Text auditing and modifying method and device, electronic equipment and storage medium

Abstract

The invention relates to a text auditing and modifying method and device, electronic equipment and a storage medium. The method comprises the steps of obtaining a to-be-audited initial text; the initial text is audited in different dimensions through a plurality of audit nodes in the large model, modification suggestions are given, and the modification suggestions are used for referring to the modification suggestions. The text verification is not passed; modifying the initial text according to the modification suggestion through a manuscript modification node in the large model to obtain a modified text; and reviewing the modified text through the plurality of reviewing nodes until no revision suggestion is generated, so as to obtain a modified final text, and no revision suggestion is generated for indicating that the text is reviewed successfully. The accuracy of the text auditing result is improved.

Description

Technical Field

[0001] The present application relates to the field of natural language and artificial intelligence technology, and in particular to a method, device, electronic device and storage medium for text review and modification. Background Art

[0002] In the current information society, article content review has become a crucial link, especially in the fields of news, publishing, social media platforms, etc., where the accuracy, legality and suitability of content need to be strictly reviewed. However, traditional article content review mainly relies on manual review, which may lead to missed or wrong review of content due to factors such as fatigue and subjective bias, and cannot effectively guarantee the accuracy and compliance of content. Summary of the invention

[0003] The present application provides a method, device, electronic device and storage medium for text review and modification to solve the problem of inaccurate text review results.

[0004] In a first aspect, the present application provides a method for text review and modification, the method comprising:

[0005] Get the initial text to be reviewed;

[0006] The initial text is reviewed in different dimensions through multiple review nodes in the large model, and modification suggestions are given, wherein the modification suggestions are used to indicate that the text review has not passed;

[0007] Modifying the initial text according to the modification suggestion through the revision node in the large model to obtain a modified text;

[0008] The modified text is re-reviewed through the multiple review nodes until no modification suggestions are generated, thereby obtaining a modified final text, wherein no modification suggestions are generated to indicate that the text review has been passed.

[0009] Optionally, the initial text is reviewed in different dimensions through multiple review nodes in the large model, and modification suggestions are given including:

[0010] Conducting compliance review on the initial text through the first review node in the large model and providing compliance modification suggestions;

[0011] Performing a logic review on the initial text through the second review node in the large model and providing logic modification suggestions;

[0012] Conducting a professional review of the initial text through the third review node in the large model and providing professional modification suggestions;

[0013] The compliance modification suggestions, the logic modification suggestions and the professional modification suggestions are combined to obtain a final modification suggestion.

[0014] Optionally, performing compliance review on the initial text through the first review node in the large model includes:

[0015] Using a first audit node in the large model, a text anti-spam checking algorithm is used to detect whether the initial text contains spam content;

[0016] Using preset sensitive words to detect whether the initial text contains sensitive words;

[0017] Natural language processing technology is used to detect whether the initial text has any grammatical errors.

[0018] Optionally, performing a logical review on the initial text through the second review node in the large model includes:

[0019] Through the second review node in the large model, natural language processing technology is used to detect whether the semantics of the initial text is accurate, whether the logic of the text content is consistent, and whether the text content structure is reasonable.

[0020] Optionally, performing professional review on the initial text through the third review node in the large model includes:

[0021] Through the third review node in the large model, it is detected whether the initial text is written in the specified language, whether the professional terms are accurate, and whether the quoted content is standardized.

[0022] Optionally, modifying the initial text according to the modification suggestion through the revision node in the large model includes:

[0023] Analyze the overall modification suggestions through the revision nodes in the large model to determine the modification details of each position in the initial text, wherein the modification details include: the audit rules violated, specific problems and corresponding suggestions;

[0024] Analyze the violated audit rules and specific issues according to the preset prompts, and make modifications based on the suggestions.

[0025] Optionally, re-examining the modified text through the multiple review nodes includes:

[0026] Adjusting the review criteria by setting prompt words, wherein the set prompt words are used to set that no additional modification suggestions will be provided when the importance of the error is lower than the importance threshold;

[0027] Re-review the modified text according to the adjusted review standards.

[0028] In a second aspect, the present application provides a device for text review and modification, the device comprising:

[0029] The acquisition module is used to obtain the initial text to be reviewed;

[0030] An audit module, used to audit the initial text in different dimensions through multiple audit nodes in the large model, and give modification suggestions, wherein the modification suggestions are used to indicate that the text has failed the audit;

[0031] A modification module, used to modify the initial text according to the modification suggestion through the revision nodes in the large model to obtain a modified text;

[0032] A generation module is used to re-examine the modified text through the multiple review nodes until no more modification suggestions are generated, thereby obtaining a modified final text, wherein no more modification suggestions are generated to indicate that the text review has passed.

[0033] In a third aspect, the present application provides an electronic device comprising: at least one communication interface; at least one bus connected to the at least one communication interface; at least one processor connected to the at least one bus; and at least one memory connected to the at least one bus.

[0034] In a fourth aspect, the present application also provides a computer storage medium storing computer executable instructions, wherein the computer executable instructions are used to execute the text review and modification method described in any one of the above items of the present application.

[0035] The above technical solution provided by the embodiment of the present application has the following advantages compared with the prior art: by allowing the large model to focus on processing a single task in each auditor role, the error of the model in processing multiple complex tasks can be reduced, thereby improving the performance of the overall audit system. In addition, the iterative audit in this application ensures that the modification suggestions for each round of audit are fully implemented and tested, which can effectively reduce the risk of misreview and missed review in content audit and improve the accuracy of text audit results. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0037] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0038] One or more embodiments are exemplarily described by pictures in the corresponding drawings, and these exemplified descriptions do not constitute limitations on the embodiments. Elements with the same reference numerals in the drawings represent similar elements, and unless otherwise stated, the figures in the drawings do not constitute proportional limitations.

[0039] Figure 1 A schematic diagram of a system for text review and modification provided for an embodiment of the present application;

[0040] Figure 2 A flowchart of a method for text review and modification provided in an embodiment of the present application;

[0041] Figure 3 A schematic diagram of the overall process of text review and modification provided in an embodiment of the present application;

[0042] Figure 4 A schematic diagram of the structure of a text review and modification device provided in an embodiment of the present application;

[0043] Figure 5 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0044] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0045] The disclosure below provides many different embodiments or examples to implement different structures of the present invention. In order to simplify the disclosure of the present invention, the parts and settings of specific examples are described below. Of course, they are only examples, and the purpose is not to limit the present invention. In addition, the present invention can repeat reference numbers and / or letters in different examples. This repetition is for the purpose of simplification and clarity, and does not itself indicate the relationship between the various embodiments and / or settings discussed.

[0046] Optionally, in the embodiment of the present application, the above-mentioned text review and modification method can be applied to Figure 1 In the hardware environment composed of the terminal 101 and the server 103 shown in FIG. Figure 1As shown, the server 103 is connected to the terminal 101 via a network, and can be used to provide services for the terminal or a client installed on the terminal. A database 105 can be set on the server or independently of the server to provide data storage services for the server 103. The above-mentioned network includes but is not limited to: a wide area network, a metropolitan area network or a local area network, and the terminal 101 includes but is not limited to a PC, a mobile phone, a tablet computer, etc.

[0047] A text review and modification method in an embodiment of the present application can be executed by the server 103 or by the terminal 101.

[0048] The following will describe in detail a method for reviewing and modifying text provided by an embodiment of the present application in conjunction with a specific implementation method. Figure 2 As shown, the specific steps are as follows.

[0049] Step 201: Obtaining the initial text to be reviewed;

[0050] Step 202: The initial text is reviewed in different dimensions through multiple review nodes in the large model, and modification suggestions are given, wherein the modification suggestions are used to indicate that the text review has not passed;

[0051] Step 203: modify the initial text according to the modification suggestion through the revision node in the large model to obtain a modified text;

[0052] Step 204: re-examine the modified text through multiple review nodes until no more modification suggestions are generated, thereby obtaining the modified final text, wherein no more modification suggestions are generated to indicate that the text review has passed.

[0053] Users submit article content that needs to be reviewed through the system interface. As the starting point of the review process, this process is usually completed on the front end of the system. Users can upload text files to the system or directly paste text content. After the system receives the article content, it will automatically trigger the subsequent review process.

[0054] Optionally, the user's input can be in plain text, HTML format, or other common document formats (such as Word, PDF, etc.). The system will first pre-process the article and convert it into a standard text format that can be processed by the large model.

[0055] Optionally, this text may include various types of content such as news articles, social media posts, academic papers, etc.

[0056] After the system obtains the initial text, it uses its built-in large-scale machine learning model, which contains multiple nodes that specialize in reviewing different aspects of the text content. These nodes may focus on multiple dimensions such as compliance, logic, and professionalism. Each node operates independently, conducts a comprehensive assessment of the text, and makes specific modification suggestions when problems are found. These suggestions are intended to guide the proofreading nodes on how to improve the text to meet publishing standards or other relevant requirements. If a node does not find any problems, it will not make any modification suggestions, indicating that the review of that dimension has passed.

[0057] Once the revision suggestions from different review nodes are received, the system will automatically apply these suggestions using the revision node in the big model. The revision node is responsible for actually revising the text according to the instructions provided by the review node, which may include correcting grammatical errors, updating outdated factual information, adjusting the writing style to meet specific publication standards, etc. After completing these revisions, the system generates a new version of the text, namely the revised text.

[0058] The revised text needs to be re-evaluated again at all review nodes. This process ensures that all suggested changes have been properly addressed and that the text now meets all preset review criteria. If no new suggested changes are generated during the re-review process, this means that the text has passed all necessary checks and can be considered the final version. However, if there are still suggested changes, the system will continue to repeat steps 202 and 203 until the text reaches a state where no further changes are required.

[0059] This process may involve multiple iterations, with each iteration making subtle adjustments and improvements to the text until all review requirements are met. This cycle ensures that the final published text not only follows language norms and social guidelines, but is also accurate in content and optimized in style.

[0060] This application allows the large model to focus on handling a single task in each reviewer role, which can reduce the model's errors when handling multiple complex tasks, thereby improving the performance of the overall review system. In addition, the iterative review in this application ensures that the modification suggestions for each round of review are fully implemented and tested, which can effectively reduce the risk of misreview and missed review in content review and improve the accuracy of the review results.

[0061] As an optional implementation, the initial text is reviewed in different dimensions through multiple review nodes in the large model, and modification suggestions are given, including:

[0062] Conduct compliance review of the initial text through the first review node in the large model and provide compliance modification suggestions;

[0063] Conduct a logical review of the initial text through the second review node in the large model and provide logical modification suggestions;

[0064] Conduct professional review of the initial text through the third review node in the big model and provide professional modification suggestions;

[0065] Comprehensive compliance modification suggestions, logical modification suggestions and professional modification suggestions to obtain the final modification suggestions.

[0066] This application adopts three different dimensions for text review. The following is a detailed description of the three review dimensions.

[0067] 1. Compliance review

[0068] At this stage, the system uses the first review node in its built-in large-scale machine learning model to perform a preliminary review, focusing on basic issues such as sensitive word filtering, basic legality review, and obvious grammatical errors. This node will carefully check the text content, and once problems are found, it will provide specific compliance modification suggestions to guide the author on how to adjust the text to comply with relevant regulations and standards. The review content of the first review node specifically includes the following:

[0069] Sensitive word filtering: Before review, preset several sensitive words into variables, and submit the initial text and sensitive word variables to the big model at the same time, and the first review node of the big model will identify them. The first review node will use the sensitive word variables to check whether there are sensitive words in the initial text, such as politically sensitive words, pornographic words, violent and terrorist words, etc. If sensitive words are found, specific modification suggestions will be provided, such as replacing sensitive words or deleting related content.

[0070] Basic legality review: Use text anti-spam checking algorithms to detect whether the initial text contains spam content, and ensure that the text content does not violate laws and regulations, such as copyright law, privacy law, etc. If potential legal issues are found, corresponding modification suggestions will be provided, such as adding necessary disclaimers or deleting infringing content.

[0071] Obvious grammatical errors: Natural language processing technology is used to check obvious grammatical errors in the text, such as spelling errors, punctuation errors, etc. If these problems are found, specific modification suggestions will be provided, such as correcting spelling errors or adjusting punctuation.

[0072] 2. Logical review

[0073] The second review node conducts a logical review of the initial text. This node focuses on evaluating the semantic accuracy, logical consistency, and rationality of the text structure. For example, it checks whether the arguments in the text are coherent and whether there are logical contradictions or unreasonable inferences. If logical problems are found, the node will make logical modification suggestions to help the author optimize the argumentation process and enhance the persuasiveness and credibility of the text. The review content of the second review node specifically includes the following:

[0074] Semantic accuracy: The second review node will ensure that the text content is semantically accurate, without ambiguous or misleading expressions, or specifying the use of certain persons for expression. If unclear or misleading semantics are found, specific modification suggestions will be provided, such as reorganizing sentences or clarifying ambiguous expressions.

[0075] Logical consistency: Check whether the logical relationship in the text is reasonable and whether the parts are coherent. For example, whether the content of the title is consistent, avoid the situation where the title content is more than the article content, and the article must cover all the content of the title. For example, whether the content of the paragraphs is too different, or there is too much repetition between paragraphs. If there are logical inconsistencies or contradictions, specific modification suggestions will be provided, such as adjusting the order of paragraphs or adding necessary background information.

[0076] Reasonable structure of the article: Evaluate the overall structure of the article to ensure that the transition between paragraphs is natural, the logic is clear, whether the number of words in the article meets the requirements, whether the number of words in the paragraphs meets the requirements, and whether the article conforms to the general format. If it is found that the structure is unreasonable or the connection between paragraphs is not smooth, specific modification suggestions will be provided, such as re-dividing paragraphs or adding transition sentences.

[0077] 3. Professional review

[0078] The third review node is responsible for professional review, which evaluates whether the text shows the necessary expertise and accurate technical details. This node will check whether the article adopts the specified language style, whether the terminology is used correctly, and whether the text citation content is standardized. If any professional deficiencies are found, it will provide corresponding professional modification suggestions to ensure that the text is authoritative and accurate in front of professional readers. The review content of the third review node includes the following:

[0079] Language style: The third review node will check whether the language style of the text meets the expectations of the target reader group, such as formal, academic, colloquial, etc., or detect whether it is written in a specified language, such as Chinese. If the language style is found to be inconsistent or inappropriate, specific modification suggestions will be provided, such as adjusting the tone or using more professional terminology.

[0080] Use of professional terminology: Ensure that professional terminology is used correctly to avoid misuse or abuse. If inappropriate use of terminology is found, specific modification suggestions will be provided, such as replacing the wrong term or providing the correct definition.

[0081] Citation format standardization: Check whether the format of the cited literature meets the standard, avoid citing specific data analysis, because large models may have errors in citing data, and avoid specific real information, such as hospital names, patient names, etc. If the citation format is found to be non-standard, specific modification suggestions will be provided, such as adjusting the citation format or supplementing necessary references.

[0082] Finally, the system will summarize all the modification suggestions from the above three review nodes to ensure that each suggestion is fully considered. The final modification suggestions will cover compliance, logic and professionalism to ensure that the text meets high standards in all dimensions. The final modification suggestions will be summarized into a detailed report, including the violated review rules, specific issues and corresponding suggestions. This report will guide the subsequent text revision work to ensure that all problems can be effectively resolved.

[0083] This application integrates the recommendations in the three areas of compliance, logic, and professionalism. In this way, a comprehensive revision plan can be formed to guide the author to make necessary revisions and improvements to the text. This process ensures that the final revision recommendations not only cover the requirements of each review dimension, but also take into account their interrelationships and priorities, so that the revised text can maintain the consistency and integrity of the content while meeting all review standards. The multi-dimensional review process adopted in this application helps to ensure that the text is not only legal and compliant, but also logically rigorous and professionally reliable, thereby improving the overall quality and credibility of the text.

[0084] The review order of the above three review nodes can be executed sequentially or in parallel.

[0085] If it is executed sequentially, after the first audit node completes the audit, the text with the compliance modification suggestions will be passed to the second audit node, the second audit node will perform a logical audit on the text, and then the text with the compliance modification suggestions and logical modification suggestions will be passed to the third audit node, the third audit node will perform a professional audit on the text, and finally a final text containing compliance modification suggestions, logical modification suggestions and professional modification suggestions will be obtained.

[0086] If it is executed in parallel, the initial text will be sent to three review nodes at the same time. The three review nodes will use different dimensions to review the text at the same time, and then send their respective review results to any one of the review nodes for summary.

[0087] As an optional implementation method, modifying the initial text according to the modification suggestions through the revision nodes in the large model includes: analyzing the overall modification suggestions through the revision nodes in the large model to determine the modification details of each position in the initial text, wherein the modification details include: violated review rules, specific problems and corresponding suggestions; analyzing the violated review rules and specific problems according to preset prompt words, and modifying according to the suggestions.

[0088] The system uses the revision node in its built-in large-scale machine learning model to comprehensively analyze all the revision suggestions collected from different review nodes. The task of the revision node is to identify the specific locations in the initial text that need to be improved and determine the detailed revision details for each location. This includes clearly indicating which review rules have been violated, what the specific problems are, and specific suggestions for these problems.

[0089] The revision nodes further refine the analysis and provide clear guidance for each location in the initial text that needs to be revised. This involves identifying specific sentences or paragraphs in the text and pointing out which specific review standards they violate, such as grammatical errors, factual inaccuracies, logical inconsistencies, etc. At the same time, it will provide corresponding suggestions for each problem. These suggestions may include replacing vocabulary, restructuring sentence structure, adding references or correcting data.

[0090] For example, the audit rule that was violated is a grammatical error, the specific problem is that a word was misspelled, and the corresponding suggestion is to correct the spelling error.

[0091] For another example, the review rule that was violated is: incoherent logic; the specific problem is: the logic of a certain sentence is unclear; the corresponding suggestion is: reorganize the sentence.

[0092] In order to improve the efficiency and accuracy of revisions, the revision node will use a series of preset prompt words to assist the analysis and revision process. These prompt words are pre-defined keywords or phrases that are used to quickly locate problems and provide solutions.

[0093] Examples of prompt words for revision nodes are as follows:

[0094] Suppose you are a professional medical content modification expert, and your task is to modify the medical article according to the modification suggestions provided by the review assistant. You need to carefully understand each modification suggestion and make corresponding adjustments to the original article. Please follow the steps below to make modifications: 1. Read the original article: Carefully read the title and content of the original article to ensure that you have a full understanding of its core information, style and structure. 2. Analyze the review results: Carefully analyze the review results provided by the review assistant, including failed rules, specific problems and corresponding modification suggestions. 3. Make modifications: Make necessary adjustments to the article based on the specific problems and modification suggestions in the review results. Ensure that each problem is effectively resolved while retaining the core information, style and article structure (such as paragraph information and word count) of the original article as much as possible. 4. Return the modification results: Return the modified article in the specified JSON format, ensuring that the format is correct and the content is complete.

[0095] Please modify the medical article according to the review results provided by the following review assistant, and return the results in the specified JSON format: original article title: {draft_title}, original article content: {draft_content}, review results: {content_review}. Please carefully analyze each failed rule, specific issues and corresponding modification suggestions, and make necessary modifications to the article. At the same time, please try to retain the core information, style and article structure (such as paragraph information and word count) of the original article.

[0096] According to the detailed modification suggestions and prompts provided by the revision node, the system actually modifies the initial text. In this way, it can ensure that the text complies with all review rules while improving its logic and professionalism, and ultimately meeting the publishing standards. This large-model-based revision node operation not only improves the efficiency of text modification, but also ensures that the modified text can meet various review requirements, thereby improving the overall quality and reliability of the text.

[0097] As an optional implementation method, re-reviewing the modified text through multiple review nodes includes: adjusting the review standards by setting prompt words, wherein the setting prompt words is used to set that no additional modification suggestions will be provided when the error importance is lower than the importance threshold; and re-reviewing the modified text according to the adjusted review standards.

[0098] After the text has been initially revised, in order to more precisely control the review process and avoid excessive revisions, the system will adjust the review criteria by setting specific prompt words. These prompt words are designed to indicate the level of importance of the errors, helping the review nodes determine which errors are critical and must be corrected, and which may be relatively minor and can be ignored without affecting the overall quality. For example, for some non-critical improper use of punctuation or minor grammatical problems, if they have little impact on understanding the content, the system may choose not to provide additional revision suggestions. Such adjustments ensure that the review process is both strict and flexible, and can be adaptively adjusted according to the actual situation of the text and publication requirements.

[0099] Once the review criteria are adjusted, all revised texts will be evaluated again by multiple review nodes. This time, the review nodes will judge whether each part of the text still needs further revision according to the new criteria. Since prompt words have been applied to define the importance threshold of errors, the review nodes will pay more attention to those issues that have a significant impact on the quality and compliance of the text. This targeted re-review not only improves efficiency, but also ensures that the text meets the basic standards without sacrificing creativity and fluency due to excessive nitpicking about details.

[0100] During the re-review process, if there are areas for improvement, the revision node will continue to fine-tune the text based on the latest review suggestions. This process may be repeated until all review nodes confirm that the text has reached the predetermined publishing standards. At this point, the text is considered to be the final version and can be used for publication or other purposes. This cyclical iterative approach ensures that the quality of the text is fully guaranteed, and also reflects the flexibility and efficiency of the large model in handling complex text tasks.

[0101] In this way, the system is able to avoid unnecessary excessive modifications while maintaining the quality of the text, thereby saving time and resources and speeding up the entire process from text creation to publication.

[0102] This application also provides a schematic diagram of the overall process of text review and modification, such as Figure 3 As shown, the following steps are included.

[0103] Step 1: The user enters text content.

[0104] Step 2: Joint audit by multiple audit nodes.

[0105] Each review node is responsible for different review tasks, ensuring that the text meets the review standards in all aspects.

[0106] First review node: Perform a preliminary review, focusing on basic issues such as sensitive word filtering, basic legality review, and obvious grammatical errors.

[0107] Second review node: Conduct a detailed review, focusing on semantic accuracy, contextual logical consistency, and the rationality of the text structure.

[0108] The third review node: responsible for in-depth review, checking the language style of the text, the correctness of the use of professional terms, and the standardization of the citation format.

[0109] Step 3: Modify the content of the revision node.

[0110] The revision node comprehensively revises the text content based on the modification suggestions put forward by the previous review nodes. The revision node is driven by the big model and its task is to integrate all review opinions and implement revision operations.

[0111] Step 4: Circular review.

[0112] The revised text content is returned to the review node for a new round of review. The review node confirms whether the revised content solves the previously pointed out problems. If there are still problems with the revised text, a new round of review and revision is carried out according to steps 2-3. This process will be repeated until the text passes the inspection of all review nodes. In each round of review, the review node may fine-tune the review criteria based on the results of the previous review to more accurately identify and solve problems.

[0113] Step 5: Review completion and text output.

[0114] When the text has passed multiple rounds of review and all issues have been effectively revised at the revision node, the system will confirm that the review is complete and output the final version of the text. The system generates the final approved text for users to use or publish, and notifies users that the review is complete.

[0115] This application proposes a method for text review and modification, which, by introducing multiple virtual reviewers and virtual editors, realizes the automation, multi-level and cyclic iteration of the review process, thereby improving the efficiency, accuracy and consistency of the review. Specifically, the overall idea and working principle of this technical solution are as follows:

[0116] 1. Role setting and division of labor: First, multiple virtual auditor roles (audit nodes) are set according to different audit requirements and scenarios. Each audit node focuses on a certain aspect of the audit article content according to specific audit conditions or standards. In this way, complex audit tasks can be broken down into multiple simple subtasks, allowing the large model to focus on processing specific audit standards and improve the accuracy of the audit.

[0117] 2. Review process design: The review process is designed as multiple review nodes to review the content of the article in turn. After completing its own review task, each review node will pass the modification suggestions to the next review node. In this way, the review task is passed between multiple review nodes to ensure that each review standard is fully considered and verified.

[0118] 3. Introduction of the revision role: After the review node proposes revision opinions, the role of the virtual revision person (revision node) is introduced. The revision node makes corresponding modifications to the article content based on the revision opinions of the previous review node. The revision node role ensures that the opinions of the review node are effectively implemented and the article content is revised to meet the review requirements.

[0119] 4. Circular review and revision: The modified content at the revision node will be resubmitted to the first review node for review. This circular review method can further improve the accuracy and consistency of the review and ensure that the revision opinions of each round of review are fully implemented and tested.

[0120] 5. Task focus and performance improvement: By allowing the large model to focus on processing a single task at each review node, the model's errors in processing multiple complex tasks can be reduced, thereby improving the performance and efficiency of the overall review system. At the same time, through multiple rounds of review and revision, the risk of misreview and missed review in content review can be effectively reduced.

[0121] This application achieves the following beneficial effects:

[0122] Improve audit efficiency: Traditional text content audits mostly rely on manual operations, and the audit efficiency is limited by the speed and energy of manual work. In contrast, this application uses the efficient computing power of large models to complete the audit task of a large amount of text in a short period of time, greatly improving the audit efficiency.

[0123] Ensure consistency in review standards: Manual review is subjective, and the judgment standards of different reviewers may be inconsistent, resulting in differences in review results. This application uses the standardized review process of a large model through a design of joint review by multiple reviewers to ensure that each text can be processed according to unified review standards, thereby reducing misreviews and missed reviews, while ensuring that the content of the article is comprehensively and carefully checked in multiple dimensions. Compared with the single review process that may be used in existing solutions, this solution achieves more efficient and accurate review results through the joint work of multiple reviewers. In addition, the role of the reviser driven by the large model not only improves the efficiency of revisions, but also ensures the uniformity and consistency of revision quality.

[0124] Reduce labor costs: Traditional manual review requires a lot of human resources, especially auditors with professional backgrounds, which is costly. This application uses a large model to achieve automated review, significantly reducing the reliance on manual reviewers, thereby reducing the labor costs in the review process.

[0125] High accuracy of audit results: The large model has strong capabilities in semantic analysis, logical judgment, and format standard checking, and can accurately identify various problems in the text. Compared with omissions and misjudgments that are prone to occur in manual audits, this application not only ensures accuracy, but also improves the comprehensiveness and depth of the audit.

[0126] Support multiple rounds of review: This application has designed a multiple round review and revision cycle mechanism to ensure that the content after each round of revision will be reviewed again until the text meets all standards. This cycle review mechanism ensures the rigor and thoroughness of the review process and is a guarantee for high-quality article review. Compared with the traditional review process that only conducts one review, this cycle review mechanism provides a more in-depth review and improvement process, which can more effectively identify and correct potential problems and ultimately output high-quality text content.

[0127] Strong flexibility and scalability: The audit standards and models in this application can be customized and adjusted according to different application scenarios and requirements, and are suitable for different types of text content audit tasks. This flexibility makes the solution not only suitable for regular text audits, but also adapts to special fields or high-standard audit requirements, providing highly targeted audit services.

[0128] Based on the same technical concept, the present application provides a device for text review and modification, such as Figure 4 As shown, the device comprises:

[0129] The acquisition module 401 is used to acquire the initial text to be reviewed;

[0130] The review module 402 is used to review the initial text in different dimensions through multiple review nodes in the large model and give modification suggestions, wherein the modification suggestions are used to indicate that the text review has not passed;

[0131] A modification module 403 is used to modify the initial text according to the modification suggestions through the revision nodes in the large model to obtain a modified text;

[0132] The generation module 404 is used to re-examine the modified text through multiple review nodes until no more modification suggestions are generated, thereby obtaining the modified final text, wherein no more modification suggestions are generated to indicate that the text review has passed.

[0133] Optionally, the audit module 402 is used to:

[0134] Conduct compliance review of the initial text through the first review node in the large model and provide compliance modification suggestions;

[0135] Conduct a logical review of the initial text through the second review node in the large model and provide logical modification suggestions;

[0136] Conduct professional review of the initial text through the third review node in the big model and provide professional modification suggestions;

[0137] Comprehensive compliance modification suggestions, logical modification suggestions and professional modification suggestions to obtain the final modification suggestions.

[0138] Optionally, the audit module 402 is used to:

[0139] Through the first review node in the large model, a text anti-spam checking algorithm is used to detect whether the initial text contains spam content;

[0140] Use preset sensitive words to detect whether the initial text contains sensitive words;

[0141] Natural language processing technology is used to detect whether the initial text has grammatical errors.

[0142] Optionally, the audit module 402 is used to:

[0143] Through the second review node in the large model, natural language processing technology is used to detect whether the semantics of the initial text is accurate, whether the logic of the text content is consistent, and whether the text content structure is reasonable.

[0144] Optionally, the audit module 402 is used to:

[0145] Through the third review node in the big model, it is checked whether the initial text is written in the specified language, whether the professional terminology is accurate, and whether the quoted content is standardized.

[0146] Optionally, the modification module 403 is used to:

[0147] The overall modification suggestions are analyzed through the revision nodes in the large model to determine the modification details of each position in the initial text, where the modification details include: the audit rules violated, specific problems and corresponding suggestions;

[0148] Analyze the violated audit rules and specific issues according to the preset prompts, and make modifications based on the suggestions.

[0149] Optionally, the generating module 404 is used to:

[0150] Adjust the review criteria by setting prompt words, where setting prompt words is used to set that no additional modification suggestions will be provided when the error importance is below the importance threshold;

[0151] Re-review the revised text according to the adjusted review standards.

[0152] like Figure 5 As shown, an embodiment of the present application provides an electronic device, including a processor 501, a communication interface 502, a memory 503 and a communication bus 504, wherein the processor 501, the communication interface 502, and the memory 503 communicate with each other through the communication bus 504.

[0153] The memory 503 is used to store computer programs.

[0154] In one embodiment of the present application, the processor 501 is used to implement the text review and modification method provided by any of the aforementioned method embodiments when executing the program stored in the memory 503.

[0155] An embodiment of the present application also provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the method for text review and modification provided in any of the aforementioned method embodiments are implemented.

[0156] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0157] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a general hardware platform, and of course, by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the relevant technology can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0158] It should be understood that the terms used herein are only for the purpose of describing specific example embodiments and are not intended to be limiting. Unless the context clearly indicates otherwise, the singular forms "one", "an" and "said" as used herein may also be meant to include plural forms. The terms "include", "comprise", "contain", and "have" are inclusive, and therefore specify the existence of stated features, steps, operations, elements and / or parts, but do not exclude the existence or addition of one or more other features, steps, operations, elements, parts, and / or combinations thereof. The method steps, processes, and operations described herein are not interpreted as necessarily requiring them to be performed in the specific order described or illustrated, unless the execution order is clearly indicated. It should also be understood that additional or alternative steps may be used.

[0159] The foregoing is merely a specific embodiment of the present invention, which enables those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be 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 present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A method for text review and modification, characterized in that: The method comprises: Get the initial text to be reviewed; The initial text is reviewed in different dimensions through multiple review nodes in the large model, and modification suggestions are given, wherein the modification suggestions are used to indicate that the text review has not passed; Modifying the initial text according to the modification suggestion through the revision node in the large model to obtain a modified text; The modified text is re-reviewed through the multiple review nodes until no modification suggestions are generated, thereby obtaining a modified final text, wherein no modification suggestions are generated to indicate that the text review has been passed.

2. The method according to claim 1, characterized in that The initial text is reviewed in different dimensions through multiple review nodes in the large model, and modification suggestions are given, including: Conducting compliance review on the initial text through the first review node in the large model and providing compliance modification suggestions; Performing a logic review on the initial text through the second review node in the large model and providing logic modification suggestions; Conducting a professional review of the initial text through the third review node in the large model and providing professional modification suggestions; The compliance modification suggestions, the logic modification suggestions and the professional modification suggestions are combined to obtain a final modification suggestion.

3. The method according to claim 2, characterized in that Performing compliance review on the initial text through the first review node in the large model includes: Using a first audit node in the large model, a text anti-spam checking algorithm is used to detect whether the initial text contains spam content; Using preset sensitive words to detect whether the initial text contains sensitive words; Natural language processing technology is used to detect whether the initial text has any grammatical errors.

4. The method according to claim 2, characterized in that Performing a logical review on the initial text through the second review node in the large model includes: Through the second review node in the large model, natural language processing technology is used to detect whether the semantics of the initial text is accurate, whether the logic of the text content is consistent, and whether the text content structure is reasonable.

5. The method according to claim 2, characterized in that: The professional review of the initial text through the third review node in the large model includes: Through the third review node in the large model, it is detected whether the initial text is written in the specified language, whether the professional terms are accurate, and whether the quoted content is standardized.

6. The method according to claim 1, characterized in that Modifying the initial text according to the modification suggestion through the revision node in the large model includes: Analyze the overall modification suggestions through the revision nodes in the large model to determine the modification details of each position in the initial text, wherein the modification details include: the audit rules violated, specific problems and corresponding suggestions; Analyze the violated audit rules and specific issues according to the preset prompts, and make modifications based on the suggestions.

7. The method according to claim 1, characterized in that Re-examining the modified text through the multiple review nodes includes: Adjusting the review criteria by setting prompt words, wherein the set prompt words are used to set that no additional modification suggestions will be provided when the importance of the error is lower than the importance threshold; Re-review the modified text according to the adjusted review standards.

8. A device for text review and modification, characterized in that: The device comprises: The acquisition module is used to obtain the initial text to be reviewed; An audit module, used to audit the initial text in different dimensions through multiple audit nodes in the large model, and give modification suggestions, wherein the modification suggestions are used to indicate that the text has failed the audit; A modification module, used to modify the initial text according to the modification suggestion through the revision nodes in the large model to obtain a modified text; A generation module is used to re-examine the modified text through the multiple review nodes until no more modification suggestions are generated, thereby obtaining a modified final text, wherein no more modification suggestions are generated to indicate that the text review has passed.

9. An electronic device, characterized in that: It includes a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other through the communication bus; Memory, used to store computer programs; A processor, for implementing any of the methods described in claims 1-7 when executing a program stored in a memory.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images