Attachment auditing method, device and equipment and storage medium

Abstract

The invention relates to the technical field of artificial intelligence, is used for an artificial intelligence scene, a financial service scene and a medical health service scene, and provides an attachment auditing method, device and equipment and a storage medium, applied to an attachment auditing platform, the attachment auditing method comprises the steps of obtaining a target attachment; the target attachment comprises a plurality of first texts; when the current time is within the first auditing time period, first semantic information of the first text is obtained based on a semantic understanding network of a preset auditing model; based on an information auditing network of a preset auditing model, according to the first text and the first semantic information, auditing the target attachment to obtain an auditing result corresponding to the target attachment; the auditing result is used for indicating the score of the risk of damaging the consumer rights and interests and the reason of the risk of damaging the consumer rights and interests of the target accessory, so that the auditing convenience when auditing whether the target accessory damages the consumer rights and interests or not is realized, and the convenience of protecting and supervising the consumer rights and interests of the target accessory is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence and the technical field of natural language processing, in particular to an attachment auditing method and device, equipment and a storage medium. BACKGROUND

[0002] With the increasing importance of consumer rights protection, the supervision of consumer rights protection is becoming increasingly strict. Based on this, enterprises need to integrate the corresponding supervision measures of consumer rights protection into the business processes of enterprises. For example, for the relevant business of an enterprise, it is necessary to manually audit whether the attachments related to the corresponding business have the risk of damaging consumer rights, so as to subsequently supervise the consumer rights protection of the corresponding business. However, with the increase in the number of attachments, manual auditing of the attachments requires a large amount of manpower and time, which can easily result in poor auditing convenience when auditing whether the attachments damage consumer rights, thereby leading to poor convenience of consumer rights protection supervision of the attachments.

[0003] Since different business scenarios such as financial business scenarios and medical health business scenarios can generate corresponding attachments, different business scenarios need to be audited for corresponding attachments. Therefore, in the case of poor auditing convenience when auditing whether the attachments damage consumer rights, thereby leading to poor convenience of consumer rights protection supervision of the attachments, it is easy to cause adverse effects on the business promotion of the financial business scenarios and the medical health business scenarios. SUMMARY

[0004] The main purpose of the present application is to provide an attachment auditing method, device, equipment and storage medium, which aims to improve the auditing convenience when auditing whether the target attachment damages consumer rights, thereby improving the convenience of consumer rights protection supervision of the target attachment.

[0005] In a first aspect, the present application provides an attachment auditing method applied to an attachment auditing platform, the auditing method comprising: obtaining, by the attachment auditing platform, a target attachment related to consumer rights protection; the target attachment comprising a plurality of first texts; when the current time is within a first auditing time period, performing semantic recognition processing on the first text based on a semantic understanding network of a preset auditing model to obtain first semantic information of the first text; performing auditing on the target attachment based on an information auditing network of the preset auditing model according to the first text and the first semantic information to obtain an auditing result corresponding to the target attachment; the auditing result being used to indicate a score of the target attachment having the risk of damaging consumer rights and a reason of the target attachment having the risk of damaging consumer rights.

[0006] In a second aspect, the present application also provides an attachment auditing device, which comprises: An attachment obtaining module is configured to obtain a target attachment related to consumer rights protection through the attachment auditing platform; the target attachment comprises a plurality of first texts; A semantic recognition module is configured to, when the current time is within a first auditing time period, perform semantic recognition processing on the first texts based on a semantic understanding network of a preset auditing model, to obtain first semantic information of the first texts; An information auditing module is configured to perform auditing on the target attachment based on an information auditing network of the preset auditing model according to the first texts and the first semantic information, to obtain an auditing result corresponding to the target attachment; the auditing result is used to indicate a score of the target attachment having a risk of damaging consumer rights and a reason of the target attachment having the risk of damaging consumer rights.

[0007] In a third aspect, the present application also provides a computer device, which comprises a memory and a processor; The memory is configured to store a computer program; The processor is configured to execute the computer program and implement the steps of the above-mentioned attachment auditing method when executing the computer program.

[0008] In a fourth aspect, the present application also provides a computer readable storage medium, which stores a computer program; the computer program is executed by a processor to implement the steps of the above-mentioned attachment auditing method.

[0009] The present application provides an attachment auditing method, device, equipment and storage medium, the attachment auditing method is applied to an attachment auditing platform, and the attachment auditing method comprises the following steps: obtaining a target attachment related to consumer rights protection through the attachment auditing platform; the target attachment comprises a plurality of first texts; when the current time is within a first auditing time period, performing semantic recognition processing on the first texts based on a semantic understanding network of a preset auditing model, to obtain first semantic information of the first texts; performing auditing on the target attachment based on an information auditing network of the preset auditing model according to the first texts and the first semantic information, to obtain an auditing result corresponding to the target attachment; the auditing result is used to indicate a score of the target attachment having a risk of damaging consumer rights and a reason of the target attachment having the risk of damaging consumer rights.

[0010] In a case where the target attachment is acquired, the attachment review platform can utilize a semantic understanding network of the preset review model to perform semantic recognition processing on the first text included in the target attachment, to obtain corresponding first semantic information. The first text and the first semantic information can be used by an information review network of the preset review model to review whether the target attachment has content that damages the rights and interests of consumers, to obtain a corresponding review result. The review result can be used to indicate a score of the target attachment having the risk of damaging the rights and interests of consumers and a reason why the target attachment has the risk of damaging the rights and interests of consumers. Based on this, in a case where the attachment review platform utilizes the preset review model to review the target attachment, the review convenience when reviewing whether the target attachment damages the rights and interests of consumers is improved, and thus the convenience of consumer rights and interests protection supervision on the target attachment is improved.

[0011] In a case where different business scenarios such as a financial business scenario and a medical health business scenario need to perform corresponding attachment review, based on the improvement of the review convenience when reviewing whether the target attachment damages the rights and interests of consumers, the convenience of consumer rights and interests protection supervision on the target attachment in the promotion of the financial business and the medical health business is improved, and thus the development of the financial business and the medical health business is promoted. For example, the target attachment involved in the financial business can be related to business, commerce, promotion, sale, and other business activities of an enterprise, or can be related to investment, credit, stocks, asset portfolio, and other financial conditions of the enterprise. The target attachment involved in the medical health business can be related to medical devices maintenance, medical drug consumption, old-age facility maintenance, and old-age pension data involved in medical data. In a case where the attachment review method is used to review the target attachment involved in the financial business and the medical health business, the target attachment can be more conveniently reviewed whether it damages the rights and interests of consumers, and thus the convenience of consumer rights and interests protection supervision on the target attachment in the financial business and the medical health business is improved, without relying on manual review, and thus the development of the financial business and the medical health business is promoted. BRIEF DESCRIPTION OF DRAWINGS

[0012] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creating laborious work.

[0013] Figure 1 is an application environment schematic diagram of an attachment review method in an embodiment of the present application; Figure 2 is a flowchart of an attachment review method in an embodiment of the present application; Figure 3is a schematic diagram of an attachment review process of an attachment review platform involved in an embodiment of the present application; Figure 4 is a structural schematic diagram of an attachment review device in an embodiment of the present application; Figure 5 is a structural schematic diagram of a computer device in an embodiment of the present application; Figure 6 is another structural schematic diagram of a computer device in an embodiment of the present application. DETAILED DESCRIPTION

[0014] The technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all the other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of the present application.

[0015] The attachment review method provided by the embodiments of the present application can be applied to, for example, Figure 1In an application environment, a client communicates with the attachment review platform through a network, and a server communicates with the attachment review platform through the network, and the client and the server can interact with each other through the attachment review platform. The attachment review platform can obtain the target attachment related to consumer rights protection through the client, and upload the target attachment to the server, so that when the server is in the first review time period at the current time, the first text included in the target attachment is subjected to semantic recognition processing based on the semantic understanding network of the preset review model, to obtain the first semantic information of the first text; based on the information review network of the preset review model, the target attachment is reviewed according to the first text and the first semantic information, to obtain the review result corresponding to the target attachment; the review result is used to indicate the score of the target attachment having the risk of damaging consumer rights and the reason of the target attachment having the risk of damaging consumer rights, and then the server can feed back the review result corresponding to the target attachment to the attachment review platform, and then to the client, so that the client can inform the user of the review result corresponding to the target attachment. In this application, the preset review model includes an artificial intelligence model, which is used to sequentially perform semantic recognition processing and review on the target attachment to obtain the corresponding review result. In the process of determining the review result corresponding to the target attachment by using the preset review model, it is not necessary to rely on manual review of whether the target attachment damages consumer rights, which is conducive to improving the convenience of reviewing whether the target attachment damages consumer rights, and thus improving the convenience of consumer rights protection supervision of the target attachment. Based on the improvement of the convenience of reviewing whether the target attachment damages consumer rights, and the convenience of consumer rights protection supervision of the target attachment, it is conducive to promoting the development of financial business and medical health business. For example, the target attachment involved in the financial business can be reviewed by using the attachment review method, and then the review result corresponding to the target attachment is obtained. The target attachment involved in the financial business includes, for example, attachments related to business, commerce, promotion, sale, etc. of enterprises, and attachments related to investment, credit, stocks, asset portfolio, etc. of enterprises. Based on this, it is more convenient to review whether there is content with the risk of damaging consumer rights in the target attachment involved in the financial business, and then inform the user of the review result corresponding to the target attachment, so as to supervise the target attachment for consumer rights protection, thereby promoting the development of financial business. The target attachment involved in the medical health business can be reviewed by using the attachment review method, and then the review result corresponding to the target attachment is obtained. The target attachment involved in the medical health business includes, for example, target attachments related to maintenance of medical equipment, consumption of medical drugs, maintenance of old-age facilities, pension data, etc. involved in medical data, etc.Based on this, it can be more convenient to audit whether there is content with damage to consumer rights and interests risk in the target attachment involved in the medical and health business, and then inform the user of the audit result corresponding to the target attachment, so as to protect the consumer rights and interests, and promote the development of medical and health business. Of course, it is not limited to this, the server can also deploy the preset audit model to the client, and then the client can use the attachment audit method to determine the audit information corresponding to the target attachment when obtaining the target attachment to be audited. The client can be, but is not limited to, various personal computers, notebook computers, smart phones, tablet computers and portable wearable devices. The server can be implemented by an independent server or a server cluster composed of multiple servers. The present application will be described in detail below through specific embodiments.

[0016] Please refer to Figure 2 , Figure 2 A flowchart of an attachment audit method provided by an embodiment of the present application is shown, which includes the following steps: S101: Obtain a target attachment related to consumer rights protection through an attachment audit platform; the target attachment includes a plurality of first texts.

[0017] Illustratively, relevant personnel can upload the target attachment related to consumer rights protection to the attachment audit platform. The attachment audit platform can directly obtain the target attachment for auditing when it needs to audit the target attachment.

[0018] The target attachment related to consumer rights protection can be used to indicate various types of proof documents and materials required to be submitted, audited or left in the consumer rights protection assessment (CPA) process. The target attachment includes, for example, product risk disclosure, customer risk assessment report, marketing materials, audio and video recordings in the sales process, customer contracts and terms, internal approval process records, etc., without limitation.

[0019] For different business scenarios, such as financial business scenarios, medical and health business scenarios, etc., the target attachment can be different.

[0020] For example, for financial business in the financial business scenario, the target attachment can include attachments related to business, business, promotion, sale, etc. of the enterprise, and attachments related to investment, credit, stock, asset portfolio, etc. of the enterprise, etc. without limitation. For medical and health business in the medical and health business scenario, the target attachment can include attachments related to maintenance of medical equipment, consumption of medical drugs, maintenance of pension facilities, pension data, etc. involved in medical data, etc. without limitation.

[0021] Exemplarily, the target attachment can include a plurality of first texts. The first texts can be used to interpret and refine the target attachment.

[0022] In some embodiments, the target attachment can be uploaded to the attachment review platform in different forms. For example, the target attachment can be uploaded to the attachment review platform in the form of text, picture, video, etc. When the attachment review platform obtains the target attachment, if the target attachment is uploaded to the attachment review platform in a form other than text, the attachment review platform can perform text conversion processing on the target attachment to convert non-text objects of the target attachment into corresponding text for subsequent processing.

[0023] In an exemplary embodiment, the attachment review platform can perform optical character recognition (OCR) on the target attachment to convert the target attachment in picture form into the target attachment in text form, and then obtain the first text included in the target attachment. Of course, it is not limited thereto, and is not limited herein.

[0024] The attachment review platform can then use the first text included in the target attachment to evaluate whether the target attachment harms the consumer rights and interests, which is conducive to improving the convenience of subsequent review of whether the target attachment harms the consumer rights and interests, thereby facilitating the convenience of consumer rights and interests protection supervision of the target attachment.

[0025] S102: When the current time is within the first review time period, performing semantic recognition processing on the first text based on a semantic understanding network of a preset review model to obtain first semantic information of the first text.

[0026] When the attachment review platform obtains the target attachment, if the current time is within the first review time period, the attachment review platform can start the automatic review process of the target attachment.

[0027] Exemplarily, the preset review model can use an artificial intelligence model with an agent (Agent) capability. The preset review model can also be referred to as Agent AI. The preset review model can identify different forms of text content in the target attachment, such as identifying text content in the target attachment in the form of Excel, Word, etc. for the review of the target attachment, thereby reviewing whether the target attachment harms the consumer rights and interests, reducing the workload of manual review, and effectively avoiding business security risks in different business scenarios.

[0028] For example, the attachment review platform can input the target attachment into a preset review model. The preset review model can perform semantic recognition processing on the first text included in the target attachment through a semantic understanding network to obtain first semantic information of the first text when the target attachment is obtained. The semantic understanding network of the preset review model can use a large language model (LLM) for text understanding and generation to perform semantic recognition processing on the first text to obtain the first semantic information of the first text. Of course, it is not limited to this, and is not limited herein.

[0029] For example, the first semantic information of the first text can be used to indicate whether the target attachment has risk prompt information, whether the target attachment involves inducement sales information, whether the target attachment has unfair terms, whether the target attachment has false information, etc. At least one of the above, then the attachment review platform can subsequently cooperate with the first text and the first semantic information to review whether the target attachment has the risk of damaging the consumer rights and interests.

[0030] For example, the first semantic information of the first text can be used to indicate whether the target attachment has risk prompt information, whether the target attachment involves inducement sales information, whether the target attachment has unfair terms, whether the target attachment has false information, etc. At least one of the above, then the attachment review platform can subsequently cooperate with the first text and the first semantic information to review whether the target attachment has the risk of damaging the consumer rights and interests.

[0031] Taking financial insurance in a financial service scenario as an example, the first text in the target attachment related to the financial insurance can include the insurance clause of the financial insurance, the claim settlement clause of the financial insurance, the exemption clause of the financial insurance, the promotional text of the financial insurance, etc. without limitation. Correspondingly, when the preset review model obtains the first text in the target attachment related to the financial insurance, the first semantic information of the first text can be determined. The first semantic information is used to indicate, for example, whether one or more of the insurance clause, the claim settlement clause, the exemption clause, and the promotional text of the financial insurance involve at least one of risk prompt information, inducement consumption information, unfair terms, and false information.

[0032] Taking medical insurance in a medical health service scenario as an example, the first text in the target attachment related to the medical insurance can include the medical insurance medical security range description, the medical insurance medical reimbursement amount, the medical insurance promotional text, etc. without limitation. Correspondingly, when the preset review model obtains the first text in the target attachment related to the medical insurance, the first semantic information of the first text can be determined. The first semantic information is used to indicate, for example, whether one or more of the medical insurance medical security range description, the medical reimbursement amount, and the promotional text involve at least one of risk prompt information, inducement consumption information, unfair terms, and false information.

[0033] When the current time is within the first review time period, the semantic understanding network of the preset review model can perform semantic recognition processing on the first text included in the target attachment in response to the attachment review platform starting automatic review of the target attachment, to obtain first semantic information of the first text, and the first text and the first semantic information can be used for subsequent evaluation by the preset review model of whether the target attachment has a risk of damaging consumer rights, thereby improving the convenience of subsequent consumer rights protection supervision of the target attachment.

[0034] S103: Based on the information review network of the preset review model, the target attachment is reviewed according to the first text and the first semantic information to obtain an audit result corresponding to the target attachment; the audit result is used to indicate a score of the target attachment having a risk of damaging consumer rights and a reason of the target attachment having a risk of damaging consumer rights.

[0035] In some embodiments, after determining the first semantic information of the first text, the first text and the first semantic information can be input into the information review network of the preset review model to review the target attachment according to the first text and the first semantic information through the information review network to obtain an audit result corresponding to the target attachment.

[0036] For example, in the case of inputting the first text and the first semantic information into the information review network, the information review model can evaluate whether the target attachment has a risk of damaging consumer rights according to the first text and the first semantic information.

[0037] Taking financial insurance in a financial service scenario as an example. The first text in the target attachment related to the financial insurance, for example, includes the exclusion clause of the financial insurance. In the process of obtaining the exclusion clause of the financial insurance and the first semantic information of the exclusion clause, if it is determined according to the exclusion clause of the financial insurance and the first semantic information of the exclusion clause that the exclusion clause lacks risk prompt information and the exclusion clause has unfair terms, it can be determined that the target attachment has a risk of damaging consumer rights, and further determine the audit result corresponding to the target attachment. For example, the information review network can determine a first score corresponding to the exclusion clause according to the exclusion clause lacking risk prompt information, and determine a second score corresponding to the exclusion clause according to the exclusion clause having unfair terms, and then combine the first score and the second score to determine a score of the target attachment having a risk of damaging consumer rights. The score can be determined according to one of the average value, weighted average value, and intermediate value between the first score and the second score, without limitation. The information review network can determine the reason of the target attachment having a risk of damaging consumer rights according to the first semantic information of the exclusion clause, such as the exclusion clause lacking risk prompt information and the exclusion clause having unfair terms.

[0038] Taking medical insurance in a medical health business scenario as an example. The first text in the target attachment related to medical insurance, for example, includes a promotional text of medical insurance. In the process of obtaining the promotional text of medical insurance and the first semantic information of the promotional text, if it is determined according to the promotional text of medical insurance and the first semantic information of the promotional text that the promotional text contains induced sales information and the promotional text contains false information, it can be determined that the target attachment has a risk of damaging the rights and interests of consumers, and then the audit result corresponding to the target attachment is determined. For example, the information audit network can determine the third score corresponding to the promotional text according to the existence of induced sales information in the promotional text, and determine the fourth score corresponding to the promotional text according to the existence of false information in the promotional text, and then combine the third score and the fourth score to determine the score of the target attachment with the risk of damaging the rights and interests of consumers. The score can be determined according to one of the average value, weighted average value, and intermediate value between the third score and the fourth score, without limitation. The information audit network can determine the reason why the target attachment has the risk of damaging the rights and interests of consumers according to the first semantic information of the promotional text, such as the existence of induced sales information in the promotional text and the existence of false information in the promotional text. By analogy.

[0039] The information audit network of the preset audit model can be used to audit the target attachment to obtain the audit result corresponding to the target attachment. In the case where the audit result is used to indicate the score of the target attachment with the risk of damaging the rights and interests of consumers and the reason why the target attachment has the risk of damaging the rights and interests of consumers, the audit result can be used to determine the score of the target attachment with the risk of damaging the rights and interests of consumers and the reason why the target attachment has the risk of damaging the rights and interests of consumers, and then the consumer rights protection supervision of the target attachment can be performed. Based on this, the attachment audit platform uses the preset audit model to audit the target attachment, which is beneficial to improve the convenience of auditing whether the target attachment damages the rights and interests of consumers, and then is beneficial to improve the convenience of consumer rights protection supervision of the target attachment.

[0040] In some embodiments, based on the semantic understanding network, the second semantic information similar to the first semantic information is determined according to the first semantic information, and the second text corresponding to the second semantic information is determined.

[0041] Based on the information audit network, the target attachment is audited according to the first text, the first semantic information, the second text, and the second semantic information, and the audit result corresponding to the target attachment is obtained.

[0042] For example, in a case where the semantic understanding network determines the first semantic information of the first text, the target attachment can contain one or more of the following: a text with similar or close semantics, a text with similar or close semantic scope, and a text with partially overlapping semantic scope, which can be used to replace the first text to evade consumer rights protection supervision. Based on this, in a case where the semantic understanding network determines the first semantic information, the semantic understanding network can continue to determine second semantic information similar to the first semantic information, and determine second text corresponding to the second semantic information. In a case where the second text and the second semantic information of the second text are determined, the second text and the second semantic information can be used to expand the review range of the preset audit model for subsequent consumer rights protection supervision of the target attachment, thereby improving the comprehensive review of whether the target attachment harms the consumer rights.

[0043] Taking medical insurance in the medical health business scenario as an example. The first text in the target attachment related to medical insurance, for example, includes a medical insurance medical care range description. In a case where the semantic understanding network determines that the first semantic information of the medical insurance medical care range description is used to indicate the medical expenses covered by the medical care range, the semantic understanding network can determine second semantic information similar to the first semantic information according to the first semantic information. For example, the second semantic information similar to the first semantic information includes a supplementary explanation related to medical expenses. The supplementary explanation related to medical expenses, for example, includes one or more of the respective numerical ranges, categories, etc. corresponding to the medical insurance reimbursement range, the drug or material range, the regular medication dosage, the hospital room reimbursement range, etc., so that the second text corresponding to the supplementary explanation related to medical expenses can include one or more of the medical insurance reimbursement range, the drug or material range, the regular medication dosage, the hospital room reimbursement range, etc.

[0044] In a case where the semantic understanding network determines the second semantic information similar to the first semantic information and determines the second text corresponding to the second semantic information, the preset audit model can review whether the target attachment harms the consumer rights by comprehensively considering the first text, the first semantic information, the second text, and the second semantic information.

[0045] Taking medical insurance in a medical health business scenario as an example. When the information auditing network audits the target attachment related to the medical insurance, the information auditing network can evaluate whether the medical insurance medical guarantee scope description, the medical expenses covered by the medical guarantee scope, the supplementary explanation related to the medical expenses, the numerical range corresponding to each of the supplementary explanation related to the medical expenses, the category, and the like have problems such as ambiguous wording, unclear range limitation, and the like. If there are, it can be determined that the target attachment related to the medical insurance has unfair terms, i.e., it can be determined that the target attachment has the risk of damaging the consumer rights and interests, and then the auditing result corresponding to the target attachment is determined. For example, the information auditing network can determine the fifth score corresponding to the medical guarantee scope description according to the problem of ambiguous wording and unclear range limitation in the medical guarantee scope description, and determine the sixth score corresponding to the supplementary explanation related to the medical expenses according to the problem of ambiguous wording and unclear range limitation in the supplementary explanation related to the medical expenses, and then combine the fifth score and the sixth score to determine the score of the risk of damaging the consumer rights and interests of the target attachment. The score can be determined according to one of the average value, the weighted average value, and the median value between the fifth score and the sixth score, without limitation. The information auditing network can determine the reason why the target attachment has the risk of damaging the consumer rights and interests according to the first semantic information of the medical guarantee scope description and the second semantic information of the supplementary explanation related to the medical expenses. For example, the reason why the target attachment has the risk of damaging the consumer rights and interests can include at least one of the problems of ambiguous wording and unclear range limitation in the medical expenses covered by the medical guarantee scope, the problems of unclear range limitation in the numerical range corresponding to each of the supplementary explanation related to the medical expenses, the category, and the like. Similarly.

[0046] After the semantic understanding network of the preset auditing model determines the first semantic information of the first text, it can determine second semantic information similar to the first semantic information, and determine a second text corresponding to the second semantic information, so that when the information auditing network of the preset auditing model audits the target attachment in combination with the first text, the first semantic information, the second text, and the second semantic information, and obtains the auditing result corresponding to the target attachment, the second text and the second semantic information can be used to expand the review range of the information auditing network when the consumer rights and interests protection supervision of the target attachment is performed, and thus the review comprehensiveness when the target attachment is reviewed for whether it damages the consumer rights and interests is improved, thereby improving the comprehensiveness of the consumer rights and interests protection supervision of the target attachment.

[0047] In some embodiments, the first text and the first semantic information are input into the information auditing network, so that the information auditing network determines a matching degree between the first text and a preset text in a preset attachment text library according to the first semantic information, the preset text is used to indicate a text with a risk of damaging consumer rights, when the matching degree between the first text and the preset text is greater than or equal to a preset matching degree threshold, a score of the target attachment with the risk of damaging consumer rights is determined according to a preset score corresponding to the preset text, and a reason of the target attachment with the risk of damaging consumer rights is determined according to the first semantic information and a preset semantic information corresponding to the preset text.

[0048] For example, the information auditing network can be trained in advance according to the preset text in the preset attachment text library, so that the information auditing network can evaluate the matching degree between the first text and the preset text in the preset attachment text library according to the first semantic information when the first text and the first semantic information are obtained. The preset text is used to indicate a text with a risk of damaging consumer rights. The preset text can have a preset semantic information. For example, the information auditing network can determine the matching degree between the first text and the preset text according to the matching degree between the first semantic information and the preset semantic information. When the semantics of the first semantic information and the preset semantic information are more similar or even the same, the matching degree between the first text and the preset text is higher. When the semantics of the first semantic information and the preset semantic information are greatly different, the matching degree between the first text and the preset text is lower.

[0049] When the information auditing network determines that the matching degree between the first text and the preset text is greater than or equal to the preset matching degree threshold, the information auditing network can determine that the target attachment has the risk of damaging consumer rights, the score of the target attachment with the risk of damaging consumer rights matches the preset score corresponding to the preset text, and the reason of the target attachment with the risk of damaging consumer rights matches the reason of the risk of damaging consumer rights indicated by the preset semantic information corresponding to the preset text, and then the information auditing network can determine the score of the target attachment with the risk of damaging consumer rights according to the preset score, and determine the reason of the target attachment with the risk of damaging consumer rights according to the first semantic information and the preset semantic information.

[0050] Taking financial insurance in a financial service scenario as an example. The first text in the target attachment related to the financial insurance includes, for example, a promotional text of the financial insurance. The promotional text of the financial insurance includes, for example, at least one of a coverage of the financial insurance, a premium of the financial insurance, and the like. The first semantic information of the coverage of the financial insurance can be used to indicate that the coverage of the financial insurance is up to a target coverage threshold. The first semantic information of the premium of the financial insurance can be used to indicate that the premium of the financial insurance is down to a target premium threshold. The preset text in the preset attachment text library includes, for example, at least one of a coverage of a preset insurance, a premium of the preset insurance, and the like under different types. The preset semantic information of the coverage of the preset insurance can be used to indicate that the coverage of the preset insurance is up to a preset coverage threshold. The preset semantic information of the premium of the preset insurance can be used to indicate that the premium of the preset insurance is down to a preset premium threshold. However, in a case where the first semantic information of the coverage of the preset insurance fails to be used to indicate one or more of whether the coverage of the preset insurance is continuously up to the preset coverage threshold during an insurance application period, a condition required to be met for the coverage of the preset insurance to be up to the preset coverage threshold, and the like, there is a situation of inducing sales by using ambiguous words for the coverage of the preset insurance. Correspondingly, in a case where the first semantic information of the premium of the preset insurance fails to be used to indicate one or more of whether the premium of the preset insurance is continuously up to the preset premium threshold during the insurance application period, a condition required to be met for the premium of the preset insurance to be down to the preset premium threshold, and the like, there is also a situation of inducing sales by using ambiguous words for the premium of the preset insurance. The information auditing network can determine, according to the first semantic information of the coverage of the financial insurance, that a matching degree between the first semantic information of the coverage of the financial insurance and the preset semantic information of the coverage of the preset insurance in the preset attachment text library is greater than or equal to a preset matching degree threshold, and then the information auditing network can determine that there is a situation of inducing sales by using ambiguous words for the coverage of the financial insurance. Correspondingly, the information auditing network can determine, according to the first semantic information of the premium of the financial insurance, that a matching degree between the first semantic information of the premium of the financial insurance and the preset semantic information of the premium of the preset insurance in the preset attachment text library is greater than or equal to the preset matching degree threshold, and then the information auditing network can determine that there is a situation of inducing sales by using ambiguous words for the premium of the financial insurance. Based on this, the information auditing network can determine a score of the target attachment related to the financial insurance having a risk of impairing consumer rights according to a first preset score corresponding to the coverage of the preset insurance and a second preset score corresponding to the premium of the preset insurance, and determine a reason why the target attachment related to the financial insurance has the risk of impairing consumer rights according to the first semantic information of the coverage of the financial insurance, the first semantic information of the premium of the financial insurance, the preset semantic information of the coverage of the preset insurance, and the preset semantic information of the premium of the preset insurance corresponding to the preset semantic information of the preset text. The reason includes, for example, a risk of inducing sales by using ambiguous words, thereby impairing consumer rights. By analogy.

[0051] Correspondingly, in a case where the information auditing network determines that the matching degree between the first text and the preset text is less than the preset matching degree threshold, the information auditing network can determine that the target attachment does not have the risk of impairing consumer rights, that the score of the risk of the target attachment impairing consumer rights does not match the preset score corresponding to the preset text, and that the reason for the risk of the target attachment impairing consumer rights does not match the reason for the risk of impairing consumer rights indicated by the preset semantic information corresponding to the preset text, and then the information auditing network can continue to determine whether the matching degrees between the first text and other preset texts are greater than or equal to the preset matching degree threshold. In a case where the matching degrees between the first text and all preset texts in the preset attachment text library are less than the preset matching degree threshold, it can be determined that the target attachment does not have the risk of impairing consumer rights, and then the information auditing network can output the score and the reason for the target attachment not having the risk of impairing consumer rights.

[0052] Based on this, the information auditing network can determine the score of the risk of the target attachment impairing consumer rights and the reason for the risk of the target attachment impairing consumer rights by using the preset text in the preset attachment text library, in combination with the first text and the first semantic information, which is beneficial to improving the convenience of auditing whether the target attachment impairs consumer rights, and is further beneficial to improving the convenience of consumer rights protection supervision of the target attachment.

[0053] In some embodiments, a preset score weight corresponding to the preset text is determined according to a risk level of the preset text having the risk of impairing consumer rights, and a score of the risk of the target attachment impairing consumer rights is determined according to a product of the preset score weight and the preset score.

[0054] For example, different preset scores can be preset for different preset texts. The risk levels of different preset texts each having a risk of damaging consumer rights can be different. The risk levels of different preset texts each can be determined according to preset semantic information of the preset texts. For example, for a case where the preset semantic information is used to indicate that the preset text does not have risk prompt information, the preset text can have a first risk level. For a case where the preset semantic information is used to indicate that the preset text involves inducement sales information, the preset text can have a second risk level. For a case where the preset semantic information is used to indicate that the preset text has an unfair clause, the preset text can have a third risk level. Different preset texts under different risk levels can correspond to different preset score weights. For example, the risk levels include the first risk level to the third risk level, the preset score weight corresponding to the preset text under the first risk level is a first score weight, the preset score weight corresponding to the preset text under the second risk level is a second score weight, and the preset score weight corresponding to the preset text under the third risk level is a third score weight. The first risk level is lower than the second risk level, the second risk level is lower than the third risk level, and thus the first score weight is lower than the second score weight, and the second score weight is lower than the third score weight.

[0055] Taking financial insurance in a financial service scenario as an example. The first text in the target attachment related to the financial insurance, for example, includes the insurance clause of the financial insurance. In a case where the matching degree between the insurance clause and the preset text of the first risk level is greater than or equal to the preset matching degree threshold, the information auditing network can determine a score of the target attachment related to the financial insurance having a risk of damaging consumer rights according to a product of the first score weight corresponding to the first risk level and the preset score of the preset text. In a case where the matching degree between the insurance clause and the preset text of the second risk level is greater than or equal to the preset matching degree threshold, the information auditing network can determine a score of the target attachment related to the financial insurance having a risk of damaging consumer rights according to a product of the second score weight corresponding to the second risk level and the preset score of the preset text. Similarly, the information auditing network can determine a score of the target attachment related to the financial insurance having a risk of damaging consumer rights according to a product of the third score weight corresponding to the third risk level and the preset score of the preset text.

[0056] The same target attachment can include multiple first texts. Different first texts can be matched with different preset texts, and preset scores and preset score weights of preset texts respectively associated with different first texts can be different or the same, which is not limited herein. Accordingly, the information review network can count the number of first texts included in the same target attachment and matched with preset texts of the same risk level and greater than or equal to a preset matching threshold. In a case where the number is greater than or equal to a first number threshold, the preset score weight corresponding to the preset text at the risk level is adjusted; in a case where the number is less than or equal to a second number threshold, the preset score weight corresponding to the preset text at the risk level is adjusted; the second number threshold is less than the first number threshold. For example, in a case where the number is greater than or equal to the first number threshold, the information review network can infer that the target attachment frequently uses the same way to harm the consumer rights and interests, which is equivalent to an intentional behavior of harming the consumer rights and interests, and thus needs to strictly supervise the target attachment. For another example, in a case where the number is less than or equal to the second number threshold, the information review network can infer that the target attachment rarely uses the same way to harm the consumer rights and interests, which is equivalent to an unintentional behavior of harming the consumer rights and interests, and thus can appropriately relax the consumer rights and interests protection supervision on the target attachment. Based on this, by dynamically adjusting the preset score weight corresponding to the preset text at the risk level, the determination flexibility of the score of the target attachment with the risk of harming the consumer rights and interests is improved.

[0057] Based on this, in a case where the information review network determines the score of the target attachment with the risk of harming the consumer rights and interests by using the risk level of the preset text matched with the first text, determining the preset score weight corresponding thereto, and combining the preset score corresponding to the preset text with the preset score weight, the determination convenience of the score of the target attachment with the risk of harming the consumer rights and interests is improved, and thus the review convenience when reviewing whether the target attachment has the risk of harming the consumer rights and interests is improved, thereby improving the convenience of the consumer rights and interests protection supervision on the target attachment.

[0058] In some embodiments, when there is a preset attachment uploaded to the attachment review platform, if the uploading time of the preset attachment is within a preset uploading time period and the current time is within a first review time period, the preset attachment is determined as the target attachment to be reviewed.

[0059] For example, the attachment review platform can be pre-provided with corresponding review restriction information. The review restriction information can include at least one of a preset upload time period of the attachment, a first review time period, a second review time period, an attachment upload role, an attachment scoring role, and an attachment query role. The attachment upload time period can include an upload start time and an upload end time. The second review time period can include a manual review start time and a manual review end time. The first review time period can be between the preset upload time period and the second review time period, such as covering a period after the upload end time to a period before the manual review start time. The attachment upload role can be used to indicate personnel who can upload the preset attachment to the attachment review platform, i.e., personnel who have an attachment upload permission. The attachment scoring role can be used to indicate personnel who can manually review the target attachment in the attachment review platform, i.e., personnel who have an attachment review permission. The attachment query role can be used to indicate personnel who can view at least one of the preset attachment and the target attachment, i.e., personnel who have an attachment query permission. The review restriction information can be pre-set or set by the corresponding personnel, which is not limited herein.

[0060] In the case where the preset attachment is uploaded to the attachment review platform and the upload time of the preset attachment is within the preset upload time period, it can be ensured that the preset attachment is successfully uploaded to the attachment review platform, and the preset attachment can be determined as the target attachment to be reviewed when the current time is within the first review time period, so that the attachment review platform can perform automatic review. For example, the attachment review platform can perform automatic review on the target attachment by using a preset review model within the first review time period.

[0061] Of course, it is not limited to this. In the case where the preset attachment is uploaded to the attachment review platform, if the upload time of the preset attachment is within the preset upload time period and the current time is not within the first review time period, the attachment upload role can still edit and re-upload the edited preset attachment to the attachment review platform. If the upload time of the preset attachment is later than the preset upload time period, the attachment review platform does not allow the preset attachment to be uploaded to the attachment review platform, i.e., an upload failure prompt information corresponding to the preset attachment can be output to prompt that the upload time of the preset attachment is later than the preset upload time period. This is not limited herein.

[0062] Based on this, in the case where the preset attachment is determined as the target attachment according to that the upload time of the preset attachment is within the preset upload time period and the current time is within the first review time period, the attachment review platform can start automatic review on the target attachment, which is beneficial to improve the convenience of automatic review on the target attachment.

[0063] In some embodiments, when the current time is within the second review time period, the review result is adjusted in response to an adjustment operation of the preset personnel on the review result, to obtain an adjusted review result; the second review time period is later than the first review time period.

[0064] For example, in the case of determining the review result corresponding to the target attachment, the review result corresponding to the target attachment can also be manually reviewed. In the case where the preset personnel assesses the review result corresponding to the attachment score role of the target attachment and finds that the review result needs to be adjusted, the review result can be adjusted accordingly. Accordingly, the attachment review platform can adjust the review result in response to the adjustment operation of the preset personnel on the review result, to obtain an adjusted review result. The second review time period is later than the first review time period.

[0065] Based on this, the attachment review platform can adjust the review result in response to the adjustment operation of the preset personnel on the review result, which is beneficial to improve the convenience of adjusting the review result, and thus, on the basis of automatically reviewing the target attachment by using the preset review model, manually reviewing the review result corresponding to the target attachment by the preset personnel, which is beneficial to improve the convenience of reviewing whether the target attachment harms the consumer rights and interests, and improve the accuracy of reviewing whether the target attachment harms the consumer rights and interests.

[0066] In some embodiments, the preset review model is trained according to the review result, to obtain a trained preset review model.

[0067] For example, in the case of determining the adjusted review result in response to the adjustment operation of the preset personnel on the review result, the preset review model can also be trained according to the adjusted review result, to obtain a trained preset review model.

[0068] In the process of training the preset review model, at least one of the review result and the adjusted review result can be used as input data for optimizing the algorithm of the preset review model, to optimize the network parameters of at least one of the semantic understanding network and the information review network in the preset review model, and thus, to optimize at least one of the semantic recognition processing capability of the semantic understanding network and the review capability of the information review network, to improve the convenience and accuracy of reviewing whether the target attachment harms the consumer rights and interests, and thus, to improve the convenience and accuracy of consumer rights protection supervision on the target attachment.

[0069] The accessory review method provided in the application can determine the first semantic information of the first text included in the target accessory by using a preset review model and determine the review result corresponding to the target accessory, that is, the accessory review method can be implemented by means of artificial intelligence (AI), thereby reducing labor costs and improving the convenience of reviewing whether the target accessory damages the consumer rights and interests.

[0070] In an exemplary embodiment, as shown in Figure 3 The accessory review process of the accessory review platform includes: determining whether the current time is in a preset uploading time period; if yes, obtaining a target accessory; determining whether the current time is in a first review time period; if yes, determining a review result corresponding to the target accessory by using a preset review model; determining whether the current time is in a second review time period; if yes, adjusting the review result in response to an adjustment operation of a preset personnel on the review result to obtain an adjusted review result.

[0071] Therefore, the accessory review platform can enhance the daily review of the consumer rights and interests protection of the target accessory according to the corresponding accessory review process, and can leave traces of the review of whether the target accessory damages the consumer rights and interests, thereby improving the convenience of reviewing the review result.

[0072] It should be understood that the size of the serial number of each step in the above embodiments does not mean the order of execution, and the execution order of each process should be determined according to its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the application.

[0073] In an embodiment, an accessory review device is provided, which corresponds to the accessory review method in the above embodiments. As shown in Figure 4 The accessory review device includes an information acquisition module 110, a semantic recognition module 120, and an information review module 130. The functions of each functional module are described in detail as follows. The information acquisition module 110 is configured to acquire a target accessory related to consumer rights protection through the accessory review platform; the target accessory includes a plurality of first texts; The semantic recognition module 120 is configured to, when the current time is in a first review time period, perform semantic recognition processing on the first text based on a semantic understanding network of a preset review model to obtain first semantic information of the first text; The information review module 130 is configured to review the target accessory based on an information review network of the preset review model according to the first text and the first semantic information to obtain a review result corresponding to the target accessory; the review result is used to indicate a score of the target accessory having a risk of damaging consumer rights and interests and a reason of the target accessory having the risk of damaging consumer rights and interests.

[0074] In an embodiment, the attachment auditing apparatus is further configured to: determine, based on the semantic understanding network, second semantic information similar to the first semantic information according to the first semantic information, and determine second text corresponding to the second semantic information; the information auditing module 130 is configured to: audit the target attachment based on the information auditing network according to the first text, the first semantic information, the second text, and the second semantic information, to obtain an auditing result corresponding to the target attachment.

[0075] In an embodiment, the information auditing module 130 is configured to: input the first text and the first semantic information into the information auditing network, so that the information auditing network determines a matching degree between the first text and a preset text in a preset attachment text library according to the first semantic information; the preset text is used to indicate a text with a risk of harming consumer rights; when the matching degree between the first text and the preset text is greater than or equal to a preset matching degree threshold, determine a score of the target attachment with the risk of harming consumer rights according to a preset score corresponding to the preset text, and determine a reason of the target attachment with the risk of harming consumer rights according to preset semantic information corresponding to the preset text and the first semantic information.

[0076] In an embodiment, the information auditing module 130 is configured to: determine a preset score weight corresponding to the preset text according to a risk level of the risk of harming consumer rights possessed by the preset text; determine the score of the target attachment with the risk of harming consumer rights according to a product of the preset score weight and the preset score.

[0077] In an embodiment, the information obtaining module 110 is configured to: when a preset attachment is uploaded to the attachment auditing platform, if an uploading time of the preset attachment is within a preset uploading time period and a current time is within a first auditing time period, determine the preset attachment as a target attachment to be audited.

[0078] In an embodiment, the attachment auditing apparatus is configured to: when the current time is within a second auditing time period, in response to an adjustment operation of the preset personnel on the auditing result, adjust the auditing result to obtain an adjusted auditing result; the second auditing time period is later than the first auditing time period.

[0079] In an embodiment, the attachment auditing apparatus is configured to: According to the auditing result, the preset auditing model is trained to obtain a trained preset auditing model.

[0080] The specific limitations of the attachment auditing apparatus can be referred to the limitations of the attachment auditing method described above, which will not be repeated here. Each module in the above attachment auditing apparatus can be realized by software, hardware and their combination in whole or in part. The above modules can be embedded in or independent of the processor in the computer device in hardware form, or can be stored in the memory in the computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0081] In an embodiment, a computer device is provided, which can be a server, and its internal structure diagram can be as shown in Figure 5 The computer device includes a processor, a memory, a network interface and a database connected through a system bus. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile and / or volatile storage medium, an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is configured to communicate with an external client through a network connection. The computer program is executed by the processor to implement the functions or steps of the attachment auditing method on the server side.

[0082] In an embodiment, a computer device is provided, which can be a client, and its internal structure diagram can be as shown in Figure 6 The computer device includes a processor, a memory, a network interface, a display screen and an input device connected through a system bus. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is configured to communicate with an external server through a network connection. The computer program is executed by the processor to implement the functions or steps of the attachment auditing method on the client side.

[0083] In an embodiment, a computer device is provided, which includes a memory, a processor and a computer program stored in the memory and executable on the processor, and the processor implements the following steps when executing the computer program: Through the attachment auditing platform, a target attachment related to consumer rights protection is obtained; the target attachment includes a plurality of first texts; When the current time is within the first auditing time period, a semantic understanding network based on a preset auditing model is used to perform semantic recognition processing on the first text to obtain first semantic information of the first text; An information auditing network based on the preset auditing model is used to perform auditing on the target attachment according to the first text and the first semantic information to obtain an auditing result corresponding to the target attachment; the auditing result is used to indicate a score of the target attachment having a risk of damaging consumer rights and interests and a reason of the target attachment having the risk of damaging consumer rights and interests.

[0084] In one embodiment, a computer readable storage medium is provided, and the computer readable storage medium stores a computer program. The computer program is executed by a processor to implement the following steps: A target attachment related to consumer rights protection is obtained through the attachment auditing platform; the target attachment includes a plurality of first texts; When the current time is within the first auditing time period, a semantic understanding network based on a preset auditing model is used to perform semantic recognition processing on the first text to obtain first semantic information of the first text; An information auditing network based on the preset auditing model is used to perform auditing on the target attachment according to the first text and the first semantic information to obtain an auditing result corresponding to the target attachment; the auditing result is used to indicate a score of the target attachment having a risk of damaging consumer rights and interests and a reason of the target attachment having the risk of damaging consumer rights and interests.

[0085] It should be noted that the functions or steps that the computer readable storage medium or the computer device can implement correspond to the related descriptions of the server side and the client side in the foregoing method embodiments. To avoid repetition, they will not be described one by one here.

[0086] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium, and when executed, can include the processes of the above-mentioned embodiment methods. Any reference to memory, storage, database or other medium used in the embodiments provided by the present application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0087] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the above-mentioned division of functional units and modules is exemplified. In actual application, the above-mentioned functions can be completed by different functional units or modules according to needs, i.e. the internal structure of the device is divided into different functional units or modules to complete all or part of the above-mentioned functions.

[0088] The above-mentioned embodiments are only used to illustrate the technical solutions of the present application, but not to limit them. Although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacements to some technical features. These modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A method for reviewing attachments, characterized in that, The attachment review method, applied to the attachment review platform, includes: The attachment review platform is used to obtain target attachments related to consumer rights protection; the target attachments include multiple first texts. When the current time is within the first review period, the semantic understanding network based on the preset review model performs semantic recognition processing on the first text to obtain the first semantic information of the first text; Based on the information review network of the preset review model, the target attachment is reviewed according to the first text and the first semantic information to obtain the review result corresponding to the target attachment; the review result is used to indicate the score of the target attachment having the risk of harming consumer rights and the reasons why the target attachment has the risk of harming consumer rights.

2. The attachment review method according to claim 1, characterized in that, After the semantic understanding network based on the preset review model performs semantic recognition processing on the first text to obtain the first semantic information of the first text, it further includes: Based on the semantic understanding network, according to the first semantic information, second semantic information similar to the first semantic information is determined, and the second text corresponding to the second semantic information is determined; The information review network based on the preset review model reviews the target attachment according to the first text and the first semantic information, and obtains the review result corresponding to the target attachment, including: Based on the information review network, the target attachment is reviewed according to the first text, the first semantic information, the second text, and the second semantic information to obtain the review result corresponding to the target attachment.

3. The attachment review method according to claim 1, characterized in that, The information review network based on the preset review model reviews the target attachment according to the first text and the first semantic information, and obtains the review result corresponding to the target attachment, including: The first text and the first semantic information are input into the information review network so that the information review network determines the matching degree between the first text and a preset text in a preset attachment text library based on the first semantic information; the preset text is used to indicate text that has the risk of harming consumer rights. When the matching degree between the first text and the preset text is greater than or equal to the preset matching degree threshold, the target attachment is determined to have a risk of harming consumer rights based on the preset score corresponding to the preset text, and the reason why the target attachment has a risk of harming consumer rights is determined based on the preset semantic information corresponding to the first semantic information and the preset semantic information corresponding to the preset text.

4. The attachment review method according to claim 3, characterized in that, The step of determining the risk of the target attachment harming consumer rights based on the preset score corresponding to the preset text includes: Based on the risk level of the preset text that it poses to harm consumer rights, the preset scoring weight corresponding to the preset text is determined; The score indicating whether the target accessory poses a risk of harming consumer rights is determined by multiplying the preset scoring weight with the preset score.

5. The method for reviewing attachments according to any one of claims 1 to 4, characterized in that, The step of obtaining the target attachment to be reviewed through the attachment review platform includes: When a preset attachment is uploaded to the attachment review platform, if the upload time of the preset attachment is within a preset upload time period and the current time is within the first review time period, then the preset attachment is determined to be the target attachment to be reviewed.

6. The attachment review method according to claim 5, characterized in that, In the information review network based on the preset review model, after reviewing the target attachment according to the first text and the first semantic information to obtain the review result corresponding to the target attachment, the method further includes: When the current time falls within the second review period, in response to the preset personnel's adjustment operation on the review result, the review result is adjusted to obtain the adjusted review result; the second review period is later than the first review period.

7. The method for reviewing attachments according to any one of claims 1 to 4, characterized in that, The auditing method also includes: Based on the audit results, the preset audit model is trained to obtain the trained preset audit model.

8. An attachment review device, characterized in that, The attachment review device includes: The information acquisition module is used to acquire target attachments related to consumer rights protection through the attachment review platform; the target attachments include multiple first texts; The semantic recognition module is used to perform semantic recognition processing on the first text based on the semantic understanding network of the preset review model when the current time is within the first review time period, so as to obtain the first semantic information of the first text. The information review module is used to review the target attachment based on the information review network of the preset review model, according to the first text and the first semantic information, and obtain the review result corresponding to the target attachment; the review result is used to indicate the score of the target attachment having the risk of harming consumer rights and the reason why the target attachment has the risk of harming consumer rights.

9. A computer device, characterized in that, The computer device includes a memory and a processor; The memory is used to store computer programs; The processor is configured to execute the computer program and, in executing the computer program, implement the appendix review method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the attachment review method as described in any one of claims 1 to 7.

Images