Electronic archive auxiliary editing and research method, system and equipment based on generative AI and medium

Abstract

The invention provides an electronic archive auxiliary editing and research method, system and device based on generative AI and a medium, and belongs to the technical field of electronic archive editing and research, and the method comprises the steps: obtaining a theme, summary information or selected archive data input by a user; generating an AI cue word according to the editing and research operation, generating an archive data query instruction or theme keyword by the generative AI, and querying a database to obtain target data; then desensitizing the target data and generating abstract cues, and sequentially generating editing and research abstracts by the generation formula AI; the abstract desensitization information is restored, editing and research operation suggestions are generated to the user, and the user adjusts and modifies according to the suggestions. According to the method, the query instruction and the keyword can be efficiently generated, and the target archive data can be accurately obtained. The desensitization processing ensures the data security, restores the desensitization information and generates suggestions, so that the user can adjust the editing and research based on the complete information, the editing and research efficiency and accuracy are improved, and the editing and research requirements are met.

Description

Technical Field

[0001] This invention belongs to the field of electronic archives compilation and research technology, specifically relating to a method, system, equipment, and medium for electronic archives assisted compilation and research based on generative AI. Background Technology

[0002] Archival compilation and research involves archives departments compiling reference materials, archiving documents, participating in historical compilation, local history writing, or writing papers and monographs based on their collections and societal needs, while also studying the content of the archives. Users typically begin in two ways: given a research topic and abstract, they then select relevant archives from the archives; or conversely, given a set of archives, they write a reasonable topic and abstract based on their content. After determining the topic, abstract, and archival content, users compile them into a book within the archival system.

[0003] Currently, generative artificial intelligence technology is demonstrating enormous application potential in multiple fields, and its semantic understanding and content generation capabilities offer new insights for archival research and compilation. However, some related technologies generate prompts based on pre-set templates without considering user-input terminology, sentence style, or context, resulting in significant discrepancies between AI-generated query commands or summaries and actual needs. For example, when a user inputs "organize project acceptance documents," the tool only generates commands to retrieve project acceptance files, ignoring the user's implicit need to link them to project budgets, thus failing to meet the user's requirements.

[0004] When using AI, data anonymization is generally required to ensure data security. However, if the anonymization rules are not differentiated from the research and compilation scenarios, semantic breaks may occur after anonymization. This necessitates tracing back to the original files for verification, which not only increases the workload but also causes subsequent summaries to fail to reflect the chronological logic. As a result, the research and compilation results are missing key information, requiring the anonymization rules to be readjusted and the work to be reworked, thus affecting the quality and efficiency of the research and compilation. Summary of the Invention

[0005] This invention provides a generative AI-based method for assisting in the compilation and research of electronic archives. The method can efficiently generate query commands and keywords to accurately obtain target archive data. Data security is ensured through anonymization processing, and the anonymized information is restored and suggestions are generated, allowing users to adjust the compilation and research based on complete information, thereby improving efficiency and accuracy.

[0006] The methods include: S101: Obtain the user-inputted topic, summary information, or selected archival data, define the compilation and research operations for the compilation and research task, and provide the execution basis for subsequent prompt word generation, archival retrieval, and desensitization processing; S102: Generate corresponding AI prompt words based on the above editing and research operations; S103: Generative AI generates file data query instructions or topic-related keywords based on the AI ​​prompts; S104: Query the archive database according to the query instruction or topic-related keywords to obtain the target archive data; S105: De-identify the target archive data and generate prompt words for summary generation; S106: Generative AI generates a research summary based on the prompt words; S107: Restore the de-identified information in the compilation summary and generate compilation operation suggestions to present to the user; S108: The user makes adjustments and modifications based on the aforementioned editing and research operation suggestions.

[0007] Preferably, step S108 specifically includes the following steps: Collect user feedback data on editing and research operation suggestions, and generate feedback feature tags by combining user historical editing and research behavior; Based on feedback feature labels and the archive database, an adjustment suggestion data package is generated; During the user's modification operation, the location, order, and deleted / added content are tracked in real time, and guidance prompts are generated. A pre-evaluation is conducted on the edited content modified by the user. By comparing the completeness of information and the degree of matching with user needs before and after the modification, a pre-evaluation report is generated. Once the user confirms the changes, final verification will be performed.

[0008] Preferably, step S102 specifically includes the following steps: Collect user input habit data from past editing operations and context information of the current editing operation, and establish a correlation mapping between the two; The system evaluates the user's current input content based on the association mapping results. If the input is a core element of the editing operation, it is determined to be accurate input, and a prompt word fragment is generated. If the input only mentions the research direction, it is judged as fuzzy input, and a guiding prompt word fragment is generated; The initially generated prompt word fragments are processed to extract information related to the compilation and research objectives, remove redundant expressions that are irrelevant to the compilation and research, and strengthen the expression weight of the core theme of the compilation and research in the prompt words; The intent information in the user's editing and research operation is broken down. If the user input includes editing and research requirements, it is broken down into two sub-intents: file selection and summary generation. Then, based on the editing and research topic association, the editing and research extension content related to the sub-intents is supplemented to enrich the coverage of the prompt words. The simulation process of generative AI responding to the current prompt word is used to check whether the generated output meets the requirements of the editing and research task. If it does not meet the requirements, the prompt word content is adjusted, and the simulation and adjustment process is repeated until the prompt word can guide the AI ​​to generate output that meets the editing and research needs.

[0009] Preferably, step S103 specifically includes the following steps: After receiving the AI ​​prompt words generated in step S102, the generative AI identifies the editing and research scene attributes in the prompt words; Based on the defined output dimensions, the generative AI initially generates archival data query instructions or topic-related keywords; it then retrieves the structural feature information of the archival database in the archival system and matches the initially generated query instruction fields and keyword descriptions with the database structural features. If the compilation and research operation involves multiple types of archives, the generative AI will adjust the field settings of the query command and the expression of keywords according to the information recording characteristics of different types of archives; In the generative AI comparison step S102, the explicit research requirements in the prompt words are checked to see if the currently generated query command contains constraints to achieve the requirements and whether the topic-related keywords cover the requirements information. If there are any omissions, the corresponding content is added. Generative AI acquires information about the preset query methods in the archival system, adjusts the format of the query instructions and the word segmentation rules of the keywords according to the query methods, and generates the final archival data query instructions or topic-related keywords.

[0010] Preferably, step S105 specifically includes the following steps: Retrieve the compilation and research scenario information recorded in steps S101-S104, and match it with the scenario-based desensitization rule set preset by the archive system. The rule set contains the types of sensitive information that need to be desensitized under different scenarios. Based on the target archive data obtained in step S104, locate the carriers of sensitive information in various types of archives; The target file data was desensitized according to the scenario-based desensitization rules; Based on the compilation and research requirements in step S101, generate prompt words for abstract generation; the prompt words clearly indicate the key compilation and research information that needs to be retained, and limit the expression style of the abstract; Check whether the anonymized target archive data contains the elements that support the generation of the research summary; if elements are missing due to anonymization, adjust the anonymization method to ensure that the anonymized data still meets the information requirements for summary generation.

[0011] Preferably, step S106 specifically includes the following steps: After receiving the summary prompts generated in step S105, the generative AI extracts the compilation and research scenarios and target file data types associated with the prompts and matches them with the preset summary module framework. For the target archive data obtained in step S104, information from different types of archives is merged; For the desensitized identifiers in step S105, supplement the semantic descriptions in conjunction with the file context; Retrieve users' historical preference data for compiling and studying abstracts to adapt to users' personalized information presentation needs; Compare the current summary with the initial requirements of the user's compilation and research operation in step S101, and check whether the current summary fully covers the required dimensions to ensure that the summary is consistent with the initial compilation and research goals.

[0012] Preferably, step S107 specifically includes the following steps: Establish a semantic mapping relationship between the anonymized tags of research summaries and the original archival data; Based on the needs of the editing and research scenario, hierarchical information restoration and semantic reconstruction are performed; Generate multi-dimensional editing and research operation suggestions and establish operation-related paths.

[0013] This application also provides an electronic archives-assisted compilation and research system based on generative AI, the system comprising: The compilation and research operation input module is used to obtain the topic, summary information or selected archival data input by the user, define the compilation and research operations of the compilation and research task, and provide the execution basis for subsequent prompt word generation, archival query and de-identification processing; The prompt generation module is used to generate corresponding AI prompt words based on the compilation and research operations; The query generation module is used by the generative AI to generate archive data query instructions or topic-related keywords based on the AI ​​prompts. The archival data retrieval module is used to query the archival database according to the query command or topic-related keywords to obtain target archival data; The data anonymization and processing module is used to anonymize the target archive data and generate prompt words for summary generation; The research summary generation module generates research summaries based on generative AI according to the prompt words; The compilation and research information restoration module is used to restore the de-identified information in the compilation and research summary and generate compilation and research operation suggestions to present to the user; The editing and optimization suggestion module is used by users to make adjustments and modifications based on the editing and optimization operation suggestions.

[0014] According to another embodiment of this application, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the generative AI-based electronic document-assisted compilation method.

[0015] According to another embodiment of this application, a storage medium is also provided, on which a computer program is stored, which, when executed by a processor, implements the steps of the generative AI-based electronic archive assisted compilation method.

[0016] As can be seen from the above technical solutions, the present invention has the following advantages: This invention clarifies the user's initial compilation and research needs by having the user fill in the title, abstract information, or select archival data. This avoids AI from engaging in untargeted auxiliary behavior due to vague needs, preserves the user's control in the compilation and research process, and ensures that the final result meets the user's actual compilation and research goals.

[0017] This invention analyzes user input habits and operational context to tailor prompts to users' personalized editing habits. It evaluates input content and differentiates between explicit input to generate precise instructions and fuzzy input to generate guiding descriptions, thus covering the input capabilities of different users and improving the usage of prompts. Generative AI, based on prompts, transforms users' natural language needs into structured query instructions recognizable by the database, ensuring a high degree of match between query direction and editing needs.

[0018] Based on targeted query commands / keywords, effective data can be quickly located and extracted from archival databases, reducing the amount of irrelevant data returned and improving data acquisition efficiency and accuracy. Locating the location of sensitive information carriers and filling in semantic gaps can solve the problem of incoherent sentences after anonymization, improving the fluency of compilation. Abstract prompts are generated based on compilation needs, setting the tone for abstract generation. Core elements after anonymization are checked to ensure compilation quality. The initial abstract covers core information, ensuring no basic compilation elements are omitted; semantic analysis optimizes sentence fluency and logic, improving readability; style is adjusted based on historical preferences to meet personalized needs. Secondary verification of content accuracy avoids discrepancies between the abstract and original archival information, improving the credibility of the compilation results. Restoring anonymized information allows users to see complete data, facilitating judgment on whether the anonymization scope needs adjustment or additional information needs to be added. The user's editing methods for the compilation results are retained; users can adjust them independently based on operation suggestions to ensure the practicality of the compilation results. Attached Figure Description

[0019] To more clearly illustrate the technical solution of the present invention, the accompanying drawings used in the description will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 Example diagram of an electronic archives compilation and research system based on generative AI; Figure 2 A flowchart of an electronic archives-assisted compilation method based on generative AI; Figure 3 This is a schematic diagram of an embodiment of an electronic archives-assisted compilation and research system based on generative AI; Figure 4 This is a schematic diagram of an electronic device. Detailed Implementation

[0021] The generative AI-based electronic archives compilation and research method involved in this application utilizes the intent understanding, instruction generation, and text summarization capabilities of generative AI to achieve intelligent suggestions and verification for electronic archives compilation and research operations.

[0022] like Figure 1 As shown, the system involved in this invention involves the interaction of three roles: user, archival system, and generative AI. The user initiates and verifies the compilation and research operations, which are divided into two types: filling in titles and abstract information, and selecting archives from the archive database. In addition to basic operations such as retrieving and compiling archival information, the archival system is also responsible for interacting with the generative AI, including generating AI prompts based on the user's compilation and research operation instructions and preset templates, desensitizing the prompts, restoring the results generated by the AI, and presenting compilation and research operation suggestions. The generative AI, as an external module, is connected to the archival system, understands and analyzes the questions raised by the archival system, and provides corresponding answers, including generating database query instructions and summarizing the content of documents given in the abstract.

[0023] The following describes in detail the generative AI-based electronic record compilation method of this application. Specific details, such as particular system structures and technologies, are presented for illustrative purposes and not for limitation, to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application can also be implemented in other embodiments without these specific details.

[0024] It should be understood that, when used in this specification, the term "comprising" indicates the presence of the described feature, integral, step, operation, element, and / or component, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components, and / or collections thereof. The terms "comprising," "including," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0025] The terms "one embodiment" or "some embodiments" used in this application mean that one or more embodiments of this application include the specific features, structures, or characteristics described in that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this application do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized.

[0026] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] Please see Figure 2 The diagram shows a flowchart of a generative AI-based electronic archives compilation method in a specific embodiment. The method includes: S101: Obtain the user-inputted topic, summary information, or selected archival data, define the compilation and research operations for the compilation and research task, and provide the execution basis for subsequent prompt word generation, archival retrieval, and desensitization processing.

[0028] In some embodiments, when users perform compilation and research operations, when filling in the title around specific compilation and research objectives, they can clarify the scope of the topic and its association with the archive category, and select the system's preset title template; when filling in the summary information, they can initially extract the core direction and key information dimensions of the compilation and research, and mark the information types to be presented first; when selecting archive data, they can accurately locate it through multi-dimensional filtering conditions, support batch selection or selection after previewing a single item, and view the core fields of the archive during preview to help determine whether it meets the compilation and research requirements, and allow users to mark statuses such as priority use or pending confirmation.

[0029] S102: Generate corresponding AI prompt words based on the compilation and research operation.

[0030] In some embodiments, the archival system generates AI prompts based on the user's operation type in S101: if the user only fills in the title, the archival system extracts elements from the title and supplements them to meet the requirements of the compilation and research scenario.

[0031] If a user fills in both title and summary information, first verify their consistency, then expand the prompt words based on the summary information; if a user selects archive data, read the common attributes of the selected archives and specify the relevant information in the prompt words.

[0032] Optionally, this embodiment can generate a rule base based on natural language processing (NLP) technology and prompt words in the field of archives, parse the text information and data attributes of user operations in S101, and then call the adapted compilation scenario template in the rule base to integrate the scattered operation information, ensuring that the prompt words are logically coherent and the instructions are clear, so that they can be accurately understood by generative AI.

[0033] S103: Generative AI generates file data query instructions or topic-related keywords based on the AI ​​prompts.

[0034] It should be noted that generative AI is implemented based on OpenAI's GPT series (GPT-3.5 / GPT-4), Google's PaLM 2, or T5 (Text-to-Text Transfer Transformer) or BART models.

[0035] In some embodiments, generative AI determines the demand type corresponding to the prompt words and outputs an adaptation result. In some specific embodiments, step S103 specifically includes the following steps: S1031: After receiving the AI ​​prompt words generated in step S102, the generative AI identifies the editing and research scene attributes in the prompt words; The compilation and research scenario attributes include the reporting and display of compilation and research results, historical archiving, and domain characteristics of compilation and research topics, and determine the output dimensions that match the scenario attributes.

[0036] In this embodiment, the identification of research and development scenario attributes extracts key information from the description of prompt words. For example, if the prompt word includes "for annual report," the intended use is determined to be a report presentation; if it includes "completion of R&D project," the domain characteristic is determined to be a scientific research project. The determination of the output dimension in this embodiment needs to correspond to the actual needs of the scenario attributes. By associating scenario attributes with output dimensions, subsequent file queries can be more closely aligned with the research and development goals, avoiding the acquisition of irrelevant information.

[0037] S1032: Based on the defined output dimension, the generative AI initially generates archival data query instructions or topic-related keywords; it retrieves the structural feature information of the archival database in the archival system and matches the initially generated query instruction fields and keyword descriptions with the database structural features.

[0038] The retrieval of database structure features in this embodiment requires specifying the specific field names and field value specifications of the archive database; during the matching process, it is checked whether the fields of the query command exist in the database and whether the keyword descriptions conform to the field value specifications.

[0039] S1033: If the compilation and research operation involves multiple types of archives, the generative AI will adjust the field settings of the query command and the expression of keywords according to the information recording characteristics of different types of archives.

[0040] Optionally, the query instructions for document archives can be adjusted to include text keyword fields, and the keywords for audio-visual archives can be adjusted to cover the subject of the shooting and the description of the storage format.

[0041] It should be noted that the information recording characteristics of different types of archives differ significantly. Therefore, this embodiment designs dedicated fields and keywords for core information during the adjustment process. For example, the document archive query command is supplemented with a keyword field to locate the text content. Through differentiated adjustments, queries for each type of archive can focus on its key information.

[0042] S1034: Generative AI comparison steps S102 prompts clearly define the compilation requirements. Check whether the currently generated query command contains the necessary constraints to achieve the requirements and whether the topic-related keywords cover the requirements information. If there are any missing items, supplement the corresponding content.

[0043] It should be noted that the comparison method for the compilation and research requirements can be to check the key requirements in the prompts one by one. During the check, confirm whether the query command contains the corresponding constraints and whether the keywords cover the requirements. If anything is missing, add it. Through targeted verification and supplementation, it can be ensured that the output content fully serves the compilation and research requirements, reducing the workload of subsequent data screening.

[0044] S1035: Generative AI acquires the preset query method information of the archive system, adjusts the format specifications of the query command and the word segmentation rules of the keywords according to the query method, and generates the final archive data query command or topic-related keywords.

[0045] It should be noted that this embodiment can clearly define the specific query type of step S104; the format adjustment conforms to the database query syntax, and the word segmentation rules match the processing logic of the full-text retrieval system.

[0046] Optionally, the specific query type in step S104 can be specified such that the archive database only supports exact match queries, in which case the query command cannot use fuzzy match symbols; or it supports full-text search queries, in which case keywords need to be processed according to the system's word segmentation rules. Different query methods in this embodiment have different requirements for command format and keyword expression. If they are not compatible, it will lead to query failure or result deviation. By adapting to subsequent query methods in advance, it can be ensured that step S104 can directly use the generated query command and guarantee the continuity of the compilation and research process.

[0047] S104: Query the archive database according to the query instruction or topic-related keywords to obtain the target archive data.

[0048] In some embodiments, if the query is based on a query command, the command validity check is performed first.

[0049] For example, the system checks whether the query contains illegal fields, whether the user has permission to access sensitive fields, and whether the filter conditions conform to the database logic. If the verification passes, the system calls the file database interface to execute the query. If the query result is empty, the system prompts the user to adjust the filter conditions and recommends similar files.

[0050] If the query is based on a topic keyword, a multi-field matching method is used, supporting fuzzy matching. Optionally, R&D investment can be matched with R&D funding or investment amount. Query results are sorted by the number of keyword matches and the weight of the matching fields. Each file is marked with matching keywords, and a file preview function is provided, allowing users to choose whether to download the complete file based on the preview.

[0051] This embodiment targets key keywords and uses a full-text search engine to quickly locate files containing those keywords. It then uses the TF-IDF algorithm to calculate the matching degree between the files and the keywords, and sorts the results. At the same time, it uses preview caching technology to store the core information of the files, ensuring that the preview function responds quickly.

[0052] S105: De-identify the target file data and generate prompt words for summary generation.

[0053] In some embodiments, sensitive information identification can be based on a rule base and text recognition methods to ensure the location of sensitive information. This embodiment can also call corresponding algorithms based on data type and compilation scenario, and ensure subsequent recovery by encrypting and storing the original data; the summary prompt word generation uses semantic analysis technology to determine the information integrity of the de-identified data and adjusts the direction of the prompt word instructions to ensure that the prompt words are adapted to the state of the de-identified data.

[0054] Step S105 in this embodiment specifically includes the following steps: S1051: Retrieve the compilation and research scenario information recorded in steps S101-S104. The compilation and research scenario information may involve the compilation and research results being used for cross-departmental collaboration, internal archiving, and public reporting. Match the scenario-based desensitization rule set preset by the archive system. The rule set contains the types of sensitive information that need to be desensitized under different scenarios.

[0055] In this embodiment, the information on the research and development scenario is determined from the prompt words in step S102. Combined with the scenario-based desensitization method, the priority of sensitive information in different scenarios is clarified. By matching with the scenario, desensitization can be achieved on demand.

[0056] S1052: Based on the target archive data obtained in step S104, locate the carriers of sensitive information in various types of archives; Specifically, this can involve document archives focusing on names, unit addresses, and classified data in the main text paragraphs, and audio-visual archives focusing on sensitive markings and key technical parameter markings in file names, video subtitles, and audio narration.

[0057] It should be noted that the carriers of sensitive information in different types of archives are fundamentally different. Sensitive information in written archives is carried by text paragraphs, while sensitive information in audio-visual archives is carried by audio-visual supplementary information. When locating sensitive information, it is necessary to adopt an identification method based on the characteristics of the carrier. Sensitive words are retrieved for written archives, while metadata and subtitle text are parsed for audio-visual archives.

[0058] S1053: Perform desensitization processing on the target file data according to the scenario-based desensitization rules. For example, replace the name with the corresponding role name, mask the trade secret data as 'confidential business data', and at the same time, complete and annotate the semantically broken information after desensitization to ensure that the desensitized content retains the contextual relevance.

[0059] The anonymization process in this embodiment uses a method of associating identifiers, roles, and scenarios, and is deduced based on the file context. This ensures that the generative AI can understand the actual meaning of the anonymized information and avoids distortion of the summary content.

[0060] S1054: Based on the compilation and research operation requirements in step S101, generate prompt words for abstract generation; the prompt words clearly indicate the key compilation and research information that needs to be retained, and limit the expression style of the abstract.

[0061] For example, the key information to be compiled can be the technological breakthroughs of the R&D project in 2024 that need to be highlighted, and the anonymized role names correspond to their responsibilities in the project.

[0062] S1055: Check whether the de-identified target archive data contains the elements that support the generation of the research summary; if the elements are missing due to de-identification, adjust the de-identification method to ensure that the data after de-identification can still meet the information requirements for the generation of the summary.

[0063] The elements checked here are compared against the research and compilation requirements list. If any are missing after de-identification, the de-identification granularity is adjusted. Through verification and adjustment, it is ensured that the de-identified data can still support the generation of research and compilation summaries, balancing security and practicality.

[0064] S106: Generative AI generates a compiled summary based on the prompts. In some embodiments, generative AI parses the main instructions in the prompts, which may involve extracting non-sensitive information, annotating the meaning of placeholders, ensuring logical coherence, and integrating them with the requirements of the research and development scenario to generate a summary from the target data.

[0065] For example, the prompt word points to the sales contract file, such as the signing time, partner, performance status, and labeled as contract amount. The AI ​​will integrate the data in chronological order. For example, from August to December 2024, a total of 45 sales contracts were obtained. Among them, a contract was signed with Company A in August, with the performance status being "completed" and the contract amount being "contract amount"; a contract was signed with Company B in September, with the performance status being "in progress" and the contract amount being "contract amount" being "contract amount" and so on.

[0066] This embodiment also verifies data consistency; if the prompt word points to the R&D project file, the generative AI organizes the content according to the logic of project initiation - technology R&D - verification results.

[0067] It can be seen that generative AI is based on a pre-trained model specifically for the field of archival compilation and research. It extracts information, format requirements, and logical requirements from prompts through instruction parsing, integrates key information from target data, and finally organizes language according to compilation and research standards through natural language.

[0068] S107: Restore the de-identified information in the compilation summary and generate compilation operation suggestions to present to the user.

[0069] In some embodiments, the de-identified information restoration process first performs user permission verification. After permission is granted, the de-identified record stored in S105 is called to extract the corresponding sensitive information from the encrypted original file data, and the placeholders in the digest are replaced with the actual data. A restoration operation record is generated for subsequent auditing.

[0070] In this way, the original data is encrypted and stored at the address associated with the de-identified record of S105. After decryption, sensitive information is extracted and restored, and operation logs are recorded to ensure traceability. The compilation operation suggestions are generated based on the identification of summary defects. The user habit matching method is to identify information gaps and logical loopholes in the summary, and then combine them with the user's historical compilation operation data to match the corresponding optimization suggestions from the suggestion rule base, and present them after being sorted by priority.

[0071] S108: The user makes adjustments and modifications based on the aforementioned editing and research operation suggestions.

[0072] In some embodiments, after viewing the editing and research operation suggestions, users can directly execute the suggestions, and the system will automatically jump to S103 to generate corresponding AI prompts and query instructions based on the suggestions, without requiring users to perform any operations again.

[0073] This embodiment also allows for customized adjustments. Combined adjustments are also possible, with some implementation suggestions and some custom modifications. During the adjustment process, a real-time preview function is provided, supporting undo / redo operations. If the user deems the abstract unnecessary to adjust, the system automatically archives the compilation abstract and operation records to the compilation results library, generating a compilation report.

[0074] In one embodiment of the present invention, based on step S108, the following will provide a possible embodiment and describe its specific implementation in a non-limiting manner. Step S108 specifically includes the following steps: S1081: Collect user feedback data on editing and research operation suggestions, and generate feedback feature tags by combining user's historical editing and research behavior.

[0075] In this embodiment, user feedback includes not only textual annotations of satisfaction / dissatisfaction, but also data on the time spent modifying a suggestion, the number of times it was repeatedly modified, and suggestions regarding anonymization and restoration. Combined with user historical behavior, such as the user focusing on modifying the acceptance conclusion in the past three compilation and research phases, feedback feature tags are generated as follows: difficulty in modification within a time frame, neglecting anonymization and restoration, and focusing on the acceptance conclusion.

[0076] S1082: Generate adjustment suggestion data package based on feedback feature labels and archive database.

[0077] In this embodiment, when the user's current feedback tag is "difficulty in modifying the time range," the system matches successful cases of time range modification in the case library, generates an adjustment suggestion data package, and, referring to the project cycle field of the associated file, suggests adjusting the time range to XXXX year XX month to match the acceptance node. This step, through matching user characteristics with historical cases, transforms the suggestion from a general template to personalized guidance.

[0078] S1083: During the user's modification operation, the location, order, and deleted / added content are tracked in real time, and guidance prompts are generated.

[0079] In this embodiment, user modification paths are recorded through front-end event tracking. For example, if a user first modifies the acceptance conclusion, then skips the rectification suggestions and adjusts the time range from XXXX year XX month to XXXX year Y month, the system recognizes that the user repeatedly adjusts the time range without referring to the associated file and generates a dynamic guidance prompt: "We have detected that you have modified the time range multiple times. The associated file, the project initiation notice, clearly states that the project period is XXXX year Y month. We suggest directly using this time range to improve accuracy."

[0080] S1084: Conduct a pre-evaluation of the edited content modified by the user, and generate a pre-evaluation report by comparing the information completeness and user needs matching degree before and after the modification.

[0081] It should be noted that information completeness refers to the proportion of core elements covered in the revised content, such as the acceptance conclusion, rectification items, and time range, which increased from 70% to 90% before the revision. User demand matching degree is a comparison between the revised content and users' historical preferences.

[0082] For example, the pre-assessment report shows that the information completeness meets the standard, the user requirement matching degree is improved by 20%, and there are no abnormalities in format compliance. This step helps users intuitively understand the effect of the modification through quantitative assessment.

[0083] S1085: After the user confirms the modification, the final verification is performed. The verification method involves checking the accuracy of the de-identified information restoration, the consistency between the summary and the original file, and the relevance between the suggestions and the modified content. A verification report is generated and marked as passed or requiring supplementary adjustments. If passed, the compilation process ends. If supplementary adjustments are required, return to S102 to regenerate suggestions.

[0084] The final verification includes: **Anonymized Content Restoration Verification:** Verifying that the modified content accurately restores the original information and checking the consistency of the restored information with the context. **Summary Consistency Verification:** Comparing the modified summary with the original anonymized file to ensure that no core elements are missing. **Suggestion Relevance Verification:** Checking whether the suggestions adopted by the user correspond to the modified content.

[0085] In one embodiment of the present invention, based on step S102, the following is a possible embodiment and its specific implementation will be described in a non-limiting manner. Step S102 specifically includes the following steps: S1021: Collect user input habit data from past editing operations and context information of the current editing operation, and establish a correlation mapping between the two.

[0086] Optionally, input habit data may include title word preferences, abstract style, and priority of screening dimensions when selecting archives.

[0087] It's important to note that this data collection process considers factors such as whether users frequently use qualifiers in titles during past compilation and research, whether abstracts tend to be concise or detailed, and whether they prioritize the organization or date of creation when selecting archives. Simply relying on generic prompt generation logic won't align with user habits. Combining past compilation and research habits with the current context allows for more tailored prompt generation to the user's specific compilation and research scenario.

[0088] S1022: Evaluate the user's current input content based on the association mapping results. If the core elements of the editing operation are input, it is determined to be accurate input, and an instruction prompt word fragment is generated. If the input only mentions the research direction, it is judged as fuzzy input, and a guiding prompt word fragment is generated.

[0089] It should be noted that the input judgment criteria focus on elements unique to the editing and research process, while the judgment criteria for fuzzy input is the lack of unique elements. The guiding prompt word fragments need to be supplemented with the necessary elements for the editing and research process in a targeted manner; this can reduce rework in subsequent editing and research operations and improve efficiency.

[0090] S1023: Process the initially generated prompt word fragments, extract information related to the compilation and research objectives, eliminate redundant expressions unrelated to compilation and research, and strengthen the expression weight of the core theme of compilation and research in the prompt words.

[0091] Optionally, extract information related to the research objectives. When the research results are used for reporting, the core conclusions should be highlighted; when used for archiving, the integrity of the archives should be emphasized.

[0092] It should be noted that semantic analysis first identifies the research and development target, then selectively retains expressions related to the target and removes irrelevant information mentioned by chance in the user input, which allows the AI ​​to more clearly grasp the core direction of the research and development.

[0093] S1024: Deconstruct the intent information in the user's compilation and research operation. If the user input includes compilation and research requirements, it is deconstructed into two sub-intents: document selection and summary generation. Then, based on the compilation and research topic association, supplement the compilation and research extension content related to the sub-intents and enrich the coverage of prompt words.

[0094] It should be noted that when breaking down complex intentions, it is necessary to distinguish between selection, summarization, and integration in the compilation and research process. For example, when a user selects and summarizes research and development files, it is broken down into two sub-intentions: selection and summarization. Thematic association expansion should be based on the selection of research and development files in the compilation and research process, associating them with supporting files that may be needed in the compilation and research process, such as approval records and acceptance materials, rather than unrelated themes. This ensures that the AI ​​understands the user's complete compilation and research intention, rather than the needs of a single stage.

[0095] S1025: Simulate the response process of generative AI to the current prompt word, and check whether the generated output meets the requirements of the editing and research task; if not, adjust the prompt word content, repeat the simulation response and adjustment process until the prompt word can guide the AI ​​to generate output that meets the editing and research requirements.

[0096] It should be noted that when simulating the AI ​​response, the query command should include the archival attributes to be compiled and studied. Keywords should cover the compilation and study topic; if missing, constraints should be added accordingly, such as requiring the query command to include the archival originating unit attribute. By simulating the response in advance and verifying it against compilation and study task standards, deficiencies in the prompts can be corrected before the official generation, improving the accuracy of the compilation and study.

[0097] In one embodiment of the present invention, based on step S106, the following is a possible embodiment and its specific implementation will be described in a non-limiting manner. Step S106 specifically includes the following steps: S1061: After receiving the summary prompts generated in step S105, the generative AI extracts the compilation scenario and target file data type associated with the prompts and matches them with the preset summary module framework.

[0098] It should be noted that the information sources of each module can be clearly defined based on the scenario and file type as the framework for the summary module. For example, the associated file module for the archival scenario may contain information from the metadata of the associated file of the target file. This is because existing technologies for generating summaries often use general modules that do not consider the differentiated information organization requirements of different scenarios, nor do they allocate module content according to the characteristics of file types. By customizing the framework, the summary information organization can be made more in line with the purpose of the scenario and the advantages of the file type.

[0099] S1062: For the target file data obtained in step S104, merge the information of different types of files.

[0100] It should be noted that the integration of archival information can combine complementary methods. For example, written archives are good at recording logical conclusions, while audio-visual archives are good at presenting scene details. Information duplication should be avoided during integration. If the technical standards of written archives are clear, the same standard descriptions in audio-visual archives will not be repeated.

[0101] It can be seen that processing different types of archives separately can easily lead to problems such as textual conclusions being disconnected from scene details and technical parameters being separated from actual applications. By complementing and integrating these different types of archives, the information advantages of various archives can be combined to form logically coherent and complete summaries, thereby enhancing the information value.

[0102] S1063: Supplement the semantic description of the identifier after the desensitization process in step S105, taking into account the file context.

[0103] It should be noted that the semantic supplementation of the anonymized label is based on the information in the archive. The supplementary content does not disclose sensitive information and ensures the interpretability of the anonymized label while guaranteeing information security.

[0104] S1064: Retrieve historical preference data of users' past compilation and research summaries to adapt to users' personalized information presentation needs.

[0105] It's important to note that user historical preference data focuses on information presentation. If a user has repeatedly placed the results data at the beginning of paragraphs in historical summaries, the current summary will also adopt this structure. If a user previously entered process details chronologically, the current summary will also present the process in chronological order. By adapting the information organization method, user acceptance and efficiency in using the summary can be improved.

[0106] S1065: Compare the current summary with the initial requirements of the user's compilation operation in step S101, and check whether the current summary fully covers the required dimensions to ensure that the summary is consistent with the initial compilation goal.

[0107] It should be noted that the compilation and research operations performed by S101 users, such as filling in titles, abstract information, and selecting archival data, reveal clear objectives from user input. If users supplement their abstract information, the objectives become even clearer. Based on these compilation and research operations, potential objectives can be deduced. For example, if users prioritize selecting 2024 subsidiary activity summary reports when selecting archives, the implicit requirement is to cover all subsidiaries and avoid focusing solely on the company level. If users emphasize major events when filling in titles, the implicit requirement is to filter core monthly activities and exclude routine subsidiary meeting notices.

[0108] This embodiment breaks down the extracted initial requirements into core and secondary dimensions, with clear verification standards for each dimension to avoid ambiguous judgments. The current research abstract is then checked point-to-point against the decomposed dimension list to avoid holistic judgment. Any incompleteness discovered during the verification process is categorized as either traceable or non-traceable supplementation.

[0109] Traceable supplementation is based on the assumption that missing information can be extracted from the target archive data obtained in S104, and is directly added to the summary. For example, if "no activity records in August 2024" are found, the company's internal archives for August 2024 can be retrieved from the target archives in S104, and the following can be extracted: "Held a special meeting on company business integration, covering 15 subsidiaries, promoting 3 business collaborations," and added to the corresponding position in the summary. Non-traceable supplementation is based on the assumption that there is no corresponding information for the missing dimension in the target archive. In this case, the missing item is accurately marked in the summary. At the same time, a prompt next to the mark indicates that users can return to S104 to extend the query to supplement the data in the sign-in table archive in March 2024, providing guidance for subsequent operations.

[0110] In one embodiment of the present invention, based on step S107, the following is a possible embodiment and its specific implementation will be described in a non-limiting manner. Step S107 specifically includes the following steps: S1071: Establish a semantic mapping relationship between the anonymized tags of the compilation and research abstracts and the original archival data.

[0111] In this embodiment, when generating the compilation summary, the de-identification tags, original information, and related data relationships are constructed. The de-identification tag can be defined as assigning a unique code to each de-identification identifier in the summary. The coding rules include the compilation type, time, and sequence number to avoid confusion between different tags. The original information reflects the specific fields and content of the original file that each de-identification tag needs to be mapped to. The related data relationships record the storage path of the original file containing the data, the de-identification rule number, the de-identification operation time, and the operator, ensuring traceability of the relationships.

[0112] S1072: Perform hierarchical information restoration and semantic reconstruction based on the needs of the compilation and research scenario.

[0113] In this embodiment, different restoration methods are executed according to the information sensitivity requirements of the research and development scenario to ensure the logical coherence of the restored statements.

[0114] Specifically, the scenario determination is based on the editing and research operation information in S101-S104, eliminating the need for manual selection by the user. This embodiment can predefine the correspondence between scenarios and information types, with each correspondence including information type, desensitization and restoration method, and desensitization retention requirements.

[0115] Text context analysis can also be used to extract key information before and after the de-identification marking in the original file, integrate the restored information with these contexts, and generate logically coherent complete sentences. This ensures that the semantic analysis and generated information are logical and meet the information usage needs of different scenarios, thereby improving reading coherence.

[0116] S1073: Generate multi-dimensional editing operation suggestions and establish operation association paths.

[0117] It should be noted that each suggestion in this embodiment is linked to a directly accessible entry point. Through text semantic analysis, it scans for discrepancies between the summary content and the compilation and research specifications and initial requirements, identifying issues such as missing information, formatting problems, and superficial content, and pinpointing the specific location of these issues within the summary. This embodiment can invoke preset R&D compilation and research issues and corresponding solutions from the archive database. The archive database stores the mapping processing methods corresponding to each type of R&D compilation and research issue; for example, chronological disorder is mapped to adjusting the structure according to the timeline, and data lacking units is mapped to supplementing compliant units, ensuring the relevance of the suggestions.

[0118] The operation path binding method in this embodiment establishes a mapping relationship between the solution and the preceding steps. Through the system's internal process interface, suggestions are bound to the corresponding step's operation interface and parameter configuration. This operation-associated path eliminates the need for users to search for preceding steps, re-enter parameters, and resume the compilation / research progress process.

[0119] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0120] The following are embodiments of the generative AI-based electronic archives assisted compilation system provided in this disclosure. This system and the generative AI-based electronic archives assisted compilation methods in the above embodiments belong to the same inventive concept. For details not described in detail in the embodiments of the generative AI-based electronic archives assisted compilation system, please refer to the embodiments of the generative AI-based electronic archives assisted compilation methods described above.

[0121] like Figure 3 As shown, the system includes: The compilation and research operation input module 201 is used to obtain the topic, summary information or selected archival data input by the user, define the compilation and research operation of the compilation and research task, and provide the execution basis for subsequent prompt word generation, archival query and de-identification processing; The prompt generation module 202 is used to generate corresponding AI prompt words based on the compilation and research operation; The query generation module 203 is used for the generative AI to generate file data query instructions or topic-related keywords based on the AI ​​prompts. The archive data retrieval module 204 is used to query the archive database according to the query instruction or topic-related keywords to obtain target archive data; The data desensitization and processing module 205 is used to desensitize the target file data and generate prompt words for summary generation; The research summary generation module 206 generates a research summary based on the prompt words using generative AI. The compilation information restoration module 207 is used to restore the desensitized information in the compilation summary and generate compilation operation suggestions to present to the user; The editing and optimization suggestion module 208 is used by users to make adjustments and modifications based on the editing and optimization operation suggestions.

[0122] like Figure 4 As shown, this application also provides an electronic device, including a display module 103, a memory 102, a processor 101, and a computer program stored in the memory and executable on the processor 101. When the processor 101 executes the program, it implements the steps of an electronic document-assisted compilation method based on generative AI.

[0123] In embodiments of the present invention, electronic devices include, but are not limited to, laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic devices 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 examples and are not intended to limit the implementation of the embodiments described and / or claimed herein.

[0124] In this embodiment, processor 101 may be implemented using at least one of an application-specific integrated circuit, a programmable logic device, a field-programmable gate array, a processor, a controller, a microcontroller, a microprocessor, or an electronic unit designed to perform the functions described herein. In some cases, such an implementation may be implemented within a controller. For software implementation, implementations such as processes or functions may be implemented with separate software modules that allow the performance of at least one function or operation. Software code may be implemented by a software application (or program) written in any suitable programming language, and the software code may be stored in memory and executed by the controller.

[0125] The display module 103 is used to display information input by the user or information provided to the user. The display module 103 may include a display panel, which may be configured in the form of a liquid crystal display, an organic light-emitting diode, or the like.

[0126] The memory 102 can be used to store software programs and various data. The memory 102 may include high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0127] This application also provides a storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the generative AI-based electronic archive assisted compilation method.

[0128] The storage medium may be any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example,, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples (a non-exhaustive list) of readable storage media include: electrical connections having one or more wires, portable disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0129] In a storage medium, a readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying readable program code. This propagated data signal may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A readable signal medium may also be any readable medium other than a readable storage medium, capable of sending, propagating, or transmitting a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0130] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for assisting in the compilation and research of electronic archives based on generative AI, characterized in that, The methods include: S101: Obtain the user-inputted topic, summary information, or selected archival data, define the compilation and research operations for the compilation and research task, and provide the execution basis for subsequent prompt word generation, archival retrieval, and desensitization processing; S102: Generate corresponding AI prompt words based on the above editing and research operations; S103: Generative AI generates file data query instructions or topic-related keywords based on the AI ​​prompts; S104: Query the archive database according to the query instruction or topic-related keywords to obtain the target archive data; S105: De-identify the target archive data and generate prompt words for summary generation; S106: Generative AI generates a research summary based on the prompt words; S107: Restore the de-identified information in the compilation summary and generate compilation operation suggestions to present to the user; S108: The user makes adjustments and modifications based on the aforementioned editing and research operation suggestions.

2. The method for assisting in the compilation and research of electronic archives based on generative AI according to claim 1, characterized in that, Step S108 specifically includes the following steps: Collect user feedback data on editing and research operation suggestions, and generate feedback feature tags by combining user historical editing and research behavior; Based on feedback feature labels and the archive database, an adjustment suggestion data package is generated; During the user's modification operation, the location, order, and deleted / added content are tracked in real time, and guidance prompts are generated. A pre-evaluation is conducted on the edited content modified by the user. By comparing the completeness of information and the degree of matching with user needs before and after the modification, a pre-evaluation report is generated. Once the user confirms the changes, final verification will be performed.

3. The method for electronic archives-assisted compilation and research based on generative AI according to claim 1, characterized in that, Step S102 specifically includes the following steps: Collect user input habit data from past editing operations and context information of the current editing operation, and establish a correlation mapping between the two; The system evaluates the user's current input content based on the association mapping results. If the input is a core element of the editing operation, it is determined to be accurate input, and a prompt word fragment is generated. If the input only mentions the research direction, it is judged as fuzzy input, and a guiding prompt word fragment is generated; The initially generated prompt word fragments are processed to extract information related to the compilation and research objectives, remove redundant expressions that are irrelevant to the compilation and research, and strengthen the expression weight of the core theme of the compilation and research in the prompt words; The intent information in the user's editing and research operation is broken down. If the user input includes editing and research requirements, it is broken down into two sub-intents: file selection and summary generation. Then, based on the editing and research topic association, the editing and research extension content related to the sub-intents is supplemented to enrich the coverage of the prompt words. The simulation process of generative AI responding to the current prompt word is used to check whether the generated output meets the requirements of the editing and research task. If it does not meet the requirements, the prompt word content is adjusted, and the simulation and adjustment process is repeated until the prompt word can guide the AI ​​to generate output that meets the editing and research needs.

4. The method for electronic archives-assisted compilation and research based on generative AI according to claim 1, characterized in that, Step S103 specifically includes the following steps: After receiving the AI ​​prompt words generated in step S102, the generative AI identifies the editing and research scene attributes in the prompt words; Based on the defined output dimensions, the generative AI initially generates archival data query instructions or topic-related keywords; it then retrieves the structural feature information of the archival database within the archival system and matches the initially generated query instruction fields and keyword descriptions with the database structural features. If the compilation and research operation involves multiple types of archives, the generative AI will adjust the field settings of the query command and the expression of keywords according to the information recording characteristics of different types of archives; In the generative AI comparison step S102, the explicit research requirements in the prompt words are checked to see if the currently generated query command contains constraints to achieve the requirements and whether the topic-related keywords cover the requirements information. If there are any omissions, the corresponding content is added. Generative AI acquires information about the preset query methods in the archival system, adjusts the format of the query instructions and the word segmentation rules of the keywords according to the query methods, and generates the final archival data query instructions or topic-related keywords.

5. The method for electronic archives-assisted compilation and research based on generative AI according to claim 1, characterized in that, Step S105 specifically includes the following steps: Retrieve the compilation and research scenario information recorded in steps S101-S104, and match it with the scenario-based desensitization rule set preset by the archive system. The rule set contains the types of sensitive information that need to be desensitized under different scenarios. Based on the target archive data obtained in step S104, locate the carriers of sensitive information in various types of archives; The target file data was desensitized according to the scenario-based desensitization rules; Based on the compilation and research requirements in step S101, generate prompt words for abstract generation; clearly indicate the key compilation and research information that needs to be retained in the prompt words, and limit the expression style of the abstract; Check whether the anonymized target archive data contains elements that support the generation of the research summary; If the desensitization process is missing, adjust the desensitization method to ensure that the desensitized data still meets the information requirements for abstract generation.

6. The method for electronic archives-assisted compilation and research based on generative AI according to claim 1, characterized in that, Step S106 specifically includes the following steps: After receiving the summary prompts generated in step S105, the generative AI extracts the compilation and research scenarios and target file data types associated with the prompts and matches them with the preset summary module framework. For the target archive data obtained in step S104, information from different types of archives is merged; For the desensitized identifiers in step S105, supplement the semantic descriptions in conjunction with the file context; Retrieve users' historical preference data for compiling and studying abstracts to adapt to users' personalized information presentation needs; Compare the current summary with the initial requirements of the user's compilation and research operation in step S101, and check whether the current summary fully covers the required dimensions to ensure that the summary is consistent with the initial compilation and research goals.

7. The method for electronic archives-assisted compilation and research based on generative AI according to claim 1, characterized in that, Step S107 specifically includes the following steps: Establish a semantic mapping relationship between the anonymized tags of research summaries and the original archival data; Based on the needs of the editing and research scenario, hierarchical information restoration and semantic reconstruction are performed; Generate multi-dimensional editing and research operation suggestions and establish operation-related paths.

8. A generative AI-based electronic archives-assisted compilation and research system, characterized in that, The system is used to implement the generative AI-based electronic archives compilation and research method as described in any one of claims 1 to 7; The system includes: The compilation and research operation input module is used to obtain the topic, summary information or selected archival data input by the user, define the compilation and research operations of the compilation and research task, and provide the execution basis for subsequent prompt word generation, archival query and de-identification processing; The prompt generation module is used to generate corresponding AI prompt words based on the compilation and research operations; The query generation module is used by the generative AI to generate archive data query instructions or topic-related keywords based on the AI ​​prompts. The archival data retrieval module is used to query the archival database according to the query command or topic-related keywords to obtain target archival data; The data anonymization and processing module is used to anonymize the target archive data and generate prompt words for summary generation; The research summary generation module generates research summaries based on generative AI according to the prompt words; The compilation and research information restoration module is used to restore the de-identified information in the compilation and research summary and generate compilation and research operation suggestions to present to the user; The editing and optimization suggestion module is used by users to make adjustments and modifications based on the editing and optimization operation suggestions.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the electronic archives-assisted compilation method based on generative AI as described in any one of claims 1 to 7.

10. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the electronic archives-assisted compilation method based on generative AI as described in any one of claims 1 to 7.

Images