Large model content auditing method and device and storage medium

Abstract

The invention discloses a content auditing method and device for a large model, and relates to the field of artificial intelligence, in particular to the technical field of data processing and the like. According to the specific implementation scheme, the method comprises the steps that a Serverless service component obtains first page content needing to be audited; performing first auditing on the first page content based on the service component set of the large model to obtain a first auditing result; the Serverless service component obtains second page content needing to be audited based on the first audit result; and performing second auditing on the second page content based on the service component set of the large model to obtain a second auditing result. According to the method and the device, various scenes in which the user uses the social software are simulated, so that the audited content is more comprehensive, and the first page content and the second page content which need to be audited are correspondingly audited through the service component set of the large model, so that the sensitive content which is deliberately avoided is audited, and the accuracy of the audited result is improved.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence, specifically to technical fields such as data processing, and in particular to a large-model content review method, device and storage medium. Background Art

[0002] With the continuous development of mobile Internet, social software has become an important way for users to understand current events and socialize, and has been widely used. Among them, there may be sensitive content in the pages of social software. Based on this, it is necessary to review the content of the pages of social software in order to monitor sensitive content in time and avoid further spread.

[0003] In the prior art, the content of the page can be detected after parsing the page of the social software. However, the above prior art cannot completely simulate the scene of users using the social software, making the audit content incomplete and reducing the accuracy of the audit results. Summary of the invention

[0004] The present invention provides a content review method, device and storage medium for a large model.

[0005] According to one aspect of the present disclosure, a content review method of a large model is provided, the method comprising: a serverless service component obtains a first page content that needs to be reviewed; a service component set based on the large model performs a first review on the first page content to obtain a first review result; the Serverless service component obtains a second page content that needs to be reviewed based on the first review result; a service component set based on the large model performs a second review on the second page content to obtain a second review result.

[0006] Optionally, in an embodiment of the present disclosure, the Serverless service component obtains the first page content that needs to be reviewed, including:

[0007] The Serverless service component obtains the first page of accessing the social software;

[0008] The Serverless service component accesses the content in the first page to obtain a corresponding preset number of content result pages;

[0009] The Serverless service component obtains the first page content that needs to be reviewed based on the content result page.

[0010] Optionally, in an embodiment of the present disclosure, the Serverless service component accesses the content in the first page to obtain a corresponding preset number of content result pages, including:

[0011] The serverless service component searches based on the keywords in the first page to obtain a corresponding preset number of content result pages; or

[0012] The Serverless service component accesses the content in the first page based on the ranking, and obtains a corresponding preset number of content result pages.

[0013] Optionally, in an embodiment of the present disclosure, the Serverless service component obtains the first page content that needs to be reviewed based on the content result page, including:

[0014] The Serverless service component determines target comment content that needs to be reviewed in the content result page;

[0015] The Serverless service component takes a screenshot of the target comment content to obtain a first screenshot result;

[0016] The Serverless service component determines the first screenshot result as the first page content that needs to be reviewed.

[0017] Optionally, in one embodiment of the present disclosure, the service component set of the large model includes a jump detection service component, a first model service component, a word vector generation component, a retrieval component, and a second model service component;

[0018] The service component set based on the big model performs a first audit on the first page content to obtain a first audit result, including:

[0019] The jump detection service component obtains, based on the first page content, a target similar text that satisfies a similarity condition with the first page content and a first text corresponding to the first page content through the first model service component, the word vector generation component and the retrieval component;

[0020] The retrieval component concatenates the first text with the target similar text, and inputs the concatenated text into the second model service component to obtain a first review result.

[0021] Optionally, in an embodiment of the present disclosure, the jump detection service component obtains, based on the first page content, a target similar text that satisfies a similarity condition with the first page content and a first text corresponding to the first page content through the first model service component, the word vector generation component, and the retrieval component, including:

[0022] The jump detection service component inputs the first page content into the first model service component to obtain a first text corresponding to the first page content;

[0023] The jump detection service component calls the word vector generation component to perform vector conversion on the first text to obtain a corresponding first vector;

[0024] The jump detection service component calls the retrieval component, calculates the similarity between the first vector and the vectors in the historical corpus, and determines the target similar text that meets the similarity condition with the first page content based on the obtained similarity result.

[0025] Optionally, in an embodiment of the present disclosure, the first audit result includes whether a potential jump exists or whether a potential jump does not exist; when the first audit result includes whether a potential jump exists, the Serverless service component obtains the second page content that needs to be audited based on the first audit result, including:

[0026] The Serverless service component determines a target account with a potential redirection situation based on the first audit result;

[0027] The Serverless service component accesses a second page corresponding to the target account, and takes a screenshot of the second page to obtain a second screenshot result;

[0028] The second screenshot result is determined as the second page content that needs to be reviewed.

[0029] Optionally, in one embodiment of the present disclosure, the service component of the large model further includes an identification detection service component, a webpage detection service component, and a third model service component;

[0030] The service component set based on the large model performs a second audit on the second page content to obtain a second audit result, including:

[0031] The jump detection service component inputs the second page content and the second text of the target account in the first page content into the identification detection service component to obtain the jump target information;

[0032] The webpage detection service component calls the Serverless service component to access based on the target information, and obtains a third page corresponding to the target information;

[0033] The Serverless service component accesses and takes a screenshot of the third page to obtain a third screenshot result that needs to be reviewed;

[0034] The web page detection service component inputs the third screenshot result into the third model service component to obtain a second review result.

[0035] Optionally, in one embodiment of the present disclosure, the identification and detection service component further includes a website identification service subcomponent and a group number identification service subcomponent;

[0036] The jump detection service component inputs the second page content and the second text of the target account in the first page content into the identification detection service component to obtain the jump target information, including:

[0037] The jump detection service component inputs the second page content and the second text into the website identification service subcomponent and the group number identification service subcomponent respectively, and obtains the website or social account where the jump exists;

[0038] The URL or social account where the jump exists is determined as the jump target information.

[0039] Optionally, in one embodiment of the present disclosure, the service component of the large model further includes a learning service component and a fourth model service component;

[0040] The second audit result includes the presence of sensitive content or the absence of sensitive content; when the second audit result includes the presence of sensitive content, the method further includes:

[0041] The learning service component calls the word vector generation component to convert the third text containing the sensitive content into a second vector;

[0042] The learning service component calls the retrieval component to calculate the similarity between the second vector and the vector in the historical corpus to obtain the corresponding prompt word text;

[0043] The learning service component concatenates the prompt word text with the third text, and inputs the concatenated text into the fourth model service component to obtain a target question text;

[0044] The learning service component calls the word vector generation component to convert the target question text into a third vector, and stores the third vector and the third text in the historical corpus.

[0045] According to another aspect of the present disclosure, a content review device of a large model is provided, including: a first acquisition module, used for a serverless service component to obtain a first page content that needs to be reviewed; a first review module, used for performing a first review on the first page content based on a service component set of the large model, and obtaining a first review result; a second acquisition module, used for the Serverless service component to obtain a second page content that needs to be reviewed based on the first review result; a second review module, used for performing a second review on the second page content based on the service component set of the large model, and obtaining a second review result.

[0046] According to another aspect of the present disclosure, an electronic device is provided, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the content review method of the large model of the present disclosure.

[0047] According to another aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to cause the computer to execute the content review method of the large model of the present disclosure.

[0048] According to another aspect of the present disclosure, a computer program product is provided, which implements the content review method of the large model of the present disclosure when the computer program is executed by a processor.

[0049] Compared with the prior art, the present invention has the following beneficial effects:

[0050] The content review method of the big model can obtain the first page content and the second page content that need to be reviewed through the Serverless service component, thereby simulating various scenarios in which users use social software, making the review content more comprehensive. At the same time, through the service component collection of the big model, the first page content and the second page content that need to be reviewed are reviewed accordingly, and sensitive content that is deliberately avoided can be reviewed, thereby improving the accuracy of the review results.

[0051] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] The accompanying drawings are used to better understand the present solution and do not constitute a limitation of the present disclosure.

[0053] Figure 1It is a flowchart of a large-model content review method provided according to an embodiment of the present disclosure;

[0054] Figure 2 It is a flowchart of a large-model content review method provided according to an embodiment of the present disclosure;

[0055] Figure 3 is a schematic diagram of a service component set of a large model provided according to an embodiment of the present disclosure;

[0056] Figure 4 It is a flowchart of a large-model content review method proposed according to an embodiment of the present disclosure;

[0057] Figure 5 It is a flowchart of a large-model content review method proposed according to an embodiment of the present disclosure;

[0058] Figure 6 It is a flowchart of a large-model content review method proposed according to an embodiment of the present disclosure;

[0059] Figure 7 It is a flowchart of a large-model content review method proposed according to an embodiment of the present disclosure;

[0060] Figure 8 It is a structural diagram of a large-scale content review device provided according to an embodiment of the present disclosure;

[0061] Fig. 9 is a block diagram of an electronic device for implementing an embodiment of the present disclosure. DETAILED DESCRIPTION

[0062] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0063] The following describes the content review method and device of the large model of the embodiment of the present disclosure with reference to the accompanying drawings.

[0064] Artificial Intelligence (AI) is a discipline that studies how computers can simulate certain thought processes and intelligent behaviors of human beings (such as learning, reasoning, thinking, planning, etc.). It includes both hardware-level technologies and software-level technologies. AI hardware technologies generally include computer vision technology, speech recognition technology, natural language processing technology, as well as learning / deep learning, big data processing technology, knowledge graph technology, and other aspects.

[0065] Natural Language Processing (NLP) is an important field in the fields of computer science and artificial intelligence. It studies various theories and methods that can achieve effective communication between humans and computers using natural language. Natural language processing is a science that integrates linguistics, computer science, and mathematics. Natural language processing is mainly used in machine translation, public opinion monitoring, automatic summarization, opinion extraction, text classification, question answering, text semantic comparison, speech recognition, etc.

[0066] Deep Learning (DL) is a new research direction in the field of Machine Learning (ML). It is introduced into machine learning to make it closer to its original goal - artificial intelligence. Deep learning is the internal laws and representation levels of learning sample data. The information obtained in the learning process is of great help in the interpretation of data such as text, images and sounds. Its ultimate goal is to enable machines to have analytical learning capabilities like humans and to recognize data such as text, images and sounds. Deep learning is a complex machine learning algorithm that has achieved results in speech and image recognition that far exceed previous related technologies.

[0067] Large models refer to models with a large number of parameters and complex structures in the field of machine learning and artificial intelligence. These models usually require huge computing resources to train and deploy, and are able to process and understand more and more complex data. Large models can be applied to many application areas such as automatic writing, chatbots, virtual assistants, voice assistants, automatic translation, etc.

[0068] Figure 1 is a schematic diagram according to the first embodiment of the present disclosure.

[0069] like Figure 1 As shown, the content review method of the large model may include:

[0070] In step 101, the serverless service component obtains the first page content that needs to be reviewed.

[0071] It should be noted that the executor of the above-mentioned large model content review method is the large model content review device, which can be implemented by software and / or hardware. The large model content review device in this embodiment can be configured in an electronic device.

[0072] In this example embodiment, the electronic device may include devices such as terminal devices and servers, and this embodiment does not limit the electronic device.

[0073] In the embodiments of the present disclosure, serverless is a model of cloud computing that enables deployment and operation of applications without the need to manage servers. Serverless service components are tools and frameworks used to build and manage serverless applications.

[0074] Among them, in the embodiment of the present disclosure, when the Serverless service component receives the review instruction, the Serverless service component can simulate various scenarios of users accessing social software (such as clicking or querying) through a simulator to obtain the page content that needs to be reviewed.

[0075] Specifically, in an embodiment of the present disclosure, the method for the above-mentioned Serverless service component to obtain the first page content that needs to be reviewed may include: the Serverless service component obtains the first page of accessing the social software; the Serverless service component accesses the content in the first page to obtain a corresponding preset number of content result pages; the Serverless service component obtains the first page content that needs to be reviewed based on the content result page.

[0076] Among them, in the embodiment of the present disclosure, the above-mentioned first page can be a default hot search page or a home screen recommendation page when accessing social software.

[0077] And, in the embodiment of the present disclosure, after the Serverless service component obtains the first page, it can access the content in the first page to obtain a preset number of content result pages. In the embodiment of the present disclosure, the preset number can be set as needed, for example, the preset number is 3, that is, the content in the first page is accessed, and the first 3 pages in the access results are determined as content result pages.

[0078] Furthermore, in the embodiment of the present disclosure, after the Serverless service component obtains the content result page, it can determine the target comment content that needs to be reviewed through the content result page, and obtain the first page content that needs to be reviewed based on the target comment content.

[0079] Step 102: Perform a first audit on the first page content based on the service component set of the large model to obtain a first audit result.

[0080] In the disclosed embodiment, Figure 2 A schematic diagram of a service component set of a large model provided by an embodiment of the present disclosure. Figure 2 As shown, the service component set of the large model may include: Serverless service component (barrier-free service), jump detection service component (jump detection service), first model service component (first model service), word vector generation component (embedding service), retrieval component (bes service), second model service component (second model service), recognition detection service component (recognition detection service), web page detection service component (web page detection service), third model service component (third model service), learning service component (learning service) and fourth model service component (fourth model service). Among them, in the embodiment of the present disclosure, the recognition detection service component may include a website recognition service subcomponent (website recognition service) and a group number recognition service subcomponent (group number recognition service).

[0081] Among them, in the embodiment of the present disclosure, the first model service component, the second model service component, the third model service component and the fourth model service component are all large models that have been trained. And, in the real-time example of the present disclosure, the large models corresponding to the first model service component, the second model service component, the third model service component and the fourth model service component can be the same or different.

[0082] Furthermore, in the embodiment of the present disclosure, after obtaining the first page content that needs to be reviewed through the above steps, a first review can be performed on the first page content based on the jump detection service component, the first model service component, the word vector generation component, the retrieval component and the second model service component in the service component set of the large model to obtain a first review result.

[0083] Further, in the embodiment of the present disclosure, the first audit result may include the existence of a potential redirection or the absence of a potential redirection. Wherein, when the first audit result includes the existence of a potential redirection, the first audit result also includes the target account of the potential redirection.

[0084] Step 103: The Serverless service component obtains the second page content that needs to be reviewed based on the first review result.

[0085] In the disclosed embodiment, after the first audit result is obtained through the above steps and the first audit result includes the existence of a potential jump situation, the Serverless service component can obtain the second page content that needs to be audited based on the first audit result.

[0086] In the disclosed embodiment, the method in which the Serverless service component obtains the second page content that needs to be reviewed based on the first review result may include: the Serverless service component determines the target account with potential jump based on the first review result; the Serverless service component accesses the second page corresponding to the target account, and takes a screenshot of the second page to obtain a second screenshot result; and determines the second screenshot result as the second page content that needs to be reviewed.

[0087] Among them, in the embodiment of the present disclosure, the second page corresponding to the above-mentioned target account can be the social homepage corresponding to the target account, so that the social software can be comprehensively reviewed based on the second page content and the first page content.

[0088] Step 104: Perform a second audit on the second page content based on the service component set of the large model to obtain a second audit result.

[0089] In the disclosed embodiment, after obtaining the second page content through the above steps, a second audit can be performed on the second page content based on the service component set of the large model to obtain a second audit result.

[0090] Among them, in the embodiment of the present disclosure, the second page content can be secondly audited by the identification and detection service component, the web page detection service component and the third model service component in the service component set of the large model to obtain a second audit result.

[0091] And, in the embodiment of the present disclosure, the second audit result may include the presence of sensitive content or the absence of sensitive content. In the embodiment of the present disclosure, if the second audit result includes the presence of sensitive content, the original comment content corresponding to the sensitive content may be stored so that the audit result of the comment content can be directly obtained later; if the second audit result includes the absence of sensitive content, the audit of the original comment content is completed, and the above steps are repeated to audit other page contents of the social software.

[0092] In the disclosed embodiment, the content review method of the big model can obtain the first page content and the second page content that need to be reviewed through the Serverless service component, thereby simulating various scenarios in which users use social software, making the reviewed content more comprehensive. At the same time, the first page content and the second page content that need to be reviewed are reviewed accordingly through the service component set of the big model, so that sensitive content that is deliberately avoided can be reviewed, thereby improving the accuracy of the review results.

[0093] Figure 3 is a schematic diagram according to a second embodiment of the present disclosure.

[0094] like Figure 3As shown, the content review method of the large model may include:

[0095] Step 301: The Serverless service component obtains the first page of accessing the social software.

[0096] In the disclosed embodiment, the Serverless service component can simulate a user's click or query operation on a social software through a simulator to access the social software and obtain the first page of accessing the social software.

[0097] Among them, in the embodiment of the present disclosure, the above-mentioned first page can be a hot search page or a home screen recommendation page.

[0098] Step 302: The Serverless service component accesses the content in the first page and obtains a corresponding preset number of content result pages.

[0099] In the disclosed embodiment, after the Serverless service component obtains the first page through the above steps, it can access the content in the first page to obtain a corresponding preset number of content result pages.

[0100] Specifically, in an embodiment of the present disclosure, the method in which the above-mentioned Serverless service component accesses the content in the first page and obtains a corresponding preset number of content result pages may include: the Serverless service component searches based on keywords in the first page to obtain a corresponding preset number of content result pages; or, the Serverless service component accesses based on the ranking of the content in the first page to obtain a corresponding preset number of content result pages.

[0101] Among them, in the embodiment of the present disclosure, if the first page is a hot search page, the Serverless service component can search based on the keywords in the first page to obtain a preset number of content result pages with high rankings.

[0102] Furthermore, in the embodiment of the present disclosure, if the first page is the home page, the Serverless service component can access the content in the first page to obtain a preset number of content result pages with high rankings.

[0103] Furthermore, in the embodiment of the present disclosure, the above preset number can be set as needed, for example, the preset number is 3.

[0104] In the disclosed embodiment, the above steps can be used to simulate various scenarios of users using social software, and obtain content result pages corresponding to more comments or higher popularity. Therefore, comprehensive target comment content can be obtained based on the content result page, making the content that needs to be reviewed later more comprehensive.

[0105] Step 303: The Serverless service component obtains the first page content that needs to be reviewed based on the content result page.

[0106] In the disclosed embodiment, after obtaining the content result page through the above steps, the Serverless service component can obtain the first page content that needs to be reviewed based on the content result page. In the disclosed embodiment, the first page content can be the comment content that needs to be reviewed in the content result page.

[0107] Specifically, in the embodiment of the present disclosure, the method for the Serverless service component to obtain the first page content that needs to be reviewed based on the content result page may include the following steps:

[0108] Step 3031: The Serverless service component determines the target comment content that needs to be reviewed in the content result page;

[0109] Step 3032: The Serverless service component takes a screenshot of the target comment content to obtain a first screenshot result;

[0110] Step 3033: The Serverless service component determines the first screenshot result as the first page content that needs to be reviewed.

[0111] Among them, in the embodiment of the present disclosure, the Serverless service component can determine the first threshold number of comment contents with the highest number of likes in the content result page and / or the first threshold number of comment contents with the highest default ranking as target comment contents that need to be reviewed. And, in the embodiment of the present disclosure, the above-mentioned first threshold number can be set manually, for example, the first threshold number is 10, that is, the top 10 comment contents with the highest number of likes in the content result page and / or the top 10 comment contents with the highest default ranking can be determined as target comment contents that need to be reviewed.

[0112] Furthermore, in the disclosed embodiment, after determining the target comment content through the above steps, the Serverless service component expands and takes a screenshot of the target comment content to obtain a first screenshot result, and determines the first screenshot result as the first page content that needs to be reviewed, so that the first page content can be subsequently reviewed based on a set of service components of a large model to determine whether there is a potential jump in the first page content, and then review sensitive content that is deliberately avoided.

[0113] Step 304: Perform a first audit on the first page content based on the service component set of the large model to obtain a first audit result.

[0114] Step 305: The Serverless service component obtains the second page content that needs to be reviewed based on the first review result.

[0115] Step 306: Perform a second audit on the second page content based on the service component set of the large model to obtain a second audit result.

[0116] For the specific implementation of steps 304 to 306, please refer to the relevant description of the above embodiment, which will not be repeated here.

[0117] In the disclosed embodiment, the first page content that needs to be reviewed is obtained through the Serverless service component, thereby simulating various scenarios in which users use social software, so that the content to be reviewed later is more comprehensive.

[0118] Figure 4 is a schematic diagram according to a third embodiment of the present disclosure.

[0119] like Figure 4 As shown, the content review method of the large model may include:

[0120] Step 401: The Serverless service component obtains the first page content that needs to be reviewed.

[0121] For the specific implementation of step 401, please refer to the relevant description of the above embodiment, which will not be repeated here.

[0122] Step 402: Based on the first page content, the jump detection service component obtains the target similar text that meets the similarity condition with the first page content and the first text corresponding to the first page content through the first model service component, the word vector generation component and the retrieval component.

[0123] In the embodiment of the present disclosure, after obtaining the first page content through the above steps, the jump detection service component can obtain the target similar text that meets the similarity conditions with the first page content and the first text corresponding to the first page content based on the first page content through the first model service component, the word vector generation component and the retrieval component.

[0124] Specifically, in the embodiment of the present disclosure, the jump detection service component obtains the target similar text that satisfies the similarity condition with the first page content and the first text corresponding to the first page content through the first model service component, the word vector generation component and the retrieval component based on the first page content, which may include the following steps:

[0125] Step 4021: The jump detection service component inputs the first page content into the first model service component to obtain a first text corresponding to the first page content;

[0126] Step 4022: The jump detection service component calls the word vector generation component to perform vector conversion on the text corresponding to the first page content to obtain a corresponding first vector;

[0127] Step 4023: The jump detection service component calls the retrieval component to calculate the similarity between the first vector and the vectors in the historical corpus, and determines the target similar text that meets the similarity condition with the first page content based on the obtained similarity result.

[0128] In the embodiment of the present disclosure, the jump detection service component can input the first page content into the first model service component to obtain the first text corresponding to the first page content. In the embodiment of the present disclosure, the first text can correspond to multiple comment contents, and the first text can be in json format.

[0129] Furthermore, in the embodiment of the present disclosure, after the jump detection service component obtains the first text, it can call the word vector generation component to perform vector conversion on the first text to obtain the corresponding first vector.

[0130] Furthermore, in the embodiment of the present disclosure, after the jump detection service component obtains the first vector, it can call the retrieval component to calculate the similarity between the first vector and the vectors in the historical corpus to obtain a similarity result, and determine the target similar text that meets the similarity conditions with the first page content based on the similarity result.

[0131] In the embodiment of the present disclosure, the method for determining the target similar text that meets the similarity condition with the first page content based on the similarity result may include: determining the top-ranked second threshold number of texts whose similarity exceeds the similarity threshold in the similarity result as the target similar text that meets the similarity condition with the first page content. In the embodiment of the present disclosure, the similarity threshold can be set as needed, for example, the similarity threshold is 0.5; the second threshold number can be set as needed, for example, the second threshold number is 5.

[0132] And, in the embodiment of the present disclosure, each comment content in the first text can be processed through the above steps to obtain the target similar text corresponding to each comment content. In the embodiment of the present disclosure, if there is no target similar text that meets the similarity condition with the corresponding comment content in the similarity result, the review of the comment content is terminated.

[0133] In the embodiment of the present disclosure, the above historical corpus can be stored using BES, and can include question text vectors, original text and question text. In the embodiment of the present disclosure, the original text and question text can be in JSON format.

[0134] In the disclosed embodiment, the above steps may be used to obtain target similar text that meets similar conditions with the first page content, thereby making the subsequent review result of the first page content more accurate.

[0135] Step 403: The search component concatenates the first text with the target similar text, and inputs the concatenated text into the second model service component to obtain a first review result.

[0136] In the disclosed embodiment, users can comment on social accounts or split the URLs in the comment area of ​​the page, so that the visiting users can jump to the social accounts or URLs in the comments to deliberately avoid the review of the comment content. For example, comment 1 is "+v34234232", and comment 2 includes {"Enter URL www.__.com", "aasdfsfas"}.

[0137] Furthermore, in the embodiment of the present disclosure, after obtaining the first text and the target similar text through the above steps, the comment content in the first text can be spliced ​​with the corresponding target similar text through the retrieval component, and the spliced ​​text is input into the second model service component to obtain the first review result corresponding to the comment content. In the embodiment of the present disclosure, the above second model service component is trained and can identify the comment content that evades review and has potential jumps.

[0138] Furthermore, in the embodiment of the present disclosure, the first audit result may include the existence of a potential redirection or the absence of a potential redirection. When the first audit result includes the existence of a potential redirection, the first audit result also includes the target account of the potential redirection. In the embodiment of the present disclosure, the target account may be the account corresponding to the comment content of the redirection.

[0139] Step 404: The Serverless service component obtains the second page content that needs to be reviewed based on the first review result.

[0140] Step 405: Perform a second audit on the second page content based on the service component set of the large model to obtain a second audit result.

[0141] Among them, regarding the specific implementation of steps 404 to 405, please refer to the relevant description of the above embodiment, which will not be repeated here.

[0142] In the disclosed embodiment, the first page content that needs to be reviewed is reviewed through the first model service component, the word vector generation component, the retrieval component and the second model service component, so as to detect the comment content with potential jumps to deliberately avoid review, so that the content to be reviewed subsequently is more comprehensive, and the accuracy of the review results is improved.

[0143] Figure 5 is a schematic diagram according to a fourth embodiment of the present disclosure.

[0144] like Figure 5 As shown, the content review method of the large model may include:

[0145] Step 501: The Serverless service component obtains the first page content that needs to be reviewed.

[0146] Step 502: Perform a first audit on the first page content based on the service component set of the large model to obtain a first audit result.

[0147] For the specific implementation of steps 501 to 502, please refer to the relevant description of the above embodiment, which will not be repeated here.

[0148] Step 503: When the first audit result includes the existence of a potential jump situation, the Serverless service component determines the target account where the potential jump situation exists based on the first audit result.

[0149] In the embodiment of the present disclosure, when the first audit result includes the existence of a potential jump situation, the Serverless service component can determine the target account where the potential jump situation exists based on the first audit result.

[0150] Step 504: The Serverless service component accesses the second page corresponding to the target account, and takes a screenshot of the second page to obtain a second screenshot result.

[0151] In the disclosed embodiment, the Serverless service component can simulate a user accessing a second page corresponding to a target account through a simulator, and take a screenshot of the second page to obtain a second screenshot result, so as to determine whether there are other contents with potential jump situations in the target account with potential jump situations based on the second page, thereby making the content to be reviewed later more comprehensive.

[0152] Step 505: determine the second screenshot result as the second page content that needs to be reviewed.

[0153] Step 506: Perform a second audit on the second page content based on the service component set of the large model to obtain a second audit result.

[0154] In the disclosed embodiment, the Serverless service component can determine the second page content that needs to be reviewed based on the first review result, so that it can be subsequently determined whether there are other potential jump content in the target account with potential jump situations based on the second page content, thereby making the content to be reviewed more comprehensive.

[0155] Figure 6 is a schematic diagram according to a fifth embodiment of the present disclosure.

[0156] like Figure 6 As shown, the content review method of the large model may include:

[0157] Step 601: The Serverless service component obtains the first page content that needs to be reviewed.

[0158] Step 602: Perform a first audit on the first page content based on the service component set of the large model to obtain a first audit result.

[0159] Step 603: The Serverless service component obtains the second page content that needs to be reviewed based on the first review result.

[0160] Step 604: The jump detection service component inputs the second page content and the second text of the target account in the first page content into the recognition detection service component to obtain the jump target information.

[0161] In the embodiment of the present disclosure, after obtaining the second page content through the above steps, the second page content and the second text of the target account in the first page content can be input into the identification detection service component through the jump detection service component to obtain the jump target information, so that the jump target address can be accessed subsequently according to the target information.

[0162] Specifically, in the embodiment of the present disclosure, the jump detection service component inputs the second page content and the second text of the target account in the first page content into the recognition detection service component, and the method for obtaining the jump target information may include the following steps:

[0163] Step 6041: The jump detection service component inputs the second page content and the second text into the website identification service subcomponent and the group number identification service subcomponent respectively, and obtains the website or social account where the jump exists;

[0164] Step 6042: determine the URL or social account where the jump exists as the jump target information.

[0165] Among them, in the embodiment of the present disclosure, the above-mentioned social account can be a social group number to guide users to join the group.

[0166] In the disclosed embodiment, the jump detection service component inputs the second page content and the second text into the URL identification service subcomponent and the group number identification service subcomponent respectively, so that the URL identification service subcomponent and the group number identification service subcomponent can obtain all URLs and / or all social accounts where jumps exist through the second page content and the second text, thereby making the content to be reviewed more comprehensive.

[0167] Step 605: The web page detection service component calls the Serverless service component to access the target information and obtains a third page corresponding to the target information.

[0168] In the disclosed embodiment, the Serverless service component can access the URL and / or social account in the target information through a simulator to obtain a third page corresponding to the target information.

[0169] Step 606: The Serverless service component accesses and takes a screenshot of the third page to obtain a third screenshot result that needs to be reviewed.

[0170] In the disclosed embodiment, the Serverless service component can access and take screenshots of the third page to obtain a third threshold number of third screenshot results that need to be reviewed. The third threshold number can be set as needed, for example, the third threshold number is 5.

[0171] Step 607: The webpage detection service component inputs the third screenshot result into the third model service component to obtain a second review result.

[0172] In the disclosed embodiment, the second audit result may include the presence of sensitive content or the absence of sensitive content. In the disclosed embodiment, if the second audit result includes the presence of sensitive content, the original comment text corresponding to the sensitive content may be stored so that the audit result of the content can be directly obtained later; if the second audit result includes the absence of sensitive content, the audit of the original comment content is completed, and the above steps are repeated to audit other page contents of the social software.

[0173] In the disclosed embodiment, a large model service component set can be used to audit all URLs and / or all social accounts that have jumps in the second page content that needs to be audited, thereby auditing deliberately avoided sensitive content and improving the accuracy of the audit results.

[0174] Figure 7 is a schematic diagram according to a sixth embodiment of the present disclosure.

[0175] like Figure 7 As shown, the content review method of the large model may include:

[0176] Step 701: The Serverless service component obtains the first page content that needs to be reviewed.

[0177] Step 702: Perform a first audit on the first page content based on the service component set of the large model to obtain a first audit result.

[0178] Step 703: The Serverless service component obtains the second page content that needs to be reviewed based on the first review result.

[0179] Step 704: Perform a second audit on the second page content based on the service component set of the large model to obtain a second audit result.

[0180] Step 705: When the second audit result includes the presence of sensitive content, the learning service component calls the word vector generation component to convert the third text containing the sensitive content into a second vector.

[0181] In the embodiment of the present disclosure, the third text containing sensitive content may be a comment of a social account containing sensitive content in the first page content.

[0182] Step 706: The learning service component calls the retrieval component to calculate the similarity between the second vector and the vectors in the historical corpus to obtain the corresponding prompt word text.

[0183] In the disclosed embodiment, the learning service component calls the retrieval component to calculate the similarity between the second vector and the vectors in the historical corpus, and concatenates the second threshold number of original texts and question texts that rank high in the similarity results to obtain the corresponding prompt word text.

[0184] Step 707: The learning service component concatenates the prompt word text with the third text, and inputs the concatenated text into the fourth model service component to obtain the target question text.

[0185] Step 708: The learning service component calls the word vector generation component to convert the target question text into a third vector, and stores the third vector and the third text in the historical corpus.

[0186] In the disclosed embodiment, the json form of the target question text, the third vector corresponding to the target question text and the json form of the third text can be stored in the historical corpus, so that the corresponding audit results can be obtained later according to the historical corpus, thereby improving the audit efficiency.

[0187] Figure 8 is a schematic diagram according to a seventh embodiment of the present disclosure.

[0188] like Figure 8 As shown, the content audit device 800 of the large model may include: a first acquisition module 801, a first audit module 802, a second acquisition module 803, and a second audit module 804, wherein:

[0189] The first acquisition module 801 is used for the serverless service component to acquire the first page content that needs to be reviewed;

[0190] A first review module 802, configured to perform a first review on the first page content based on the service component set of the large model to obtain a first review result;

[0191] The second acquisition module 803 is used for the Serverless service component to acquire the second page content that needs to be reviewed based on the first review result;

[0192] The second audit module 804 is used to perform a second audit on the second page content based on the service component set of the large model to obtain a second audit result.

[0193] It should be noted that the aforementioned explanation of the implementation example of the content review method for a large model is also applicable to this embodiment, and this embodiment will not elaborate on it.

[0194] The content review device of the big model of the embodiment of the present disclosure can obtain the first page content and the second page content that need to be reviewed through the Serverless service component, so as to simulate various scenarios of users using social software, making the reviewed content more comprehensive. At the same time, the first page content and the second page content that need to be reviewed are reviewed accordingly through the service component set of the big model, so that sensitive content that is deliberately avoided can be reviewed, thereby improving the accuracy of the review results.

[0195] In one embodiment of the present disclosure, the first acquisition module 801 is specifically used to:

[0196] The Serverless service component obtains the first page of the social software visited;

[0197] The Serverless service component accesses the content in the first page and obtains a corresponding preset number of content result pages;

[0198] The Serverless service component obtains the first page content that needs to be reviewed based on the content result page.

[0199] In one embodiment of the present disclosure, the first acquisition module 801 is further configured to:

[0200] The serverless service component searches based on the keywords in the first page to obtain a corresponding preset number of content result pages; or

[0201] The Serverless service component accesses the content based on the ranking of the first page and obtains a corresponding preset number of content result pages.

[0202] In one embodiment of the present disclosure, the first acquisition module 801 is further configured to:

[0203] The serverless service component determines the target comment content that needs to be reviewed in the content result page;

[0204] The Serverless service component takes a screenshot of the target comment content and obtains the first screenshot result;

[0205] The Serverless service component determines the first screenshot result as the first page content that needs to be reviewed.

[0206] In one embodiment of the present disclosure, the service component set of the large model includes a jump detection service component, a first model service component, a word vector generation component, a retrieval component, and a second model service component; the first audit module 802 is specifically used to:

[0207] The jump detection service component obtains, based on the first page content, a target similar text that satisfies a similarity condition with the first page content and a first text corresponding to the first page content through a first model service component, a word vector generation component, and a retrieval component;

[0208] The retrieval component concatenates the first text with the target similar text, and inputs the concatenated text into the second model service component to obtain a first review result.

[0209] In one embodiment of the present disclosure, the first audit module 802 is specifically configured to:

[0210] The jump detection service component inputs the first page content into the first model service component to obtain a first text corresponding to the first page content;

[0211] The jump detection service component calls the word vector generation component to perform vector conversion on the first text to obtain a corresponding first vector;

[0212] The jump detection service component calls the retrieval component to calculate the similarity between the first vector and the vectors in the historical corpus, and determines the target similar text that meets the similarity condition with the first page content based on the obtained similarity result.

[0213] In one embodiment of the present disclosure, the first audit result includes the existence of a potential jump situation or the absence of a potential jump situation; when the first audit result includes the existence of a potential jump situation, the second acquisition module 803 is specifically used to:

[0214] The Serverless service component determines the target account with potential redirection based on the first audit result;

[0215] The Serverless service component accesses the second page corresponding to the target account, and takes a screenshot of the second page to obtain a second screenshot result;

[0216] The second screenshot result is determined as the second page content that needs to be reviewed.

[0217] In one embodiment of the present disclosure, the service components of the large model further include an identification detection service component, a webpage detection service component and a third model service component; the second audit module 804 is specifically used to:

[0218] The jump detection service component inputs the second page content and the second text of the target account in the first page content into the recognition detection service component to obtain the jump target information;

[0219] The webpage detection service component calls the Serverless service component to access the target information and obtains the third page corresponding to the target information.

[0220] The Serverless service component accesses and takes a screenshot of the third page, and obtains the third screenshot result that needs to be reviewed;

[0221] The web page detection service component inputs the third screenshot result into the third model service component to obtain a second review result.

[0222] In one embodiment of the present disclosure, the identification and detection service component further includes a website identification service subcomponent and a group number identification service subcomponent; the second audit module 803 is specifically used to:

[0223] The jump detection service component inputs the second page content and the second text into the website identification service subcomponent and the group number identification service subcomponent respectively, and obtains the website or social account where the jump exists;

[0224] The URL or social account where the jump exists is determined as the jump target information.

[0225] In one embodiment of the present disclosure, the service component of the large model further includes a learning service component and a fourth model service component; the second audit result includes the presence of sensitive content or the absence of sensitive content; when the second audit result includes the presence of sensitive content, the above-mentioned device is further used to:

[0226] The learning service component calls the word vector generation component to convert the third text containing sensitive content into a second vector;

[0227] The learning service component calls the retrieval component to calculate the similarity between the second vector and the vector in the historical corpus to obtain the corresponding prompt word text;

[0228] The learning service component concatenates the prompt word text with the third text, and inputs the concatenated text into the fourth model service component to obtain the target question text;

[0229] The learning service component calls the word vector generation component to convert the target question text into a third vector, and stores the third vector and the third text in the historical corpus.

[0230] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium and a computer program product.

[0231] Fig. 9 A schematic block diagram of an example electronic device 900 that can be used to implement an embodiment of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0232] like Fig. 9 As shown, the device 900 includes a computing unit 901, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 902 or a computer program loaded from a storage unit 908 into a random access memory (RAM) 903. In the RAM 903, various programs and data required for the operation of the device 900 can also be stored. The computing unit 901, the ROM 902, and the RAM 903 are connected to each other via a bus 904. An input / output (I / O) interface 905 is also connected to the bus 904.

[0233] A number of components in the device 900 are connected to the I / O interface 905, including: an input unit 906, such as a keyboard, a mouse, etc.; an output unit 907, such as various types of displays, speakers, etc.; a storage unit 908, such as a disk, an optical disk, etc.; and a communication unit 909, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 909 allows the device 900 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0234] The computing unit 901 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 901 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 901 performs the various methods and processes described above, such as the content audit method of the large model. For example, in some embodiments, the content audit method of the large model may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 908. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 900 via the ROM 902 and / or the communication unit 909. When the computer program is loaded into the RAM 903 and executed by the computing unit 901, one or more steps of the method described above may be performed. Alternatively, in other embodiments, the computing unit 901 may be configured to perform the content audit method of the large model in any other appropriate manner (e.g., by means of firmware).

[0235] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0236] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0237] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0238] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0239] The systems and techniques described herein can be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), the Internet, and a blockchain network.

[0240] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship between the client and the server is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services ("Virtual Private Server", or "VPS" for short). The server may also be a server of a distributed system, or a server combined with a blockchain.

[0241] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0242] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A content audit method for a large model, the method comprising: The serverless service component obtains the first page content that needs to be reviewed; Performing a first review on the first page content based on the service component set of the large model to obtain a first review result; The Serverless service component obtains second page content that needs to be reviewed based on the first review result; A second audit is performed on the second page content based on the service component set of the large model to obtain a second audit result.

2. The method of claim 1, wherein: The Serverless service component obtains the first page content that needs to be reviewed, including: The Serverless service component obtains the first page of accessing the social software; The Serverless service component accesses the content in the first page to obtain a corresponding preset number of content result pages; The Serverless service component obtains the first page content that needs to be reviewed based on the content result page.

3. The method of claim 2, wherein: The Serverless service component accesses the content in the first page to obtain a corresponding preset number of content result pages, including: The serverless service component searches based on the keywords in the first page to obtain a corresponding preset number of content result pages; or The Serverless service component accesses the content in the first page based on the ranking, and obtains a corresponding preset number of content result pages.

4. The method of claim 2, wherein: The Serverless service component obtains the first page content that needs to be reviewed based on the content result page, including: The Serverless service component determines the target comment content that needs to be reviewed in the content result page; The Serverless service component takes a screenshot of the target comment content to obtain a first screenshot result; The Serverless service component determines the first screenshot result as the first page content that needs to be reviewed.

5. The method of claim 1, wherein: The service component set of the large model includes a jump detection service component, a first model service component, a word vector generation component, a retrieval component, and a second model service component; The service component set based on the big model performs a first audit on the first page content to obtain a first audit result, including: The jump detection service component obtains, based on the first page content, a target similar text that satisfies a similarity condition with the first page content and a first text corresponding to the first page content through the first model service component, the word vector generation component and the retrieval component; The retrieval component concatenates the first text with the target similar text, and inputs the concatenated text into the second model service component to obtain a first review result.

6. The method of claim 5, wherein: The jump detection service component obtains, based on the first page content, a target similar text that satisfies a similarity condition with the first page content and a first text corresponding to the first page content through the first model service component, the word vector generation component and the retrieval component, including: The jump detection service component inputs the first page content into the first model service component to obtain a first text corresponding to the first page content; The jump detection service component calls the word vector generation component to perform vector conversion on the first text to obtain a corresponding first vector; The jump detection service component calls the retrieval component, calculates the similarity between the first vector and the vectors in the historical corpus, and determines the target similar text that meets the similarity condition with the first page content based on the obtained similarity result.

7. The method of claim 1, wherein: The first audit result includes whether a potential jump exists or whether a potential jump does not exist. When the first audit result includes whether a potential jump exists, the Serverless service component obtains second page content that needs to be audited based on the first audit result, including: The Serverless service component determines a target account with a potential redirection situation based on the first audit result; The Serverless service component accesses a second page corresponding to the target account, and takes a screenshot of the second page to obtain a second screenshot result; The second screenshot result is determined as the second page content that needs to be reviewed.

8. The method of claim 7, wherein: The service components of the large model also include an identification detection service component, a webpage detection service component and a third model service component; The service component set based on the large model performs a second audit on the second page content to obtain a second audit result, including: The jump detection service component inputs the second page content and the second text of the target account in the first page content into the identification detection service component to obtain the jump target information; The webpage detection service component calls the Serverless service component to access based on the target information, and obtains a third page corresponding to the target information; The Serverless service component accesses and takes a screenshot of the third page to obtain a third screenshot result that needs to be reviewed; The web page detection service component inputs the third screenshot result into the third model service component to obtain a second review result.

9. The method of claim 8, wherein: The identification and detection service component also includes a website identification service subcomponent and a group number identification service subcomponent; The jump detection service component inputs the second page content and the second text of the target account in the first page content into the identification detection service component to obtain the jump target information, including: The jump detection service component inputs the second page content and the second text into the website identification service subcomponent and the group number identification service subcomponent respectively, and obtains the website or social account where the jump exists; The URL or social account where the jump exists is determined as the jump target information.

10. The method of claim 1, wherein: The service components of the large model also include a learning service component and a fourth model service component; The second audit result includes the presence or absence of sensitive content; When the second audit result includes the presence of sensitive content, the method further includes: The learning service component calls the word vector generation component to convert the third text containing the sensitive content into a second vector; The learning service component calls the retrieval component to calculate the similarity between the second vector and the vector in the historical corpus to obtain the corresponding prompt word text; The learning service component concatenates the prompt word text with the third text, and inputs the concatenated text into the fourth model service component to obtain a target question text; The learning service component calls the word vector generation component to convert the target question text into a third vector, and stores the third vector and the third text in the historical corpus.

11. A content review device for a large model, the device comprising: The first acquisition module is used by the serverless service component to acquire the first page content that needs to be reviewed; A first review module, configured to perform a first review on the first page content based on a service component set of a large model to obtain a first review result; A second acquisition module, configured for the Serverless service component to acquire second page content that needs to be reviewed based on the first review result; The second audit module is used to perform a second audit on the second page content based on the service component set of the large model to obtain a second audit result.

12. The device of claim 11, wherein: The first acquisition module is specifically used to: The Serverless service component obtains the first page of accessing the social software; The Serverless service component accesses the content in the first page to obtain a corresponding preset number of content result pages; The Serverless service component obtains the first page content that needs to be reviewed based on the content result page.

13. The device of claim 12, wherein: The first acquisition module is further used for: The serverless service component searches based on the keywords in the first page to obtain a corresponding preset number of content result pages; or The Serverless service component accesses the content in the first page based on the ranking, and obtains a corresponding preset number of content result pages.

14. The device of claim 12, wherein: The first acquisition module is further used for: The Serverless service component determines the target comment content that needs to be reviewed in the content result page; The Serverless service component takes a screenshot of the target comment content to obtain a first screenshot result; The Serverless service component determines the first screenshot result as the first page content that needs to be reviewed.

15. The device of claim 11, wherein: The service component set of the large model includes a jump detection service component, a first model service component, a word vector generation component, a retrieval component, and a second model service component; the first audit module is specifically used to: The jump detection service component obtains, based on the first page content, a target similar text that satisfies a similarity condition with the first page content and a first text corresponding to the first page content through the first model service component, the word vector generation component and the retrieval component; The retrieval component concatenates the first text with the target similar text, and inputs the concatenated text into the second model service component to obtain a first review result.

16. The device of claim 15, wherein: The first audit module is specifically used for: The jump detection service component inputs the first page content into the first model service component to obtain a first text corresponding to the first page content; The jump detection service component calls the word vector generation component to perform vector conversion on the first text to obtain a corresponding first vector; The jump detection service component calls the retrieval component, calculates the similarity between the first vector and the vectors in the historical corpus, and determines the target similar text that meets the similarity condition with the first page content based on the obtained similarity result.

17. The device of claim 11, wherein: The first audit result includes the existence of a potential jump situation or the absence of a potential jump situation; when the first audit result includes the existence of a potential jump situation, the second acquisition module is specifically used to: The Serverless service component determines a target account with a potential redirection situation based on the first audit result; The Serverless service component accesses a second page corresponding to the target account, and takes a screenshot of the second page to obtain a second screenshot result; The second screenshot result is determined as the second page content that needs to be reviewed.

18. The device of claim 17, wherein: The service components of the large model also include an identification and detection service component, a webpage detection service component and a third model service component; the second audit module is specifically used to: The jump detection service component inputs the second page content and the second text of the target account in the first page content into the identification detection service component to obtain the jump target information; The webpage detection service component calls the Serverless service component to access based on the target information, and obtains a third page corresponding to the target information; The Serverless service component accesses and takes a screenshot of the third page to obtain a third screenshot result that needs to be reviewed; The web page detection service component inputs the third screenshot result into the third model service component to obtain a second review result.

19. The device of claim 18, wherein: The identification and detection service component also includes a website identification service subcomponent and a group number identification service subcomponent; the second audit module is specifically used to: The jump detection service component inputs the second page content and the second text into the website identification service subcomponent and the group number identification service subcomponent respectively, and obtains the website or social account where the jump exists; The URL or social account where the jump exists is determined as the jump target information.

20. The device of claim 11, wherein: The service component of the large model also includes a learning service component and a fourth model service component; the second audit result includes the presence of sensitive content or the absence of sensitive content; when the second audit result includes the presence of sensitive content, the device is further used to: The learning service component calls the word vector generation component to convert the third text containing the sensitive content into a second vector; The learning service component calls the retrieval component to calculate the similarity between the second vector and the vector in the historical corpus to obtain the corresponding prompt word text; The learning service component concatenates the prompt word text with the third text, and inputs the concatenated text into the fourth model service component to obtain a target question text; The learning service component calls the word vector generation component to convert the target question text into a third vector, and stores the third vector and the third text in the historical corpus.

21. An electronic device, comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 10.

22. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to cause the computer to execute the method according to any one of claims 1 to 10.

23. A computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the method according to any one of claims 1 to 10 is implemented.

Images