Report generation method and device, electronic equipment and storage medium

Abstract

The invention provides a report generation method and device, electronic equipment and a storage medium, and relates to the technical field of artificial intelligence such as large language model processing and deep learning. The specific implementation scheme is as follows: obtaining a report generation request of a user, wherein the report generation request is used for requesting to obtain a to-be-generated report; according to the report generation request, the reference data and the first large language model, determining a question text list corresponding to the report to be generated and an answer text of a first type of question text in the question text list; obtaining answer texts of the second class of question texts, wherein the question text list comprises the second class of question texts to be answered in addition to the first class of question texts; and generating a to-be-generated report according to the reference data, the answer text of the question text in the question text list and the second large language model. Therefore, the accurate and comprehensive question text list can be determined without depending on manpower, the accuracy and integrity of the generated to-be-generated report are ensured, and the user experience is improved.

Description

Technical Field

[0001] This disclosure relates to the field of computer technology, and in particular to the fields of artificial intelligence technology such as large language model processing and deep learning, and especially to a report generation method, apparatus, electronic device and storage medium. Background Technology

[0002] Large Language Models (LLMs) are a type of deep learning-based artificial intelligence model that is trained on large amounts of text data and is able to understand, generate, and reason about natural language.

[0003] In related technologies, large language models are used to generate professional reports based on similar reports. To avoid severe illusions, it is necessary to manually extract question texts or create question text templates. The answer texts of the question texts are obtained by asking the professionals who wrote the report, and then used as input for the large language model. This method requires a lot of manpower to extract question texts or create question text templates. The manual cost is high and the efficiency is low. Moreover, it cannot guarantee that the question texts are accurate and comprehensive, and it cannot guarantee the completeness and comprehensiveness of the generated professional report. This is a problem that urgently needs to be solved. Summary of the Invention

[0004] This disclosure provides a report generation method, apparatus, electronic device, and storage medium.

[0005] According to a first aspect of this disclosure, a report generation method is provided, comprising: obtaining a user's report generation request, the report generation request being used to request the acquisition of a report to be generated; determining, based on the report generation request, reference materials, and a first major language model, a list of question texts corresponding to the report to be generated, and answer texts of first-type question texts in the question text list; acquiring answer texts of second-type question texts, the question text list including unanswered second-type question texts in addition to the first-type question texts; and generating the report to be generated based on the reference materials, the answer texts of the question texts in the question text list, and the second major language model.

[0006] According to a second aspect of this disclosure, a report generation apparatus is provided, comprising: a first information acquisition module, configured to acquire a user's report generation request, the report generation request being used to request the acquisition of a report to be generated; a first processing module, configured to determine, based on the report generation request, reference materials, and a first major language model, a list of question texts corresponding to the report to be generated, and answer texts for first-type question texts in the question text list; a second information acquisition module, configured to acquire answer texts for second-type question texts, the question text list including unanswered second-type question texts in addition to the first-type question texts; and a second processing module, configured to generate the report to be generated based on reference materials, answer texts for question texts in the question text list, and the second major language model.

[0007] According to a third aspect of this disclosure, an electronic device is provided, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the method described in the first aspect above.

[0008] According to a fourth aspect of this disclosure, a non-transitory computer-readable storage medium is provided storing computer instructions, wherein the computer instructions are configured to cause the computer to perform the method described in the first aspect above.

[0009] According to a fifth aspect of this disclosure, a computer program product is provided, comprising a computer program, wherein the computer program, when executed by a processor, implements the steps of the method described in the first aspect above.

[0010] According to the technical solution of this disclosure, the answer texts of the first type of question texts and the answer texts of the second type of question texts in the question text list corresponding to the report to be generated are determined based on the first language model and reference materials. There is no need to manually extract question texts or create question text templates. While saving labor costs and improving processing efficiency, it can ensure that the determined question text list is accurate and comprehensive. Combined with the answer texts of the question texts in the question text list, the accuracy and completeness of the generated report to be generated are ensured.

[0011] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0012] The accompanying drawings are provided to better understand this solution and do not constitute a limitation of this disclosure. Wherein:

[0013] Figure 1This is a flowchart of a report generation method provided in an embodiment of this disclosure;

[0014] Figure 2 This is a flowchart of another report generation method provided in this disclosure embodiment;

[0015] Figure 3 This is a flowchart of step S203 of the report generation method provided in this embodiment;

[0016] Figure 4 This is a block diagram of a report generation apparatus provided in an embodiment of this disclosure;

[0017] Figure 5 This is a block diagram of a first processing module in the report generation apparatus provided in this embodiment of the disclosure;

[0018] Figure 6 This is a block diagram of another report generation apparatus provided in an embodiment of this disclosure;

[0019] Figure 7 This is a block diagram of an electronic device used to implement the report generation method of the embodiments of this disclosure. Detailed Implementation

[0020] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0021] It should be noted that the collection, storage, use, processing, transmission, provision, and disclosure of user personal information involved in the technical solution disclosed herein all comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0022] It should be noted that the information (including but not limited to user device information, user personal information, etc.), data (including but not limited to data used for analysis, data stored, data displayed, etc.) and signals involved in this disclosure are all authorized by the user or fully authorized by all parties, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0023] It should also be noted that in the embodiments disclosed herein, certain software, components, models and other existing solutions in the industry may be mentioned. These should be considered as exemplary and are intended only to illustrate the feasibility of implementing the technical solutions disclosed herein. However, they do not mean that the applicant has used or necessarily used such solutions.

[0024] In this embodiment of the disclosure, a method based on a large language model is proposed to interact with users to obtain information and generate reports required by the users.

[0025] In generating reports, besides providing similar reports, the professionals writing the reports also need to provide certain information as input. Without this information, simply generating reports using a large language model would result in significant errors. Manually extracting this information and asking questions to the writers to assist in report generation would consume a large amount of manpower to create a list of question texts. Furthermore, the answer texts for some questions can be obtained from additional knowledge bases provided by the client or from already generated reports. There's no need for professionals to repeat their answers to these questions. This technology can directly extract the answer texts from the question texts, requiring only information not currently available to the system from the user.

[0026] In this embodiment, key information for report generation can be automatically extracted and used to generate corresponding question texts for user inquiry. Before generating the report, it is ensured that these question texts are answered by professionals. Finally, the final report is generated based on the answers to these question texts.

[0027] In summary, this technology can automatically extract information needed for report writing that is not yet available, and by proactively asking users questions, it reduces the illusion of reports generated by large language models.

[0028] The report generation method provided in this disclosure is applicable to product needs in scenarios such as professional reports in certain industries, such as construction and power. Most reports in these industries have fixed formats, and information such as outlines can be extracted from existing similar reports. However, some detailed textual information about the report to be generated, such as the project name and specific data, needs to be provided by the user. An automated process that proactively obtains this information from the user before generating the report, and asynchronously generates a professional report after the user provides complete information, can save users considerable time and improve work efficiency.

[0029] Prior to the solutions proposed in this disclosure, related technologies, similar to deep research, involve generating several question texts before report generation, and then generating the report based on the answers to these question texts. Another approach, similar to question text templates, involves manually extracting the question texts, editing them into templates, forcing users to input them through a fixed program or panel, and then generating the report based on the answers to the question texts. These related technologies require significant manpower for question text extraction and template creation, resulting in high labor costs and low efficiency. Furthermore, they cannot guarantee the accuracy and comprehensiveness of the question texts, thus failing to ensure the completeness and comprehensiveness of the generated professional report. This is a problem that urgently needs to be solved.

[0030] Based on this, embodiments of this disclosure provide a report generation method, apparatus, electronic device, and storage medium. The method includes: obtaining a user's report generation request, the report generation request being used to request the acquisition of a report to be generated; determining, based on the report generation request, reference materials, and a first major language model, a list of question texts corresponding to the report to be generated, and answer texts for a first type of question text in the question text list; acquiring answer texts for a second type of question text, the question text list including unanswered second type of question texts in addition to the first type of question texts; and generating the report to be generated based on the reference materials, the answer texts of the question texts in the question text list, and the second major language model. Thus, there is no need for manual question extraction or question template creation, saving labor costs and improving processing efficiency while ensuring the accuracy and comprehensiveness of the determined question text list. Combined with the answer texts of the question texts in the question text list, the accuracy and completeness of the generated report to be generated are ensured.

[0031] The report generation method, apparatus, electronic device, and storage medium of this disclosure are described below with reference to the accompanying drawings. It should be noted that the report generation method in the embodiments of this disclosure can be applied to report generation in various industries that require lengthy reports, such as construction, power, healthcare, and finance.

[0032] Figure 1 This is a flowchart of a report generation method provided in an embodiment of this disclosure. Figure 1 As shown, the method for generating this report may include, but is not limited to, the following steps.

[0033] It should be noted that the execution subject of the report generation method in this embodiment is a report generation device, which can be implemented by software and / or hardware. The device can be configured in an electronic device, which may include, but is not limited to, a terminal, a server, etc.

[0034] S101, Obtain the user's report generation request. The report generation request is used to request the report to be generated.

[0035] In this embodiment of the disclosure, a report generation request from the user is obtained, which is used to request the generation of a report to be generated.

[0036] In some embodiments, a report generation request includes at least one of the following: the type, number, and index of the report to be generated.

[0037] In this embodiment of the disclosure, by numbering the reports to distinguish different types of reports, the user's report generation request is obtained. If the report generation request includes the report number, the report to be generated requested by the user can be determined.

[0038] In this embodiment of the disclosure, by adding an index to the report to distinguish different types of reports, the user's report generation request can be obtained. If the report generation request includes an index, the report to be generated requested by the user can be determined.

[0039] In some embodiments, the report generation request may also include references. Of course, the references may also be locally stored, meaning the report generation request may not include references.

[0040] In some embodiments, the references include at least one of the following:

[0041] Information about existing reports related to the report to be generated;

[0042] A list of historical issue texts related to existing reports that are to be generated;

[0043] The report to be generated has completed the following content;

[0044] The guidelines for writing historical reports for users whose existing reports are related to the report to be generated.

[0045] In this embodiment of the disclosure, the reference materials include at least one of the following: information on existing reports related to the report to be generated, a list of historical problem texts corresponding to existing reports related to the report to be generated, the content of the completed parts of the report to be generated, and the historical report writing specifications of users corresponding to existing reports related to the report to be generated.

[0046] In some embodiments, the existing report associated with the report to be generated is a report of the same type as the report to be generated, or a report with the same number, or a report with the same index.

[0047] In this embodiment of the disclosure, the reference materials include information from existing reports related to the report to be generated, thereby enabling the adaptive identification of key information to be completed based on the information from existing reports and a large language model, so as to determine a comprehensive and accurate list of problem texts based on the key information and meet the requirements for report generation.

[0048] In this embodiment of the disclosure, the reference materials include a list of historical problem texts corresponding to existing reports related to the report to be generated. This allows the historical problem text list to be identified as the problem text list corresponding to the report to be generated, eliminating the need to use a large language model to identify the problem text list, saving the time required to determine the problem text list, and ensuring that the determined problem text list is comprehensive and accurate.

[0049] In this embodiment of the disclosure, the reference materials include the completed portion of the report to be generated, thereby enabling the generation of answer texts for a portion of the question texts in the question text list based on the completed portion of the report to be generated, reducing the number of questions that the user needs to answer, saving labor costs and improving processing efficiency.

[0050] In this embodiment of the disclosure, the reference materials include the historical report writing specifications of users corresponding to existing reports related to the report to be generated. This allows the generation of the report to be generated to be combined with the historical report writing specifications of users corresponding to existing reports related to the report to be generated. This makes the generated report to be generated more in line with the user's report writing specifications, saves the user's modification time, and improves processing efficiency.

[0051] It should be noted that existing reports can also be called historical reports, and the writing standards for historical reports can also be called historical writing habits, existing writing habits, existing writing standards, existing writing requirements, etc.

[0052] In some embodiments, the references also include existing materials. These existing materials are used to determine the answer texts for the questions in the list of question texts.

[0053] In some embodiments, reference materials are obtained by uploading by users, or stored in a knowledge base or database, or obtained from the knowledge base or database. Alternatively, they may be partially obtained by uploading by users, partially stored in a knowledge base or database, or obtained from the knowledge base or database.

[0054] S102, based on the report generation request, reference materials, and the first major language model, determine the answer text corresponding to the report to be generated, as well as the first type of question text in the question text list.

[0055] In this embodiment of the disclosure, upon receiving a user's report generation request, the report to be generated requested by the user can be determined.

[0056] Specifically, based on the report generation request, reference materials, and the first major language model, a list of question texts corresponding to the report to be generated, as well as the answer texts for the first type of question texts in the question text list, are determined.

[0057] In some embodiments, based on the report generation request, information about existing reports related to the report to be generated, the content of the completed part of the report to be generated, and existing data are obtained from the reference materials. The information about existing reports related to the report to be generated, the content of the completed part of the report to be generated, and existing data are input into the first language model to generate a list of question texts corresponding to the report to be generated, as well as the answer texts of the first type of question texts in the list of question texts.

[0058] In some embodiments, based on the report generation request, information about existing reports related to the report to be generated, the content of the completed part of the report to be generated, and existing materials are obtained from the reference materials. Based on the chapter structure, outline, key parameters, and other information of the existing reports, the report to be generated is compared to determine the missing key information to be completed. Then, the key information, the content of the completed part of the report to be generated, and the existing materials are input into the first language model to generate a list of question texts corresponding to the report to be generated, as well as the answer texts of the first type of question texts in the question text list.

[0059] In some embodiments, the problem text list includes problem texts corresponding to information missing from the report to be generated relative to existing reports.

[0060] For example, compared to an existing report, the report to be generated is missing: project name, key parameters, time range, numerical indicators, etc. The list of questions could include: What is the project name of the report to be generated? What are the values ​​of the key parameters in the report to be generated? What is the time range of the report to be generated? What are the numerical indicators in the report to be generated? etc.

[0061] The first language model can determine the project name and time range based on the completed portion of the report to be generated, and determine important parameters based on existing data. This allows it to determine the answer text for the first type of question and the text for the second type of question to be answered. Examples of the first type of question text include: What is the name of the project to be generated? What is the time range of the report to be generated? What are the values ​​of the important parameters in the report to be generated? Examples of the second type of question text include: What are the numerical indicators in the report to be generated?

[0062] In some embodiments, the list of question texts may include a second type of question text to be answered in addition to the first type of question texts. In this case, S103 is executed.

[0063] In some embodiments, the text list includes only the first type of text, in which case S103 may be omitted.

[0064] S103, Obtain the answer text for the second type of question.

[0065] In some embodiments, when a second type of question text to be answered is determined, the answer text of the second type of question text is obtained by at least one of the following methods:

[0066] Ask users questions;

[0067] Use the search service tool to perform a query;

[0068] Ask a professional a question;

[0069] Use artificial intelligence (AI) tools to perform the query.

[0070] In some embodiments, professionals include users.

[0071] In this embodiment of the disclosure, the answer text of the user's answer to the second type of question is obtained by asking the user a question.

[0072] In this embodiment of the disclosure, the answer text for the second type of question text is obtained by calling a retrieval service tool. The retrieval service tool is a search engine, a professional database and academic platform, a vertical industry platform, etc.

[0073] In some embodiments, by entering the text of the second type of question into a search engine, public information can be quickly matched and located based on keywords to obtain the answer text of the second type of question.

[0074] In some embodiments, the answer text to the second type of question can be retrieved by inputting the second type of question text into a professional database or academic platform.

[0075] In this embodiment of the disclosure, the answer text of the user's answer to the second type of question is obtained by asking questions to professionals.

[0076] In this embodiment of the disclosure, the answer text of the user's answer to the second type of question is obtained by calling an AI tool to perform a query.

[0077] In some embodiments, the AI ​​tool inputs a second type of question text and outputs the answer text to the second type of question text.

[0078] S104: Generate a report to be generated based on the reference materials, the answer texts to the questions in the question text list, and the second major language model.

[0079] In this embodiment of the disclosure, information from existing reports related to the report to be generated in the reference materials; the content of the completed part of the report to be generated; the historical report writing specifications of the user corresponding to the existing report to be generated; and the answer text of the question text in the question text list are input into the second language model to generate the report to be generated.

[0080] In some embodiments, the first major language model and the second major language model may be the same or different.

[0081] In this embodiment of the disclosure, when the first and second major language models are the same, the major language model first determines the answer texts of the first type of question texts and the unanswered second type of question texts in the question text list corresponding to the report to be generated, based on the user's report generation request and the information of existing reports related to the report to be generated in the reference materials, the historical question text list corresponding to the existing reports related to the report to be generated, and existing materials. After obtaining the answer texts of the second type of question texts, the major language model generates the report to be generated based on the information of existing reports related to the report to be generated in the reference materials; the content of the completed part of the report to be generated; the user's historical report writing specifications corresponding to the existing reports related to the report to be generated; and the answer texts of the question texts in the question text list.

[0082] By implementing the embodiments of this disclosure, a user's report generation request is obtained, which requests the retrieval of a report to be generated. Based on the report generation request, reference materials, and a first major language model, the answer texts for the first type of question texts and the unanswered second type of question texts in the question text list corresponding to the report to be generated are determined. The answer texts for the second type of question texts are obtained. Based on the reference materials, the answer texts of the question texts in the question text list, and the second major language model, the report to be generated is generated. Therefore, there is no need for manual question extraction or question template creation. This saves labor costs and improves processing efficiency while ensuring the accuracy and comprehensiveness of the determined question text list. Combined with the answer texts of the question texts in the question text list, the accuracy and completeness of the generated report to be generated are ensured, thus improving the user experience.

[0083] Figure 2 This is a flowchart of the report generation method provided in this embodiment of the disclosure. Figure 2 As shown, the method for generating this report may include, but is not limited to, the following steps.

[0084] S201, Obtain the user's report generation request. The report generation request is used to request the report to be generated.

[0085] S201 can be implemented in any of the embodiments of this disclosure. This disclosure does not limit the implementation of this method and will not elaborate further.

[0086] In some embodiments, a report generation request is used to request a specified section of the report to be generated.

[0087] In this embodiment of the disclosure, the user can request the generation of the report to be generated by chapter. Through the chapter-based generation mechanism, the generated report is guaranteed to be of sufficient length and logically complete, reaching the industry standard of tens of thousands of words.

[0088] S202, Does the reference material contain a list of historical problem texts related to the existing reports to be generated?

[0089] In this embodiment of the disclosure, when a user requests a report to be generated, it is possible to determine existing reports related to the report to be generated in the reference materials, and to determine whether there is a corresponding list of historical problem texts in the existing reports.

[0090] In some embodiments, the list of historical issue texts corresponding to existing reports is determined by the large language model based on reports related to the existing reports.

[0091] S203, if there is no corresponding historical question text list for existing reports related to the report to be generated in the reference materials, determine the question text list corresponding to the report to be generated, as well as the answer text of the first type of question text in the question text list, based on the report generation request, the reference materials and the first major language model.

[0092] S203 can be implemented in any of the embodiments of this disclosure. This disclosure does not limit this implementation and will not elaborate further.

[0093] like Figure 3 As shown, in some embodiments, S203 includes, but is not limited to, the following steps:

[0094] S2031, Based on the report generation request, determine the information of existing reports related to the report to be generated in the reference materials.

[0095] In this embodiment of the disclosure, based on the type of the report to be generated, a report of the same type as the report to be generated is identified as an existing report related to the report to be generated.

[0096] In this embodiment of the disclosure, a report with the same number as the report to be generated is identified as an existing report related to the report to be generated, based on the number of the report to be generated.

[0097] In this embodiment of the disclosure, based on the index of the report to be generated, a report with the same index as the report to be generated is identified as an existing report related to the report to be generated.

[0098] In this embodiment of the disclosure, the information in the existing report includes the chapters, outline, key parameters, title, etc. of the existing report.

[0099] S2032, Based on information from existing reports, identify the key information to be supplemented.

[0100] In this embodiment of the disclosure, the key information to be supplemented includes: information that differs from existing reports, information unique to the report to be generated, and information without a unified standard. Examples include project name, important parameters, time intervals, and numerical indicators.

[0101] S2033, based on key information, reference materials, and the first major language model, determine the list of question texts, and the answer texts for the first type of question texts in the list of question texts.

[0102] In this embodiment of the disclosure, by inputting key information and reference materials into the first language model, a list of question texts and the answer texts of the first type of question texts in the list of question texts are determined.

[0103] In this embodiment of the disclosure, without relying on manual extraction of question text or creation of question text templates, the system determines a list of question texts and the answer texts for the first type of question texts in the question text list based on key information, reference materials, and a first major language model determined from information in existing reports related to the report to be generated. This adaptively identifies missing key information to obtain a comprehensive and accurate list of question texts. Furthermore, by first determining the answer texts for the first type of question texts in the question text list, the system can save time in obtaining answer texts, reduce workload, and improve processing efficiency.

[0104] In some embodiments, S2033, determining a list of question texts and answer texts for the first type of question texts in the list of question texts based on key information, reference materials, and a first major language model includes: determining a list of question texts based on key information and a first major language model; and determining answer texts for the first type of question texts in the list of question texts based on reference materials.

[0105] In this embodiment of the disclosure, the first language model automatically generates a list of question texts based on key information. Further, the answer texts for the first type of question texts in the question text list are determined based on reference materials, while the questions in the question text list that do not have answer texts are the second type of question texts to be answered.

[0106] In this embodiment of the disclosure, the first language model can determine an accurate and comprehensive list of question texts based on key information. Furthermore, it can first determine the answer texts of the first type of question texts in the list of question texts, which can save time in obtaining answer texts, reduce workload, and improve processing efficiency.

[0107] S204, if there is a corresponding historical problem text list in the reference materials for an existing report related to the report to be generated, the historical problem text list shall be determined as the problem text list corresponding to the report to be generated.

[0108] Among them, the list of historical issue texts corresponding to existing reports was determined by the large language model based on reports related to existing reports.

[0109] S205, Based on the reference materials, determine the answer text for the first type of question text in the question text list corresponding to the report to be generated.

[0110] In this embodiment of the disclosure, if a historical problem text list exists in the reference materials that corresponds to an existing report related to the report to be generated, the historical problem text list is determined as the problem text list corresponding to the report to be generated. This eliminates the need to re-determine the problem text list based on the large language model, thereby improving the efficiency of problem text list generation.

[0111] In this embodiment of the disclosure, when a list of question texts corresponding to the report to be generated is determined, the answer texts for the first type of question texts can be determined based on reference materials. This saves time in obtaining the answer texts for the question texts in the question text list, reduces workload, and further improves processing efficiency.

[0112] In some embodiments, the list of question texts may include a second type of question text to be answered in addition to the first type of question texts. In this case, S206 is executed.

[0113] In some embodiments, the text list includes only the first type of text, in which case S206 may be omitted.

[0114] S206, Obtain the answer text for the second type of question.

[0115] In some embodiments, S206, obtaining the answer text of the second type of question text includes: providing the second type of question text to the user; obtaining the initial answer text of the user's answer to the second type of question text; verifying the initial answer text; and obtaining the answer text of the second type of question text.

[0116] In this embodiment of the disclosure, a second type of question text is provided to the user through an interactive interface or a question-and-answer process, and the initial answer text of the user in answering the second type of question text is obtained.

[0117] In this embodiment, the initial answer text is validated to ensure its completeness and consistency. For example, the accuracy of units, clarity of numerical ranges, and absence of missing questions are checked. After validation, the answer text for the second type of question is obtained. This ensures the completeness and consistency of the obtained answer text for the second type of question, improving the completeness and consistency of the generated report.

[0118] S207: Generate a report to be generated based on the reference materials, the answer texts to the questions in the question text list, and the second major language model.

[0119] S207 can be implemented in any of the embodiments of this disclosure. This disclosure does not limit this implementation and will not elaborate further.

[0120] In some embodiments, when the report generation request is used to request the acquisition of a specified chapter of the report to be generated, the list of question texts corresponding to the report to be generated is the list of question texts corresponding to the specified chapter of the report to be generated, wherein generating the report to be generated includes: generating the specified chapter of the report to be generated.

[0121] In this embodiment, the report generation request is used to request the acquisition of a specified chapter of the report to be generated. Based on the reference materials and the first major language model, the answer texts of the first type of question texts and the unanswered second type of question texts in the question text list corresponding to the specified chapter of the report to be generated are generated. The answer texts of the second type of question texts are acquired. Based on the reference materials, the answer texts of the question texts in the question text list, and the second major language model, the specified chapter of the report to be generated is generated. Thus, the user can request the generation of the report to be generated by chapter. Through the chapter-based generation mechanism, the generated report is guaranteed to be of sufficient length and logically complete, meeting the industry standard of tens of thousands of words, thereby improving the user experience.

[0122] In some embodiments, the method further includes: providing a report to be generated to a user; obtaining a modification record of the user's modifications to the report to be generated; and generating a modified report to be generated based on the modification record.

[0123] In this embodiment of the disclosure, after the second language model generates the report to be generated, the report is provided to the user for review and inspection. The user can modify the report, and the modification record of the user's modifications is obtained. The modified report is then generated based on the modification record. This allows the user to check the report generated by the second language model, ensuring the completeness and accuracy of the final modified report and improving the user experience.

[0124] In some embodiments, the method further includes storing a record of modifications made by the user to the report to be generated in a reference.

[0125] In some embodiments, modification records are used by the second language model to generate a report in conjunction with the modification records, thereby reducing the time spent on modification by users and improving processing efficiency.

[0126] In some embodiments, the method further includes: determining the user's report writing specifications based on the modification records; and updating the user's historical report writing specifications corresponding to existing reports related to the report to be generated in the reference materials based on the user's report writing specifications.

[0127] In this embodiment of the disclosure, the user's report writing specifications are determined based on the modification records.

[0128] In some embodiments, modifying a record includes modifying at least one of the date format, table of contents format, font format, and title format.

[0129] For example, if a user's modification record is that they changed the date format from year-month-day to day-month-year, then the user's report writing standard is determined to be that the date format is day-month-year.

[0130] For example, the user's modification record is to change the directory format from the same format for all directories to different formats for different levels of directories, thus determining the user's report writing standard as using different formats for different levels of directories.

[0131] For example, if a user's modification record shows that they changed the font format from other fonts to KaiTi or SongTi, then the user's report writing standard is to use KaiTi or SongTi font.

[0132] In this embodiment of the disclosure, the historical report writing specifications of the user corresponding to the existing report related to the report to be generated in the reference materials are updated according to the user's report writing specifications. As a result, when the second language model generates the report to be generated, it can combine the user's report writing specifications to generate the report, thereby avoiding repeated modifications by the user to the report to be generated. This makes the generated report to be generated more in line with the user's writing specifications, improves the quality of report generation, and enhances the user's experience.

[0133] For example, in the case where the first and second large language models in this disclosure embodiment are the same large language model, the process of generating the report to be generated by the large language model specifically includes: clarifying the report's objectives and audience: First, the analytical purpose of the report must be clearly defined, and the target audience must be determined, such as professionals in various industries such as construction, power, healthcare, and finance. Simultaneously, the main content framework of the report must be clarified, such as data sources, analytical methods, key findings, etc., as well as listing the text of the problems to be solved and the text of the desired answers.

[0134] Data preparation: Collect relevant data from sources such as databases, APIs, and CSV files, and then handle missing values, outliers, and duplicate values ​​to ensure data quality. Next, convert the data into a format suitable for analysis, such as numericalization or normalization, and finally upload the data to a local computing environment or cloud service.

[0135] Design prompts: Guided prompts inform the model in advance about the semantics and generation order of structured fields. For example, "You are an experienced financial analyst. Please write a concise quarterly report based on the following metrics. Metrics include: (Field 1:xxx), (Field 2:xxx)... Output format: Paragraph 1-xxx; Paragraph 2-xxx..." This approach can significantly reduce field omission rates and language style drift.

[0136] Generate preliminary content: Utilize a large language model to generate descriptive analysis content for the data, including data distribution, key features, etc. Simultaneously, select appropriate data analysis methods, such as linear regression and logistic regression, and have the model generate explanations and theoretical background for the analysis methods, as well as interpretations and discussions of the analysis results.

[0137] Content Integration and Optimization: The generated content is integrated according to the report structure, including introduction, methods, results, discussion, and conclusions. A large language model is used to generate coherent paragraphs and transitional sentences to ensure clear report logic. The generated content is then refined and revised, checking accuracy and logic. Based on review comments, the content can also be revised and optimized using the large language model.

[0138] Data visualization: Use tools such as Matplotlib, Seaborn, Plotly, or Tableau to generate charts and provide chart data and related descriptions. Utilize large language models to generate detailed interpretations of the charts, explaining key trends and findings.

[0139] Final review and publication: The report undergoes a comprehensive manual review, checking for accuracy, logic, and language quality. After further revisions and optimizations based on review comments, the report is published through appropriate channels, such as the company intranet or academic journals.

[0140] By implementing the embodiments of this disclosure, there is no need to manually extract question text or create question text templates. This saves labor costs and improves processing efficiency while ensuring that the determined question text list is accurate and comprehensive. Combined with the answer text of the question text in the question text list, the accuracy and completeness of the generated report are ensured, thus improving the user experience.

[0141] To enable those skilled in the art to better understand this disclosure, the following exemplary embodiments are provided.

[0142] In an exemplary embodiment, a report generation method based on a large language model is provided. Before generating a professional report using a large language model (LLM), key information is automatically generated and confirmed, and the accuracy and completeness of the generated report are ensured through interaction with professionals.

[0143] The specific process includes, but is not limited to:

[0144] 1. Input and Report Analysis:

[0145] The system reads existing similar or identical reports and automatically extracts the report structure (such as chapters, outlines, and key parameters).

[0146] By comparing the current report to be generated, identify the key missing information (such as project name, important parameters, time range, numerical indicators, etc.).

[0147] 2. Question text generation and filtering:

[0148] Based on the missing information, LLM automatically generates corresponding questions.

[0149] The system will perform a second screening of these question texts:

[0150] If the answer text can be automatically extracted from the completed parts of the knowledge base, materials, or reports uploaded by the user, then it can be added directly without asking the question again.

[0151] The question text will only be submitted to professionals when the system is unable to automatically obtain the answer text.

[0152] 3. Interaction with professionals:

[0153] The system presents a list of prepared questions to professionals through an interactive interface or a question-and-answer process.

[0154] Professionals answer each question one by one, and the system verifies the completeness and consistency of the answers in real time (such as checking units, numerical ranges, and whether there are any omissions).

[0155] 4. Generate a draft report:

[0156] Once the key information is complete, the system calls LLM to generate a report.

[0157] When writing a report, LLM combines similar reports, existing data, and newly collected information, as well as user habits recorded in the system, such as date format and table of contents format, to generate a complete report that meets the user's needs.

[0158] 5. Results Optimization and Feedback:

[0159] The system provides drafts for professionals to review, and professionals can make minor modifications in the interface.

[0160] The system records these modifications for future optimization of the question text generation and answer text extraction processes (e.g., date format).

[0161] The system will record the information of the user during the report generation process, optimize the process, and facilitate future generation, such as (1) recording the user's text habits, such as date format, directory format, etc. (2) recording the problem text list of the same type of report. When the user generates the same type of report again, these lists can be quickly retrieved from storage to improve the response speed, etc.

[0162] Using the above embodiments, key information extraction can be automated: without relying on manually created question text templates, the system can adaptively identify missing information points. Intelligent question-and-answer filtering is possible: avoiding repeated questions and reducing user burden. Completeness checks are performed: ensuring all key information is obtained before generating the report, reducing the "illusion" risk of LLM. Optimization for long documents is possible: this technology can generate chapters sequentially, avoiding the "input-output imbalance" problem text of large language models—where the output tokens are relatively few, making it difficult to generate long documents. Simultaneously, when generating later chapter content, the already generated chapter content is referenced to ensure consistency in long document content.

[0163] By employing the above embodiments, we can reduce illusions: ensuring report generation is based on true and complete information. We achieve efficiency: professionals only need to supplement necessary information, reducing unnecessary workload. We improve versatility: applicable to various industries requiring lengthy reports, such as construction, power, healthcare, and finance. We enable intelligent evolution: as usage increases, the system automatically learns user preferences and industry standards, continuously improving the quality of generated reports.

[0164] Figure 4 This is a block diagram of the report generation apparatus 100 provided in an embodiment of this disclosure. Figure 3 As shown, the report generation device 100 includes: a first information acquisition module 1, a first processing module 2, a second information acquisition module 3, and a second processing module 4.

[0165] The first information acquisition module 1 is used to acquire the user's report generation request, which is used to request the acquisition of the report to be generated.

[0166] The first processing module 2 is used to determine the list of question texts corresponding to the report to be generated, as well as the answer texts of the first type of question texts and the second type of question texts to be answered, based on the report generation request, reference materials and the first major language model.

[0167] The second information acquisition module 3 is used to acquire the answer text of the second type of question text. In addition to the first type of question text, the question text list also includes the second type of question text to be answered.

[0168] The second processing module 4 is used to generate a report to be generated based on the reference materials, the answer texts of the questions in the question text list, and the second major language model.

[0169] In some embodiments, the first processing module 2 is further configured to: determine the historical problem text list as the problem text list corresponding to the report to be generated if there is a corresponding historical problem text list in the reference materials related to the existing report to be generated; and determine the problem text list corresponding to the report to be generated, and the answer text of the first type of problem text in the problem text list, according to the reference materials.

[0170] like Figure 5 As shown, in some embodiments, the first processing module 2 includes: a first information determination submodule 21, a second information determination submodule 22, and a processing submodule 23.

[0171] The first information determination submodule 21 is used to determine the information of existing reports related to the report to be generated in the reference materials based on the report generation request.

[0172] The second information determination submodule 22 is used to determine the key information to be supplemented based on the information in the existing reports.

[0173] Processing submodule 23 is used to determine the list of question texts and the answer texts of the first type of question texts in the list of question texts based on key information, reference materials and the first major language model.

[0174] In some embodiments, the processing submodule 23 is specifically used to: determine a list of question texts based on key information and a first major language model; and determine the answer texts for the first type of question texts in the list of question texts based on reference materials.

[0175] In some embodiments, the second information acquisition module 3 is specifically used to: provide the user with a second type of question text; acquire the user's initial answer text for answering the second type of question text; verify the initial answer text, and acquire the answer text for the second type of question text.

[0176] like Figure 6 As shown, in some embodiments, the report generation device may further include: a report providing module 5, a modification record acquisition module 6, and a modification processing module 7.

[0177] Among them, the report provision module 5 is used to provide users with the report to be generated.

[0178] Modification record acquisition module 6 is used to acquire the modification records of user modifications to the report to be generated.

[0179] Modification processing module 7 is used to generate a modified report to be generated based on the modification records.

[0180] like Figure 6As shown, in some embodiments, the report generation apparatus may further include: a specification determination module 8 and a specification update module 9.

[0181] The specification determination module 8 is used to determine the user's report writing specifications based on the modification records.

[0182] The specification update module 9 is used to update the user's historical report writing specifications for existing reports related to the report to be generated in the reference materials, based on the user's report writing specifications.

[0183] Regarding the report generation apparatus in the above embodiments, the specific methods by which each module performs its operations have been described in detail in the embodiments of the report generation method described above, and will not be repeated here.

[0184] The beneficial effects achieved by the report generation device in the above embodiments are the same as those achieved by the report generation method described above, and will not be repeated here.

[0185] According to embodiments of this disclosure, this disclosure also provides an electronic device and a readable storage medium.

[0186] like Figure 7 The diagram shown is a block diagram of an electronic device according to an embodiment of the present disclosure. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0187] like Figure 7 As shown, the electronic device includes one or more processors 701, a memory 702, and interfaces for connecting the components, including high-speed interfaces and low-speed interfaces. The components are interconnected via different buses and can be mounted on a common motherboard or otherwise as required. The processors can process instructions executed within the electronic device, including instructions stored in or on memory to display graphical information of a GUI on an external input / output device (such as a display device coupled to the interface). In other embodiments, multiple processors and / or multiple buses can be used with multiple memories and multiple memory modules, if desired. Similarly, multiple electronic devices can be connected, each providing some of the necessary operations (e.g., as a server array, a group of blade servers, or a multiprocessor system). Figure 7 Take the 701 processor as an example.

[0188] The memory 702 is the non-transitory computer-readable storage medium provided in this disclosure. The memory stores instructions executable by at least one processor to cause the at least one processor to perform the report generation method provided in this disclosure. The non-transitory computer-readable storage medium of this disclosure stores computer instructions for causing a computer to perform the report generation method provided in this disclosure.

[0189] Memory 702, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs, non-transitory computer-executable programs, and modules, such as the program instructions / modules corresponding to the report generation method in the embodiments of this disclosure (e.g., appendix). Figure 5 The first extraction module 501, the second extraction module 502, the encoding module 503, the determination module 504, and the processing module 505 are shown. The processor 701 executes various functional applications and data processing of the server by running non-transient software programs, instructions, and modules stored in the memory 702, thereby implementing the report generation method in the above method embodiments.

[0190] The memory 702 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the electronic device. Furthermore, the memory 702 may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, the memory 702 may optionally include memory remotely located relative to the processor 701, and these remote memories can be connected to the electronic device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0191] The electronic device may also include an input device 703 and an output device 704. The processor 701, memory 702, input device 703, and output device 704 can be connected via a bus or other means. Figure 7 Taking the example of a connection between China and Israel via a bus.

[0192] Input device 703 can receive input numerical or character information, and generate key signal inputs related to user settings and function control of the electronic device, such as touch screens, keypads, mice, trackpads, touchpads, joysticks, one or more mouse buttons, trackballs, joysticks, etc. Output device 704 may include display devices, auxiliary lighting devices (e.g., LEDs), and haptic feedback devices (e.g., vibration motors). The display device may include, but is not limited to, liquid crystal displays (LCDs), light-emitting diode (LED) displays, and plasma displays. In some embodiments, the display device may be a touch screen.

[0193] Various implementations of the systems and techniques described herein can be implemented in digital electronic circuit systems, integrated circuit systems, application-specific integrated circuits (ASICs), computer hardware, firmware, software, and / or combinations thereof. These various implementations may include: implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transferring data and instructions to the storage system, the at least one input device, and the at least one output device.

[0194] These computational programs (also referred to as programs, software, software applications, or code) include machine instructions for a programmable processor and can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. As used herein, the terms “machine-readable medium” and “computer-readable medium” refer to any computer program product, device, and / or apparatus (e.g., disk, optical disk, memory, programmable logic device (PLD)) used to provide machine instructions and / or data to a programmable processor, including machine-readable media that receive machine instructions as machine-readable signals. The term “machine-readable signal” refers to any signal used to provide machine instructions and / or data to a programmable processor.

[0195] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0196] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or middleware components (e.g., application servers), or frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), the Internet, and blockchain networks.

[0197] Computer systems can include clients and servers. Clients and servers are generally geographically separated and typically interact via communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. A server can be a cloud server, also known as a cloud computing server or cloud host, a hosting product within the cloud computing service ecosystem, addressing the shortcomings of traditional physical hosts and VPS (Virtual Private Server) services, such as high management difficulty and weak business scalability. Servers can also be servers for distributed systems or servers incorporating blockchain technology.

[0198] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this application can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0199] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A report generation method, comprising: Obtain the user's report generation request, which is used to request the report to be generated; Based on the report generation request, reference materials, and the first major language model, determine the list of question texts corresponding to the report to be generated, and the answer texts of the first type of question texts in the list of question texts; Obtain the answer text for the second type of question text, wherein the list of question texts includes the second type of question texts to be answered in addition to the first type of question texts; The report to be generated is generated based on the reference materials, the answer texts of the questions in the question text list, and the second major language model.

2. The method as described in claim 1, wherein, The method further includes: If a historical problem text list exists in the reference materials that corresponds to the existing report to be generated, the historical problem text list is determined to be the problem text list corresponding to the report to be generated. Based on the reference materials, determine the answer texts for the first type of question texts in the question text list corresponding to the report to be generated.

3. The method as described in claim 1, wherein, The step of determining the list of question texts corresponding to the report to be generated, and the answer texts of the first type of question texts in the list of question texts, based on the report generation request, reference materials, and the first major language model, includes: Based on the report generation request, determine the information of existing reports related to the report to be generated in the reference materials; Based on the information in the existing reports, the key information to be supplemented has been identified; Based on the key information, the reference materials, and the first large language model, the list of question texts and the answer texts for the first type of question texts in the list of question texts are determined.

4. The method of claim 3, wherein, The step of determining the list of question texts and the answer texts of the first type of question texts in the list of question texts based on the key information, the reference materials, and the first large language model includes: Based on the key information and the first large language model, the list of question texts is determined; Based on the reference materials, determine the answer text for the first type of question text in the question text list.

5. The method of claim 1, wherein, The step of obtaining the answer text for the second type of question text includes: Provide the user with the text of the second type of question; Obtain the initial answer text from the user's response to the second type of question text; The initial answer text is validated to obtain the answer text for the second type of question text.

6. The method of claim 1, wherein, The method further includes: Provide the user with the report to be generated; Obtain the modification records of the user's modifications to the report to be generated; The modified report to be generated is generated based on the modification records.

7. The method of claim 6, wherein, The method further includes: The user's report writing guidelines are determined based on the modification records; Based on the user's report writing guidelines, update the user's historical report writing guidelines in the reference materials corresponding to the existing reports related to the report to be generated.

8. The method according to any one of claims 1 to 7, wherein, The report generation request is used to request a specified chapter of the report to be generated, and the list of question texts corresponding to the specified chapter of the report to be generated is the list of question texts corresponding to the specified chapter of the report to be generated. Generating the report to be generated includes: Generate the specified section of the report to be generated.

9. The method according to any one of claims 1 to 7, wherein, The references include at least one of the following: Information about existing reports related to the report to be generated; A list of historical issue texts corresponding to existing reports related to the report to be generated; The completed portion of the report to be generated; The historical report writing guidelines for the user corresponding to the existing reports related to the report to be generated.

10. A report generation apparatus, comprising: The first information acquisition module is used to acquire the user's report generation request, wherein the report generation request is used to request the acquisition of the report to be generated; The first processing module is used to determine, based on the report generation request, reference materials and the first major language model, the list of question texts corresponding to the report to be generated, and the answer texts of the first type of question texts in the list of question texts; The second information acquisition module is used to acquire the answer text of the second type of question text. The question text list includes the second type of question text to be answered in addition to the first type of question text. The second processing module is used to generate the report to be generated based on the reference materials, the answer texts of the questions in the question text list, and the second major language model.

11. An electronic device, comprising: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1 to 9.

12. A non-transitory computer-readable storage medium storing computer instructions, wherein, The computer instructions are used to cause the computer to perform the method according to any one of claims 1 to 9.

13. A computer program product comprising a computer program, wherein, When executed by a processor, the computer program implements the steps of the method according to any one of claims 1 to 9.

Images