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0056] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0057] In an exemplary embodiment, Figure 1 As shown, a method for generating an electric power report based on a large model and a multi-document course learning framework is provided. This embodiment uses the method applied to a terminal as an example. It can be understood that the method can also be applied to a server, and can also be applied to a system including a terminal and a server, and is implemented through the interaction between the terminal and the server. In this embodiment, the method includes the following steps 101 to 104. Among them:

[0058] Step 101, in response to a question and answer operation on a power customer service platform, obtaining a query text of a power designated item of the question and answer operation.

[0059] In practical applications, such as Figure 2 As shown, in response to the question-and-answer operation on the power customer service platform, the query information input by the user can be obtained as the power designated item query text.

[0060] Step 102: Use a preset large language model to perform semantic understanding on the power designated project query text to extract key query information.

[0061] The query key information may be used to indicate the target query scope, target query object, and target report type.

[0062] After obtaining the query text of the power specified project, such as Figure 2 As shown, a preset large language model can be used to understand and retrieve key query information, such as keywords used to indicate the target query scope, target query object, and target report type, so as to further input the keywords extracted by the model into the database for retrieval to find the corresponding documents.

[0063] Exemplarily, in the semantic understanding stage, the preset large language model can be used to perform entity recognition tasks, and by semantically understanding the input power specified project query text, the user-specified report type (i.e., target report type) and keywords required for report generation (i.e., target query range, target query object) can be extracted, such as the time range involved in the data to be queried, the name of the data set to which the data to be queried belongs, etc. Optionally, for different report types, the extracted multiple keywords can be classified and sorted, such as in chronological order, according to different data sets, etc.

[0064] Step 103 , searching according to the target query scope and the target query object in the query key information to obtain a plurality of power designated project associated files.

[0065] In the specific implementation, Figure 2 As shown, relevant documents can be retrieved in the information database according to the target query scope and target query object in the query key information, such as the time range involved in the data to be queried, the name of the data set to which the data to be queried belongs, etc., and then multiple power designated project associated files can be obtained.

[0066] Step 104 : using a multi-document course learning framework to integrate the plurality of power-specified project-related files to generate simplified file information.

[0067] As an example, the multi-document curriculum learning framework may be a curriculum learning (CL) enhancement framework for multi-document summarization (MDS), namely, MDS-CL.

[0068] After associating files in multiple power-specified projects, such as Figure 2 As shown, a multi-document course learning framework can be used to integrate the retrieved multiple power-specified project-related documents, and then streamlined document information, such as a streamlined report summary, can be generated.

[0069] Specifically, for the multi-document course learning framework, the task of generating summaries based on documents retrieved from the information database can be used as a multi-document summarization task to generate concise summaries based on multiple documents.

[0070] Step 105 : generating an electricity designated project report according to the simplified file information and the target report type as a query result of the electricity designated project query text.

[0071] As an example, the report types may include, but are not limited to, general notifications, specific project reports, project summary reports, and specific task data analysis reports.

[0072] In practical applications, the user can select the type of special report (i.e. target report type) required by the user, such as Figure 2 The report type 1 and the generated concise summary (i.e., concise file information) generate the special report required by the user, realizing that the model generates the corresponding report content based on the specified report format.

[0073] In one example, the technical solution of this embodiment, by utilizing a method of coordinating a large language model with a multi-document course learning enhancement framework, can use a large language model to read user input information for understanding, and retrieve keywords to enter the information database, and can use a multi-document summary course learning enhancement framework to efficiently generate concise summary content. By combining the data understanding ability of the multi-document course learning enhancement framework and the language retrieval and generation ability of the large model, efficient extraction, analysis and report generation of multi-source data in the power field are achieved, making the power special report generation process more intelligent and efficient, improving the generation quality and generation efficiency of the power special report, and being able to better meet industry needs.

[0074] In the above-mentioned power report generation method based on the big model and multi-document course learning framework, by responding to the question and answer operation on the power customer service platform, the query text of the power designated project of the question and answer operation is obtained, and then the preset big language model is used to perform semantic understanding on the query text of the power designated project, and the query key information is extracted. According to the target query range and the target query object in the query key information, a search is performed to obtain multiple power designated project related files, and then the multi-document course learning framework is used to integrate the multiple power designated project related files to generate streamlined file information. According to the streamlined file information and the target report type, a power designated project report is generated as the query result of the power designated project query text, and the automatic generation optimization of the power special report is realized. By adopting the collaborative mode of the big model and the multi-document course learning framework, the query text can be understood and processed based on the big model to extract key information, and the multi-document course learning framework can be used to quickly integrate and generate streamlined information from multiple retrieved files, so that the special report generation can be faster and more accurate, which is helpful to provide high-quality and multi-demand special report generation services, and effectively improves the quality and efficiency of power reports.

[0075] In an exemplary embodiment, the following steps may also be included:

[0076] Construct training data and training prompt words based on the power designated project query task; obtain a basic large language model for model training, and use the training data and the training prompt words to adjust the model parameters of the basic large language model until the training end condition is met to obtain the preset large language model.

[0077] In practical applications, key information extraction can be regarded as a named entity recognition problem. Key information can be extracted based on the basic large language model by using prompt fine-tuning. By using the open source large language model as the base framework and performing prompt fine-tuning on it, the model can adjust parameters by comparing the generated answers with the correct answers during the prompt fine-tuning process to better adapt to specific tasks.

[0078] In one example, in the prompt fine-tuning stage, fine-tuning data can be collected based on the power designated project query task and prompt words (Prompt), that is, training data and training prompt words, can be designed; through the prompt project for a few samples of fine-tuning, the large language model can effectively deal with the problem of named entity recognition in special report generation. After training the model with only a small amount of labeled data, it can also maintain a high accuracy in the testing phase, thereby realizing the fine-tuning of the large model for specific tasks based on the prompt project and the fine-tuning data set.

[0079] In this embodiment, by constructing training data and training prompt words based on the power designated project query task, a basic large language model for model training is obtained, and the model parameters of the basic large language model are adjusted using the training data and training prompt words until the training end conditions are met, and a preset large language model is obtained. The parameters of the large language model for the power designated project query task can be adjusted based on the prompt words and training data.

[0080] In an exemplary embodiment, the construction of training data and training prompt words based on the power specified project query task may include the following steps:

[0081] Based on the power designated project query task, sample power designated project query texts of different project types and sample query key information corresponding to each of the sample power designated project query texts are obtained as the training data; the sample query key information has annotated labels; based on the power designated project query task, execution instructions, power project related information, and input text indication information for the basic large language model are designed as the training prompt words.

[0082] In a specific implementation, a fine-tuning dataset for a power-specific report can be constructed and prompt words can be designed to adjust the parameters of a large language model using the prompt words and the fine-tuning dataset. For example, in order to enable the model to complete the named entity recognition problem, the fine-tuning dataset can be a task-specific dataset with labels, which can include multiple types of user input content and annotated key information corresponding to each content, that is, sample power-specified project query texts of different project types, and sample query key information corresponding to each sample power-specified project query text, and the sample query key information has annotated labels.

[0083] For another example, prompt words (i.e., training prompt words) can contain multiple elements: instructions, background information, and input text. Among them, instructions (i.e., execution instructions) can be short sentences that control the model to perform a specific task. For example, when processing a special power report, the prompt word instruction can be "As a special power report salesperson, extract keyword information based on user input content"; background information (i.e., power project related information) can be relevant information provided for few-sample learning. The background information of the prompt word can include "keywords include time such as x year / x month / x day to y year / y month / y day, the name of the data set to which the query data belongs, and the report type; input text (i.e., input text indication information) can be used to determine the text that needs to be processed by the model, such as user input content in fine-tuning data.

[0084] In this embodiment, based on the power designated project query task, sample power designated project query texts of different project types and sample query key information corresponding to each sample power designated project query text are obtained as training data, and then based on the power designated project query task, execution instructions, power project related information, and input text indication information for the basic large language model are designed as training prompts, which provide data support for the large language model to adjust parameters for the power designated project query task.

[0085] In an exemplary embodiment, the following steps may also be included:

[0086] The multi-document course learning framework is constructed by applying the course learning strategy to the multi-document summarization task and combining it with the multi-document difficulty scoring function to perform model training on the preset backbone model.

[0087] In one example, a curriculum learning enhancement framework for multi-document summarization can enhance multi-document summarization based on curriculum learning strategies by applying curriculum learning strategies to multi-document summarization tasks. In order to distinguish between easy examples and difficult examples, a multi-document difficulty scoring function can be designed for multiple source documents, which can be composed of three difficulty heuristics, such as the number of documents, the total length of documents, and the length distribution. Then, a rolling window curriculum method can be used to arrange the sorted data set, so that the preset backbone model can be effectively fine-tuned.

[0088] In this embodiment, by applying the course learning strategy to the multi-document summarization task and combining it with the multi-document difficulty scoring function, the preset backbone model is trained to construct a multi-document course learning framework. The data understanding ability of the framework can be enhanced based on the multi-document course learning, and concise summary content can be efficiently generated, thereby improving the generation quality and efficiency of power special reports.

[0089] In an exemplary embodiment, the method of applying the curriculum learning strategy to the multi-document summarization task and combining the multi-document difficulty scoring function to perform model training on the preset backbone model may include the following steps:

[0090] Acquire multiple sample document sets, and use the multi-document difficulty scoring function to calculate the difficulty score of each of the sample document sets; sort the multiple sample document sets according to the difficulty score of each of the sample document sets; and train the preset backbone model based on the sorted multiple sample document sets by means of a rolling window course.

[0091] In practical applications, for building a multi-document course learning framework, a multi-document difficulty scoring function can be used to calculate the difficulty score of each document set (i.e., a sample document set). Then, the document sets can be sorted according to the difficulty score, and then a rolling window course training preset backbone model can be used.

[0092] In one example, by defining a multi-document difficulty scoring function, a preset backbone model can be trained based on a rolling window curriculum. At this stage, the training data can be arranged into a learning curriculum based on the sorted document set to train the adopted model.

[0093] In an optional embodiment, the following course method can be used to verify the effectiveness of course learning;

[0094] 1. No curriculum: Random sampling from the original dataset for training;

[0095] 2. One-Pass course: start training from the simplest example to the most difficult example, and do not move to the next batch of data until the model converges;

[0096] 3. Baby-Steps course: Similar to the One-Pass course, the difference is that it can gradually increase the complexity of the training data instead of discarding simpler examples.

[0097] 4. Reverse course: It is the reverse course of the One-Pass course, which can start training from the most difficult examples to the simplest examples.

[0098] 5. Rolling Window course: It combines the advantages of One-Pass course and Baby-Steps course. It can sample training samples through rolling and sliding windows, so that the model can be trained more efficiently and the risk of overfitting can be avoided.

[0099] In this embodiment, by obtaining multiple sample document sets, using a multi-document difficulty scoring function to calculate the difficulty score of each sample document set, and then sorting the multiple sample document sets according to the difficulty score of each sample document set, and then based on the sorted multiple sample document sets, the preset backbone model is trained through a rolling window course, which can achieve efficient generation of concise summary content using a multi-document summary course learning enhancement framework.

[0100] In an exemplary embodiment, the following steps may also be included:

[0101] The multi-document difficulty scoring function is determined based on the number of documents in the document collection, the total length of the documents, and the distribution of the document lengths.

[0102] In the specific implementation, the difficulty score plays an important role in course learning, which can be used to distinguish between simple examples and difficult examples. Traditional difficulty indicators proposed for text examples, such as n-gram frequency, specificity, word rarity, and sentence length, cannot be directly used for multi-document summarization tasks. Since the difficulty of multi-document summarization is affected by many factors, the technical solution of this embodiment proposes a heuristic method to measure the difficulty score of each document set, such as three heuristic methods that can be considered, including the number of documents, the total length of documents, and the distribution of document lengths.

[0103] For example, a dataset can be represented as a collection of M document sets, , where each set is N about the same topic i Document sequence . Accordingly, each document set D i Each has a sequence of document length The document characteristics of interest are given below, including the number of documents, the total length of documents, and the distribution of document lengths. Based on these three indicators, a multi-document difficulty scoring function can be determined.

[0104] Regarding the number of documents, different multi-document sets have different numbers of documents. The more documents there are, the more difficult it is to summarize. i The difficulty is defined as follows:

[0105]

[0106] Among them, Diff Di Represents a multi-document set D i The difficulty score, N i It is a multi-document set D i The number of documents.

[0107] Regarding the total length of the document, the total length of the document can reflect the multi-document set D iThe total length difficulty of the summary. Multiple documents with a longer total length of summary have a higher difficulty because they contain more information. The total length difficulty of the document is defined as follows:

[0108]

[0109] Among them, L i It is a multi-document set D i The total length of the document.

[0110] Regarding the document length distribution, the document length distribution can reflect the relative proportion of the length of each document in the multi-document collection. The difficulty score of a document with a more uneven distribution of length is lower than that of a uniform distribution. For example, if the length of a document accounts for a high proportion of the total length, its difficulty score is low and it can be regarded as a pseudo-single document. Calculate the document set D i The difficulty heuristic for the document length distribution is as follows:

[0111]

[0112] in, It is the document set D i The maximum document length in .

[0113] Based on the above heuristics, we can propose a multi-document summary task difficulty scoring function (i.e., multi-document difficulty scoring function), which can be formalized as follows:

[0114]

[0115] Among them, the hyperparameters =512, Represents a multi-document difficulty scoring function.

[0116] In this embodiment, by determining a multi-document difficulty scoring function based on the number of documents in the document set, the total length of the documents, and the document length distribution, the difficulty score of each document set can be effectively measured.

[0117] In an exemplary embodiment, Figure 3 As shown, a flow chart of another method for generating a power report based on a large model and a multi-document course learning framework is provided. In this embodiment, the method includes the following steps:

[0118] In step 301, based on the power designated project query task, sample power designated project query texts of different project types and sample query key information corresponding to each sample power designated project query text are obtained as training data; the sample query key information has a labeled label. In step 302, based on the power designated project query task, execution instructions, power project related information, and input text indication information for the basic large language model are designed as training prompt words. In step 303, the basic large language model for model training is obtained, and the model parameters of the basic large language model are adjusted using the training data and the training prompt words until the training end condition is met to obtain the preset large language model. In step 304, by applying the course learning strategy to the multi-document summarization task and combining the multi-document difficulty scoring function, the preset backbone model is trained to construct a multi-document course learning framework. In step 305, in response to the question-and-answer operation on the power customer service platform, the power designated project query text of the question-and-answer operation is obtained, and the preset large language model is used to perform semantic understanding on the power designated project query text to extract the query key information. In step 306, a search is performed based on the target query scope and target query object in the query key information to obtain multiple power designated project related files. In step 307, a multi-document course learning framework is used to integrate multiple power designated project related files to generate streamlined file information, and a power designated project report is generated based on the streamlined file information and the target report type as the query result of the power designated project query text. It should be noted that the specific limitations of the above steps can be found in the specific limitations of a power report generation method based on a large model and a multi-document course learning framework above, which will not be repeated here.

[0119] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0120] Based on the same inventive concept, the embodiment of the present application also provides a device for generating an electric report based on a large model and a multi-document course learning framework for implementing the above-mentioned method for generating an electric report based on a large model and a multi-document course learning framework. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above-mentioned method, so the specific limitations in one or more embodiments of the electric report generating device based on a large model and a multi-document course learning framework provided below can be referred to the limitations of the electric report generating method based on a large model and a multi-document course learning framework above, and will not be repeated here.

[0121] In an exemplary embodiment, Figure 4 As shown, a power report generation device based on a large model and a multi-document course learning framework is provided, comprising:

[0122] The query text acquisition module 401 is used to obtain the query text of the power specified item of the question and answer operation in response to the question and answer operation on the power customer service platform;

[0123] The key information extraction module 402 is used to use a preset large language model to perform semantic understanding on the power designated project query text to extract the query key information; the query key information is used to indicate the target query scope, target query object, and target report type;

[0124] A retrieval file obtaining module 403 is used to perform a search according to the target query scope and the target query object in the query key information to obtain multiple power designated project related files;

[0125] A simplified information integration module 404 is used to integrate the plurality of power-specified project-related files using a multi-document course learning framework to generate simplified file information;

[0126] The power report generation module 405 is used to generate a power designated item report according to the simplified file information and the target report type as a query result of the power designated item query text.

[0127] In one embodiment, the apparatus further comprises:

[0128] A training information building module, used to build training data and training prompt words based on power specified project query tasks;

[0129] The model training module is used to obtain a basic large language model for model training, and use the training data and the training prompt words to adjust the model parameters of the basic large language model until the training end condition is met to obtain the preset large language model.

[0130] In one embodiment, the model training module includes:

[0131] A training data acquisition submodule is used to obtain sample power designated project query texts of different project types and sample query key information corresponding to each sample power designated project query text as the training data based on the power designated project query task; the sample query key information has a marked label;

[0132] The training prompt word obtaining submodule is used to design execution instructions, power project related information, and input text indication information for the basic large language model as the training prompt words based on the power designated project query task.

[0133] In one embodiment, the apparatus further comprises:

[0134] The multi-document course learning framework construction module is used to construct the multi-document course learning framework by applying the course learning strategy to the multi-document summarization task and combining the multi-document difficulty scoring function to perform model training on the preset backbone model.

[0135] In one embodiment, the multi-document course learning framework building module includes:

[0136] A difficulty score calculation submodule, used to obtain multiple sample document sets and calculate the difficulty score of each of the sample document sets using the multi-document difficulty score function;

[0137] A sorting submodule, used to sort the multiple sample document sets according to the difficulty score of each sample document set;

[0138] The backbone model training submodule is used to train the preset backbone model through a rolling window course based on the sorted multiple sample document sets.

[0139] In one embodiment, the apparatus further comprises:

[0140] The difficulty scoring function definition module is used to determine the multi-document difficulty scoring function based on the number of documents in the document collection, the total length of the documents, and the distribution of the document lengths.

[0141] Each module in the above-mentioned power report generation device based on the large model and multi-document course learning framework can be implemented in whole or in part by software, hardware and their combination. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0142] In an exemplary embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as shown in FIG. Figure 5 As shown. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit and an input device. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface, the display unit and the input device are connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a mobile cellular network, a near field communication (Near Field Communication, NFC) or other technologies. When the computer program is executed by the processor, a method for generating an electric power report based on a large model and a multi-document course learning framework is implemented.

[0143] Those skilled in the art will understand that Figure 5 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0144] In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented:

[0145] In response to a question-and-answer operation on the electric power customer service platform, obtaining an electric power designated item query text of the question-and-answer operation;

[0146] A preset large language model is used to perform semantic understanding on the query text of the power designated project, and key query information is extracted; the key query information is used to indicate the target query scope, target query object, and target report type;

[0147] Retrieve according to the target query scope and the target query object in the query key information to obtain multiple power designated project associated files;

[0148] A multi-document course learning framework is used to integrate the multiple power-specified project-related files to generate streamlined file information;

[0149] According to the simplified file information and the target report type, an electricity designated item report is generated as a query result of the electricity designated item query text.

[0150] In one embodiment, when the processor executes the computer program, it also implements the steps of the power report generation method based on the large model and multi-document course learning framework in the other embodiments mentioned above.

[0151] In one embodiment, a computer readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0152] In response to a question-and-answer operation on the electric power customer service platform, obtaining an electric power designated item query text of the question-and-answer operation;

[0153] A preset large language model is used to perform semantic understanding on the query text of the power designated project, and key query information is extracted; the key query information is used to indicate the target query scope, target query object, and target report type;

[0154] Retrieve according to the target query scope and the target query object in the query key information to obtain multiple power designated project associated files;

[0155] A multi-document course learning framework is used to integrate the multiple power-specified project-related files to generate streamlined file information;

[0156] According to the simplified file information and the target report type, an electricity designated item report is generated as a query result of the electricity designated item query text.

[0157] In one embodiment, when the computer program is executed by a processor, it also implements the steps of the power report generation method based on the large model and multi-document course learning framework in the other embodiments described above.

[0158] In one embodiment, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the following steps:

[0159] In response to a question-and-answer operation on the electric power customer service platform, obtaining an electric power designated item query text of the question-and-answer operation;

[0160] A preset large language model is used to perform semantic understanding on the query text of the power designated project, and key query information is extracted; the key query information is used to indicate the target query scope, target query object, and target report type;

[0161] Retrieve according to the target query scope and the target query object in the query key information to obtain multiple power designated project associated files;

[0162] A multi-document course learning framework is used to integrate the multiple power-specified project-related files to generate streamlined file information;

[0163] According to the simplified file information and the target report type, an electricity designated item report is generated as a query result of the electricity designated item query text.

[0164] In one embodiment, when the computer program is executed by a processor, it also implements the steps of the power report generation method based on the large model and multi-document course learning framework in the other embodiments described above.

[0165] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0166] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0167] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0168] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be construed as limiting the scope of the present application. It should be noted that, for a person of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for generating power reports based on a large model and multi-document course learning framework, characterized in that: The method comprises: In response to a question-and-answer operation on the electric power customer service platform, obtaining an electric power designated item query text of the question-and-answer operation; A preset large language model is used to perform semantic understanding on the power designated project query text to extract key query information; the key query information is used to indicate a target query scope, a target query object, and a target report type; Retrieve according to the target query scope and the target query object in the query key information to obtain multiple power designated project associated files; A multi-document course learning framework is used to integrate the multiple power-specified project-related files to generate streamlined file information; According to the simplified file information and the target report type, an electricity designated item report is generated as a query result of the electricity designated item query text.

2. The method according to claim 1, characterized in that: The method further comprises: Construct training data and training prompt words based on power specified project query tasks; A basic large language model for model training is obtained, and model parameters of the basic large language model are adjusted using the training data and the training prompt words until a training end condition is met, thereby obtaining the preset large language model.

3. The method according to claim 2, characterized in that The construction of training data and training prompt words based on the power specified project query task includes: Based on the power designated project query task, sample power designated project query texts of different project types and sample query key information corresponding to each of the sample power designated project query texts are obtained as the training data; the sample query key information has a marked label; Based on the power designated project query task, execution instructions, power project related information, and input text indication information for the basic large language model are designed as the training prompt words.

4. The method according to claim 1, characterized in that The method further comprises: The multi-document course learning framework is constructed by applying the course learning strategy to the multi-document summarization task and combining it with the multi-document difficulty scoring function to perform model training on the preset backbone model.

5. The method according to claim 4, characterized in that The method applies the curriculum learning strategy to the multi-document summarization task and combines the multi-document difficulty scoring function to train the preset backbone model, including: Acquire multiple sample document sets, and use the multi-document difficulty scoring function to calculate the difficulty score of each of the sample document sets; Sorting the multiple sample document sets according to the difficulty score of each of the sample document sets; Based on the sorted multiple sample document sets, the preset backbone model is trained by means of a rolling window curriculum.

6. The method according to claim 4, characterized in that The method further comprises: The multi-document difficulty scoring function is determined based on the number of documents in the document collection, the total length of the documents, and the distribution of the document lengths.

7. A power report generation device based on a large model and multi-document course learning framework, characterized in that: The device comprises: A query text acquisition module, for obtaining a query text of a power designated item of the question and answer operation in response to a question and answer operation on the power customer service platform; A key information extraction module, used to use a preset large language model to perform semantic understanding on the power designated project query text, and extract key query information; the key query information is used to indicate the target query scope, target query object, and target report type; A retrieval file obtaining module is used to perform a search according to the target query scope and the target query object in the query key information to obtain multiple power designated project related files; A simplified information integration module, used to integrate the plurality of power-specified project-related files using a multi-document course learning framework to generate simplified file information; The power report generation module is used to generate a power designated project report according to the simplified file information and the target report type as a query result of the power designated project query text.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Citation Information

Patent Citations

  • Multi-document scientific abstract generation method based on topic knowledge graph joint enhancement

    CN116821371A

  • Knowledge retrieval enhancement generation method and system based on large language model

    CN118394890A

  • Multi-document intelligent question and answer method and system based on large language model

    CN118394897A

Cited By

  • Automatic verification report generation method based on intelligent customer service

    CN120235122A