Cross-service scene content detection method and device, storage medium and equipment

Abstract

The invention provides a cross-service scene content detection method and device, a storage medium and equipment. The content detection method comprises the following steps: acquiring service trigger information submitted by a user to any service module in a plurality of service modules, and a service output result presented by any service module in a corresponding page; determining a cross-scene information verification logic uniformly formulated for the business scenes of the plurality of business modules, and constructing prompt words in combination with the business trigger information and the business output result; the cue word is input into a preset verification engine based on a generative language model, a verification result output by the verification engine is obtained, and the verification result comprises a result obtained by conducting consistency detection on the service triggering information and the target presentation content based on cross-scene information verification logic and / or a result obtained by conducting consistency detection on the service triggering information and the target presentation content based on the cross-scene information verification logic. And based on the cross-scene information verification logic, carrying out relevance detection on the target presentation content and other presentation content to obtain a result.

Description

Technical Field

[0001] This specification relates to the field of computer technology, and in particular to a content detection method, apparatus, storage medium and device for cross-business scenarios. Background Technology

[0002] In mainstream information presentation scenarios such as information search, information recommendation, and advertising, users increasingly demand higher accuracy and a more complete experience in information acquisition. However, in actual business operations, common experience problems exist, such as inconsistencies between delivery and delivery, and deficiencies in content quality. These problems directly affect user trust and business conversion rates.

[0003] However, existing information detection methods require separate detection tools for different information presentation business scenarios. The development, deployment, and maintenance of these tools require significant human and resource costs. Furthermore, since the detection logic of each tool is independent and data cannot be shared, when content from multiple business scenarios is presented on the same page, only explicit problems in a single business scenario can be detected. This results in low accuracy of content detection and makes it difficult to meet the needs of efficient and unified content quality control across multiple business scenarios. Summary of the Invention

[0004] In view of the above, one or more embodiments of this specification provide the following technical solutions: According to a first aspect of one or more embodiments of this specification, a content detection method for cross-business scenarios is proposed, comprising: Obtain business trigger information submitted by a user to any one of multiple business modules, and the business output result presented by the business module on the corresponding page; wherein, the business output result includes the target presentation content of the business scenario corresponding to the business module and other presentation content of the business scenario corresponding to other business modules; Determine a unified cross-scenario information verification logic for the business scenarios of the multiple business modules, and construct prompt words by combining the business trigger information and the business output results; The prompt words are input into a preset validation engine based on a generative language model, and the validation results output by the validation engine are obtained. The validation results include: the result of consistency detection of the business trigger information and the target presentation content based on the cross-scenario information validation logic, and / or the result of correlation detection of the target presentation content and the other presentation content based on the cross-scenario information validation logic.

[0005] According to a second aspect of one or more embodiments of this specification, a content detection device for cross-business scenarios is proposed, comprising: The acquisition module is used to acquire business trigger information submitted by the user to any one of the multiple business modules, as well as the business output result presented by the business module on the corresponding page; wherein, the business output result includes the target presentation content of the business scenario corresponding to the business module and other presentation content of the business scenario corresponding to other business modules. The module is used to determine the cross-scenario information verification logic uniformly formulated for the business scenarios of the multiple business modules, and to construct prompt words by combining the business trigger information and the business output results; The detection module is used to input the prompt words into a preset validation engine based on a generative language model and obtain the validation results output by the validation engine. The validation results include: the result of consistency detection of the business trigger information and the target presentation content based on the cross-scenario information validation logic, and / or the result of correlation detection of the target presentation content and the other presentation content based on the cross-scenario information validation logic.

[0006] According to a third aspect of one or more embodiments of this specification, an electronic device is provided, comprising: a processor; a memory for storing processor-executable instructions; wherein the processor implements the steps of the method described above by executing the executable instructions.

[0007] According to a fourth aspect of one or more embodiments of this specification, a computer-readable storage medium is provided that stores computer instructions thereon, which, when executed by a processor, implement the steps of the method described above.

[0008] According to a fifth aspect of one or more embodiments of this specification, a computer program product is provided, comprising a computer program / instructions that, when executed by a processor, implement the steps of the method described above.

[0009] As can be seen from the above embodiments, this specification obtains the business trigger information submitted by the user to any of the multiple business modules, as well as the business output results presented by the business module on the corresponding page. Based on the cross-scenario information verification logic uniformly formulated for the business scenarios of the multiple business modules, it constructs prompt words by combining the business trigger information and business output results, and inputs them into a verification engine based on a generative language model. In this way, the verification engine can accurately detect the consistency and relevance of content in multiple business scenarios. Moreover, the cross-scenario information verification logic corresponding to different business modules is the same, thereby avoiding the need to develop separate detection tools for different businesses, reducing detection costs and improving tool reusability, and improving the accuracy of content detection and the adaptability to multiple business scenarios. Attached Figure Description

[0010] Figure 1This is a schematic diagram of the architecture of a content detection service system that spans multiple business scenarios, provided in an exemplary embodiment. Figure 2 This is a flowchart illustrating a content detection method across business scenarios provided in an exemplary embodiment; Figure 3 This is an exemplary embodiment of a content detection flowchart; Figure 4 This is an exemplary embodiment of a content detection page diagram; Figure 5 This is a schematic diagram of the structure of a device provided in an exemplary embodiment; Figure 6 This is a block diagram of a content detection device for cross-business scenarios provided in an exemplary embodiment. Detailed Implementation

[0011] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments in this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this specification.

[0012] The user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this manual are all information and data authorized by the user or fully authorized by all parties. The collection, use and processing of related data shall comply with the relevant laws, regulations and standards of the relevant countries and regions, and corresponding operation portals shall be provided for users to choose to authorize or refuse.

[0013] In mainstream information presentation business scenarios such as information search, information recommendation, and advertising marketing, there are common experience problems such as inconsistency between the landing page and the content of the landing page (i.e., the content of the landing page is inconsistent with the content of the landing page) and content quality defects. However, the existing content detection system has significant shortcomings in solving these problems. On the one hand, traditional detection methods use independent detection systems for different business scenarios. This not only results in inconsistent problem handling standards and the need to develop separate detection tools, but also makes it impossible to perform correlation analysis on cross-business content. This leads to a waste of resources and low accuracy, which greatly hinders the optimization of the overall experience.

[0014] Specifically, in a single business scenario, traditional detection methods can usually only perform isolated compliance checks on the target content in that scenario based on the detection rules and data dimensions specific to that business scenario. However, when the content from multiple scenarios is presented in the same user interface, only the target content of that business scenario can be detected, while ignoring the correlation between content from different scenarios in the same interface, making it difficult to identify hidden problems between content from different scenarios.

[0015] For example, in information search scenarios, in addition to displaying search results, information search pages also display a list of recommended information and pop-up ads. Traditional detection methods can only verify the consistency between search content and search results, but ignore the impact of the relevance between the list of recommended information, pop-up ads and search results on the overall detection results and user experience.

[0016] On the other hand, traditional content inspection technologies are difficult to meet actual business needs. Among them, UI inspection technology can only identify visual anomalies (such as missing buttons or disordered layouts) and cannot detect semantic problems such as content tampering (such as an advertisement claiming "repair service" but the landing page contains illegal content) or service changes (such as a service advertised as "free" but requiring payment). Optical Character Recognition (OCR) technology is prone to failure in mixed text and image scenarios. It cannot distinguish between image text and decorative elements, mistakenly treats UI control text as main text, and has a high false positive rate due to font distortion and background interference. Furthermore, it cannot understand the contextual semantics.

[0017] In addition, manual review is difficult to cope with the scale and dynamic changes in information presentation business. The coverage of manual spot checks is low and comprehensive detection cannot be achieved. Furthermore, the tampering of landing page content is often completed within minutes, and dynamic changes such as illegal advertisements that suddenly go online at night are beyond the scope of human real-time monitoring capabilities.

[0018] Based on this, this specification provides a content detection method for cross-business scenarios. It is based on a cross-scenario information verification logic uniformly formulated for business scenarios of multiple business modules, combines business trigger information and business output results to construct prompt words, and uses a verification engine based on a generative language model to perform consistency detection. The cross-scenario information verification logic is the same for different business modules, eliminating the need to develop separate detection tools for different businesses, reducing detection costs and improving tool reusability, as well as improving the accuracy of content detection and adaptability to multiple business scenarios.

[0019] The technical solutions provided in the various embodiments of this specification are described in detail below with reference to the accompanying drawings.

[0020] Figure 1 This is a schematic diagram of the architecture of a content detection service system that spans multiple business scenarios, provided as an exemplary embodiment. For example... Figure 1As shown, the system may include a server 11, a network 12, and several electronic devices, such as a personal computer (PC) 13, a mobile phone 14, etc.

[0021] Server 11 can be a physical server containing an independent host, or it can be a virtual server hosted in a host cluster. During operation, server 11 can run server-side programs for a specific application to implement the relevant functions of that application. For example, when server 11 runs a content detection service program that crosses business scenarios, it can function as a corresponding content detection service platform that crosses business scenarios.

[0022] PC13 and mobile phone14 are just some of the types of electronic devices that users can use. In reality, users can obviously also use electronic devices such as tablets, laptops, PDAs (Personal Digital Assistants), wearable devices (such as smart glasses, smartwatches, etc.), etc., and one or more embodiments in this specification do not limit this. During operation, the electronic device can run a client-side program of an application to implement the relevant functions of that application. For example, when the electronic device runs a cross-business scenario content detection service program, it can act as a client for that cross-business scenario content detection service. The aforementioned cross-business scenario content detection service client application can be launched and run on the electronic device. This client-side program can be a native application installed on the electronic device, or it can be a mini-program, quick app, or other similar form. Of course, when using web technologies such as HTML5 or similar, the relevant functions can be implemented through a page displayed by a browser. This browser can be a standalone browser application or a browser module embedded in some applications.

[0023] As for the network 12 that enables interaction between electronic devices such as PC13 and mobile phone 14 and server 11, communication can be achieved using either wired or wireless networks, depending on the communication methods supported by the respective electronic devices. This specification does not impose any restrictions on this. For example, PC13 can support both wired and wireless communication, so it can use either wired or wireless networks as needed. Mobile phone 14 typically only supports wireless communication, so it can use a wireless network for communication.

[0024] Based on the above architecture, this specification provides a content detection method for cross-business scenarios, such as... Figure 2 As shown.

[0025] Figure 2 This is a flowchart illustrating a content detection method across business scenarios provided in an exemplary embodiment, including the following steps: S200: Obtain the business trigger information submitted by the user to any one of the multiple business modules, and the business output result presented by the business module on the corresponding page; wherein, the business output result includes the target presentation content of the business scenario corresponding to the business module and other presentation content of the business scenario corresponding to other business modules.

[0026] In this specification, the execution subject for performing the content detection method across business scenarios can be a designated device such as a server. Of course, it can also be a client for content detection or a terminal device with the client installed. For ease of description, the following will use a server as the execution subject to explain the content detection method across business scenarios provided in this specification.

[0027] The server can obtain the business trigger information submitted by the user to any of the multiple business modules, as well as the business output results presented by any of the business modules on the corresponding page.

[0028] The aforementioned multiple business modules can be different business modules within a single target application. For example, an app may offer three types of services: information search, information recommendation, and advertising marketing. Therefore, its corresponding business modules could include: information search module, information recommendation module, and advertising marketing module.

[0029] In addition, the aforementioned business modules can also be business modules corresponding to different target applications, with each business module corresponding to one target application. In this case, the business modules can include: business modules corresponding to information search applications, business modules corresponding to information recommendation applications, and business modules corresponding to advertising and marketing applications.

[0030] The business output results include the target presentation content of the current business module corresponding to the business scenario and other presentation content of other business modules corresponding to the business scenario.

[0031] For example, in the information search business, users can enter search terms on the search page corresponding to the search module and click the search button or trigger voice search. On the display page of the search module, in addition to displaying the corresponding search results (i.e. the target content), other display content corresponding to other business scenarios, such as recommendation lists and floating advertising windows, can also be displayed. For example, in the information recommendation business, after a user clicks on a certain item in the recommendation list on the page corresponding to the recommendation module (such as the homepage or personalized recommendation page), in addition to displaying the detailed information of the item, the display page of the recommendation module can also display a floating ad window, an ad notification bar, or an ad information flow card.

[0032] In practical applications, information search typically refers to a service that accurately matches and returns relevant results from a database or information pool based on user-inputted search information (such as search terms, files, images, etc.). After a user inputs search information, the system performs semantic analysis, keyword extraction, and data retrieval, ultimately presenting the corresponding web pages, articles, products, services, and other information results in the form of lists, cards, etc.

[0033] Information recommendation typically refers to a service that proactively pushes content that users may be interested in based on their historical behavior, interests, or contextual characteristics. After a user interacts with a piece of content (such as clicking, liking, or saving it), the application will further present the corresponding specific content, related topics, or a personalized recommendation list.

[0034] Advertising and marketing typically refer to services that deliver advertising messages to a specific target audience. When users open an app, browse content streams (such as homepage feeds or video playback pages), or use specific functions, the system displays ad placement pages in these scenarios. These pages often appear as feed cards, pop-ups, or splash screens. When a user clicks on the ad placement page, the system redirects them to the ad landing page. The landing page is the core page for achieving conversions, presenting more detailed information such as product details, activity rules, brand introductions, form entry points, or download links to guide users to complete actions desired by the advertiser, such as purchasing, registering, or downloading.

[0035] Based on this, when a user performs the following operations, the business trigger information submitted by the user can be determined: When a user enters search information in the information search module and performs a confirmation operation, the system confirms that the user has submitted the business trigger information. When a user selects a certain item in the information recommendation module, the user confirms the submission of business trigger information; After a user clicks on the ad placement page, the system confirms that the user has submitted the business trigger information.

[0036] It should be noted that the above only uses information search, information recommendation, and advertising marketing as typical examples to illustrate the division of business modules and the information triggered by these modules. In practical applications, other businesses and corresponding business modules may also be included, such as e-commerce transactions (e.g., order placement, order management), social interaction (e.g., adding friends, sending messages), and content creation (e.g., video publishing, article editing). This manual will not list them all.

[0037] In this specification, business triggering information can be information used to trigger the current business, such as information related to user-input search information, information recommendation content, advertising placement page, etc. Business output results can be information further presented on the client after the current business is triggered, such as information related to search results, recommendation detail page, advertising landing page, etc.

[0038] For any business module, the server can collect data from the graphic and text data displayed on the front end of the business module and the graphic and text data loaded in the data layer of the business module. Then, based on the status identifiers carried by each graphic and text data, the server can determine the business trigger information and business output results in each graphic and text data.

[0039] In addition, in order to ensure that data belonging to different business scenarios can be accurately distinguished in the subsequent cross-scenario correlation detection process, the server can determine the presentation content corresponding to different business scenarios based on the scenario identifiers (such as the entry identifier of the business module, page path identifier, etc.) carried by each graphic data corresponding to the business output results.

[0040] The image and text data displayed on the front end can include screenshots of the business trigger page, text metadata (such as title, description, UI layout information, etc.), screenshots of the business output page, text metadata, and loading performance metrics (such as page first rendering time, resource loading completion time, and interaction response latency).

[0041] For the information search module, the image and text data loaded in its data layer can include the parsing results corresponding to search logs (such as MobileSearch logs), which include the search information entered by the user, search result text and images, etc. For the information recommendation module, the image and text data loaded in its data layer can include the distribution records of recommended content (such as RecMixerII content), which includes the text and images contained in the recommended content, image resolution / color parameters, the original encoding format of the recommended text, etc.

[0042] For the advertising and marketing module, the image and text data loaded in its data layer can include metadata of the content to be delivered (such as Advanced Data Exchange (ADX) data), which includes advertising material images, text content and layout rules of advertising copy, pixel size of original material images, page URLs, etc.

[0043] It should be noted that user-submitted business trigger information can include both directly submitted and indirectly submitted business trigger information. For example, the search information entered by the user in the search module is directly submitted business trigger information, while screenshots of the search page and metadata loaded in the data layer are indirectly submitted business trigger information. As another example, when a user clicks the "View Now" button on the ad placement page in the advertising and marketing module, the click itself is directly submitted business trigger information, while the real-time screenshot of the ad placement page and the metadata loaded in the data layer are indirectly submitted business trigger information.

[0044] Furthermore, during the user's execution of the above-mentioned business processes, the business trigger information and business output results may contain user privacy data. Due to the restrictions of data transmission security specifications, it is not possible to directly collect and transmit this data. In addition, it is necessary to avoid delaying the user's current business operations (such as loading search results or redirecting to advertising pages) and to ensure the efficiency of normal business interactions on the user's end. Therefore, the user's end often cannot directly upload the business trigger information and business output results to the server, making it impossible for the server to directly obtain the business trigger information and business output results from the user's end.

[0045] In this scenario, the server can pre-build a corresponding information presentation environment, thereby reproducing the relevant content of the user-triggered business within that environment. This allows for content detection without acquiring sensitive user data or interfering with the user's actual business operations.

[0046] Specifically, the server can determine the application type of the target application corresponding to the business module, and then construct an information presentation environment that matches the application type.

[0047] For example, if the target application is a mini-program, the cloud acceptance real device cluster is used as the information presentation environment; if the target application is an H5 page, the X Engine simulator is used as the information presentation environment; if the target application is a native application, the actual device is used as the information presentation environment.

[0048] After a user submits a service trigger information, although the user client cannot directly transmit the service information to the server, the server can determine the address or identifier associated with the user client when generating the service trigger information by monitoring the user client's service trigger behavior and environmental status in real time. Based on the address or identifier, the server can determine the first page source data on which the user client generates the service trigger information, and then generate the service trigger information in the information presentation environment based on the first page source data.

[0049] In addition, the server can also determine the second page source data associated with the first page source data, and then generate the business output result in the information presentation environment based on the second page metadata. The aforementioned second page source data can also be obtained by the server through monitoring or sensing the user's business triggering behavior and environmental status. Alternatively, it can be further analyzed and derived based on the page jump links carried in the business triggering information or the interaction logic definition of the user's business modules (such as page jump rules and page flow configuration after function triggering).

[0050] The above page source data (including the first page source data and the second page source data) may include: page structure definition (such as component layout rules, nesting relationship of functional modules), business logic configuration (such as search information matching rules, recommended content filtering strategy), general interaction entry definition (such as button function binding, input box response mechanism), and other basic data used to recreate the information presented to the user.

[0051] Furthermore, to ensure the accuracy of the restored information, the server can detect information windows unrelated to the target application in the user terminal based on a preset information interception protocol, and close the detected information windows.

[0052] For example, if a system notification window unrelated to the target application's business pops up on the user's device during business execution (such as a push message from another application), the server will identify and close the window through an information interception protocol. The server can also filter dynamic elements in the presentation page corresponding to business trigger information and / or business output results based on preset element filtering rules. For example, when reproducing the search results page of the information search module, dynamic elements such as real-time updated advertising animations and scrollbars of recommended content unrelated to the search information can be filtered out. In addition, the server can monitor the stability of the page and set corresponding timeout thresholds. When the page loading time exceeds the timeout threshold (e.g., the preset maximum loading time for a certain business page is 5 seconds, but the actual loading time reaches 8 seconds), the page loading is determined to be abnormal, triggering the abnormal investigation process or using backup page source data for reproduction.

[0053] S202: Determine a cross-scenario information verification logic uniformly formulated for the business scenarios of the multiple business modules, and construct prompt words by combining the business trigger information and the business output results; S204: Input the prompt word into a preset validation engine based on a generative language model, and obtain the validation result output by the validation engine. The validation result includes: the result of consistency detection of the business trigger information and the target presentation content based on the cross-scenario information validation logic, and / or the result of correlation detection of the target presentation content and the other presentation content based on the cross-scenario information validation logic.

[0054] After determining the business trigger information and business output structure, the server can further determine the cross-scenario information verification logic uniformly formulated for business scenarios of multiple business modules, and construct prompt words by combining the business trigger information and business output results.

[0055] In this specification, the cross-scenario verification logic can be used to detect the consistency between the target content presented in the business trigger information and the business output results. An example of the verification logic corresponding to this consistency detection is as follows: Text consistency detection is performed based on the semantic similarity between the text information corresponding to the business trigger information and the text information corresponding to the target presentation content. Image-text consistency detection is performed based on the matching degree between text and image information corresponding to business trigger information and the matching degree between text and image information corresponding to target presentation content; Image consistency detection is performed based on the similarity between the image information corresponding to the business trigger information and the image information corresponding to the target presentation content.

[0056] In practical applications, it often happens that some business output content does not match the user's current needs. For example, a user searches for financial-related content, but the displayed page shows pop-ups unrelated to finance, such as game downloads, beauty product promotions, and restaurant group buying. To avoid this, the aforementioned cross-scenario validation logic can also be used to perform correlation detection on the presentation content corresponding to different business scenarios in the business output results. An example of the correlation detection validation logic is as follows: Based on the degree of correlation between the text information corresponding to the presented content belonging to different business scenarios, the text correlation between each presented content is determined; Based on the degree of correlation between the image information corresponding to the presentation content belonging to different business scenarios, the image correlation between each presentation content is determined; Based on the degree of correlation between the image and text information corresponding to the content presented in different business scenarios, the image-text correlation between each piece of content is determined.

[0057] It should be noted that the consistency detection and correlation detection mentioned above belong to two different detection dimensions, among which: Consistency detection is based on the similarity between two objects (i.e., the degree of similarity between the two objects, reflecting the direct similarity in content expression). This can be determined by comparing and calculating the textual semantic features, image visual features, and image-text matching relationships between the two objects. When the similarity is higher than a preset consistency threshold, the two objects are considered to have consistent content expression; otherwise, they are considered to have inconsistent content expression. The correlation detection is based on the degree of correlation between two objects being tested (i.e., the degree of correlation between the two objects in terms of business theme, service scope, user demand direction, etc., reflecting whether the two objects share common business goals or user needs). This can be determined through the overlap of their business classification systems, user demand relevance, or key information (such as keywords or key icons). When the correlation is higher than a preset correlation threshold, the two objects are considered to be related; otherwise, they are considered not related.

[0058] For example, in consistency testing, if a user enters the search term "query the tax declaration guide for xxx year" but the search result is "the housing price trend in xxx year", the two have low similarity and can be judged to fail the consistency test. Regarding the relevance test, if the business output contains the search results "tax declaration guide for xxx year" and the recommended information "tax deduction filing tips", although the similarity between the two is low, they both involve the keyword "tax declaration" and are highly related in terms of both business classification and user needs. Therefore, it can be determined that the relevance test between the two passes.

[0059] In addition, cross-scenario verification logic can also be used to check the content quality of business trigger information and business output results. An example of the verification logic corresponding to quality detection is as follows: Check whether the text content corresponding to the business trigger information and business output results meets the preset text quality. Check whether the image content corresponding to the service trigger information and service output results meets the preset image quality.

[0060] The above verification logic can be used to detect various business scenarios, including information search, information recommendation, and advertising marketing.

[0061] In practical applications, there are multiple ways to construct prompt words. For example, the server can input the business trigger information and business output results into a preset prompt word template. The prompt word template contains the aforementioned cross-scenario verification logic. After filling the business trigger information and business output results into the corresponding positions in the prompt word template, the prompt word can be generated.

[0062] For example, the server can input business trigger information and business output results into a validation engine based on a generative language model (such as a Large Language Model (LLM)). This validation engine is pre-configured with the cross-scenario validation logic mentioned above, and can automatically parse the relationship between business trigger information and business output results. Combining the validation dimensions of image-text consistency, text consistency and quality detection, it can dynamically generate prompt words that include validation targets, data dimensions and judgment criteria.

[0063] After constructing the prompt words, the server can input them into the validation engine of the aforementioned generative language model and obtain the validation result output by the validation engine. The validation result includes the result of consistency detection of the business trigger information and the business output result based on the cross-scenario information validation logic.

[0064] Specifically, the process by which the verification engine performs consistency checks on business trigger information and business output results based on cross-scenario information verification logic is as follows: The text consistency detection result is determined based on the semantic similarity between the text information corresponding to the business trigger information and the text information corresponding to the business output result. For example, if the business trigger information is "red sneakers" entered by the user in the search module, and the business output text information is "red hiking shoes promotion", the text consistency detection is passed if the similarity between the two is calculated to be 85% (the preset threshold is 70%). If the output text is "blue sports backpack", the semantic similarity is only 30%, and it is determined to be inconsistent.

[0065] The image-text consistency detection result is determined based on the matching degree between the text and image information corresponding to the business trigger information and the matching degree between the text and image information corresponding to the business output results. For example, in the business trigger information, the user inputs the text "mountain bike" and the accompanying reference image is a mountain bike with shock absorbers, and the matching degree between the two is 90%; in the business output result, the text description is "mountain bike" but the image shows a regular city bike, and the image-text matching degree is only 40%. Therefore, the overall judgment is that the image-text consistency test fails.

[0066] The image consistency detection result is determined based on the similarity between the image information corresponding to the business trigger information and the image information corresponding to the business output result. For example, if the image in the business trigger information is a brand logo (containing a specific font and graphic combination), and the image in the business output is a blurred and compressed version of the logo (the key graphic elements are complete but the clarity is reduced), and the similarity calculated by feature point comparison is 88% (the preset qualified threshold is 80%), then the image consistency detection is deemed to have passed; if the core graphic of the logo in the output image is replaced, then the similarity is 30%, and it is deemed to be inconsistent.

[0067] The verification engine can then determine the consistency test results for business trigger information and business output results based on the text consistency test results, image consistency test results, and image consistency test results.

[0068] The final consistency check result can be a binary decision result, i.e. consistent or inconsistent. For example, if one of the three consistency check results is determined to be inconsistent, the final consistency check result is inconsistent. However, if all three check results are determined to be consistent, the final consistency check result is consistent.

[0069] Alternatively, the final consistency test result can also be a score calculated based on the weighted average of the three consistency test results mentioned above (e.g., 0-100 points, with higher scores indicating stronger consistency). For example, if text consistency has a weight of 40%, image-text consistency has a weight of 35%, and image consistency has a weight of 25%, and scores of 80, 90, and 85 points respectively, then the overall score would be 80×40%+90×35%+85×25%=84.75 points.

[0070] Of course, the consistency detection result can also be a combination of sub-items, that is, it includes the specific judgment conclusions (consistent / inconsistent) and key evidence corresponding to each of the three detection types. For example, "Text consistency: consistent (semantic similarity 85%); Image-text consistency: inconsistent (output result image-text matching degree 40%); Image consistency: consistent (feature point similarity 88%)".

[0071] In addition, the verification results output by the aforementioned verification engine may also include: the results of correlation detection of the presentation content corresponding to different business scenarios in the business output results based on cross-scenario information logic. The process of determining the correlation detection results is as follows: Based on the degree of correlation between the text information corresponding to the presented content belonging to different business scenarios, the text correlation between each presented content is determined; For example, if the search result text is "Tax Filing Guide for XXX Year" (search scenario) and the recommended list text is "Tax Deduction Filing Tips" (recommendation scenario), both revolve around the business theme of "tax processing," so it is determined that there is textual relevance. However, if the recommended list text is "Home Appliance Purchase Guide," which has no business theme relevance to the search result text, then it is determined that there is no textual relevance.

[0072] Based on the degree of correlation between the image information corresponding to the presentation content belonging to different business scenarios, the image correlation between each presentation content is determined; For example, if the ad pop-up image is a "physical credit card image" (ad scenario) and the search result image is a "screenshot of a credit card bill query interface" (search scenario), both contain the core visual element of "credit card," so the images are considered to be related. However, if the ad pop-up image is a "snack packaging image" and has no common visual elements with the search result image, then the images are considered not related.

[0073] Based on the degree of correlation between the image and text information corresponding to the content presented in different business scenarios, the image-text correlation between each piece of content is determined.

[0074] For example, if the search result is "disassembly images of home appliance parts" and the recommended list is "home appliance purchasing guide", both the images in the search results and the text in the recommended list revolve around "home appliances", so the image-text relevance exists. However, if the recommended list is "tax deduction reporting tips", which has no business relevance to the text in the search results, then the image-text relevance does not exist.

[0075] It should be added that the above correlation detection can be applied only to the correlation between the target content and the other content, or it can be applied to the correlation between all content in the business output.

[0076] Furthermore, the final correlation detection result can be a binary judgment result, namely, correlation or no correlation. For example, if any one of the above three correlation detection results is determined to be correlated, then the final correlation detection result is correlated. If all three correlation detection results are determined to be uncorrelated, then the final correlation detection result is uncorrelated.

[0077] Alternatively, the final consistency test result can also be a score calculated based on the weighted average of the three types of relevance test results mentioned above (e.g., 0-100 points, with higher scores indicating stronger relevance). For example, if the text relevance weight is 40%, the image-text relevance weight is 35%, and the image relevance weight is 25%, with scores of 10, 90, and 20 points respectively, the overall score is 41.5 points.

[0078] In addition, the verification results output by the aforementioned verification engine may also include: the results of quality detection of the information content of business trigger information and business output results based on the cross-scenario information verification logic. The process for determining the quality detection results is as follows: Determine whether the text content corresponding to the business trigger information and business output results meets the preset text quality, and obtain the text quality detection result; For example, if the text corresponding to the business trigger information has an incomplete title description (e.g., the advertisement title only shows "Promotion" without specifying the product), or the text is truncated (e.g., "Limited-time offer" is displayed as "Limited-time offer..."), or contains illegal content (e.g., the advertisement copy contains false advertising words or inflammatory text), or contains meaningless blocks (e.g., a large number of garbled characters or repeated meaningless characters appear in the text), then the text quality check is deemed to have failed; conversely, if the text structure is complete and the content complies with the rules, then the check is deemed to have passed.

[0079] Determine whether the image content corresponding to the service trigger information and the service output result meets the preset image quality, and obtain the image quality detection result; For example, if the image corresponding to the business trigger information has insufficient clarity (such as being so blurry that key information cannot be identified), color distortion (such as the advertiser's brand color being seriously deviated), content violation, or meaningless blocks (such as a large number of noise points or blank redundant areas in the image), then the image quality test is deemed to have failed; conversely, if the image is clear, the colors are accurate, the content is compliant, and the information is complete, then it is deemed to have passed.

[0080] Correspondingly, the quality inspection result can be a binary judgment result, namely, qualified or unqualified; It can also be a score calculated based on a weighted average of text quality detection results and image quality detection results. For example, if the text quality weight is 50% and the image quality weight is 50%, and the text quality score is 90 points and the image quality score is 80 points, then the overall score is 85 points. Of course, it can also be a combination of sub-items, that is, including the specific judgment conclusions (pass / fail) and key evidence corresponding to the text quality detection and image quality detection, such as "Text quality: incomplete title description (fail); text truncation (fail); Image quality: insufficient clarity (fail); content compliance (pass)".

[0081] Furthermore, the server can determine the risk level of the business module based on the consistency detection results of the business trigger information and business output results, as well as the information quality detection results of the business trigger information and business output results. Then, based on the risk level, it can determine the target risk handling strategy and carry out risk handling for the business module through the target risk handling strategy.

[0082] For example, if there is a serious inconsistency between the business trigger information and the business output results, such as a blank landing page or service unavailability, and the image / text quality is seriously substandard, it is judged as a severe risk, and a risk handling strategy of blocking traffic in real time is implemented; if there is a discrepancy between the commitment and the actual output (such as the advertising commitment is inconsistent with the actual output), it is judged as a high-risk risk, and the person in charge is contacted for handling within 15 minutes; if there are only content quality defects (such as small-scale text truncation or slightly blurred images), it is judged as a warning risk, and a daily summary report is generated.

[0083] For example, when the consistency detection result and content detection result are scores, the comprehensive score below 60 points can be judged as a severe risk, 60-75 points as a high-risk risk, and above 75 points as a warning risk, which correspond to the handling strategies of real-time blocking, contacting the person in charge, and daily summary, respectively.

[0084] To facilitate understanding, this specification provides an overall flowchart for content detection, such as... Figure 3 As shown.

[0085] When a user triggers a business transaction via front-end event tracking, the server can recreate the current business scenario within a pre-built, user-defined presentation environment. This allows the server to determine the business trigger information and output results from the data displayed on the front-end and loaded on the back-end. Then, based on a unified cross-scenario information verification logic across multiple business modules, and combining the trigger information and output results, a prompt is constructed and input into a verification engine. The engine performs consistency and quality checks across multiple dimensions and outputs the corresponding verification results. The server can then determine the corresponding risk level based on these results and execute appropriate risk management strategies.

[0086] To ensure the continuous optimization of the verification engine's detection accuracy for business trigger information and business output results, and to reduce the false positive rate, the server can obtain historical false positive sample data of the content detection model, and iteratively adjust the verification engine parameters based on the false positive data to improve detection accuracy.

[0087] Specifically, the server can obtain misjudged sample data from the content detection model and determine the actual tag information corresponding to the misjudged sample data; wherein, the misjudged sample data includes: business trigger information and business output results that actually conform to the information verification logic but are judged not to conform to the information verification logic, and / or business trigger information and business output results that actually do not conform to the information verification logic but are judged to conform to the information verification logic; The server can construct prompt words based on the misjudged sample data and information verification logic, input the prompt words into the verification engine, and obtain the verification result to be corrected output by the verification engine. Then, the loss value is determined based on the deviation between the detection result to be corrected and the actual label information, and the parameters of the verification engine are adjusted based on the loss value.

[0088] For the above misjudged sample data, manual review can be conducted by professional auditors who will check the actual triggering information, business output results, and original test conclusions of the misjudged samples one by one or by sampling, in accordance with the information verification logic (such as consistency rules and quality standards). This will clarify whether the samples actually conform to the verification logic and assign them accurate actual labels (such as "actually conforms" or "actually does not conform"), thereby achieving the labeling of misjudged samples.

[0089] For example, if a sample is judged as "inconsistent in text consistency" by the detection model, but manual review finds that the two have the same semantic core (only the expression is different), then the actual label is marked as "compliant with the text consistency rule", and the reason for the misjudgment is recorded (such as the model's insufficient recognition of synonyms), providing a basis for subsequent adjustment of the verification engine parameters.

[0090] Furthermore, the server can display real-time verification data for business modules, enabling testing personnel to monitor the verification status of various business scenarios in real time, quickly locate abnormal samples, and trace the basis of the testing logic. For example, the testing page for each business module... Figure 4 As shown.

[0091] Figure 4 This is an exemplary embodiment of a content detection page diagram.

[0092] The content detection page can include three modules: core indicators, bad case trend, and detection result details. Core indicators are used to intuitively present the quantitative results of key data such as the total number of verifications and inconsistencies between submissions and acceptances, helping detection personnel to quickly grasp the overall verification scale and anomaly distribution. Bad case trend is used to display the fluctuation of the anomaly rate over time, making it easier for detection personnel to identify anomaly trends (such as a sudden increase in the anomaly rate during a certain period) and assisting in analyzing the temporal correlation of the problems. Detection result details are used to display the specific problem dimensions (image consistency, text consistency, etc.), related links (display image links, jump links, etc.) and unique identifiers of each anomaly sample, supporting detection personnel to accurately trace the cause of the anomaly and locate the responsible link.

[0093] As can be seen from the above, this solution focuses on the front-end graphic and textual data of business trigger information and business output results, as well as the metadata collection data loaded at the data layer. It achieves synchronous capture of multimodal data, covering core data types such as text and images, providing comprehensive data support for subsequent detection. It breaks down the detection barriers between different business modules (search, recommendation, advertising and marketing), adopts a unified verification logic, and integrates text consistency, graphic and text consistency, image consistency, and quality detection functions through the verification engine, realizing standardized and integrated verification across business scenarios. By detecting the consistency of text, graphic and textual data, and images between business trigger information and business output results, it improves the accuracy of business process verification and reduces the false judgment rate. Based on the risk level classification mechanism, it matches differentiated risk handling strategies, achieving precise risk control and efficient handling.

[0094] Figure 5 This is a schematic structural diagram of a device provided in an exemplary embodiment. For example... Figure 5 As shown, device 500 mainly consists of a communication interface 502, a user interface 504, a processor 506, and a data storage 508. These components are interconnected and communicate with each other via a system bus, network, or other connection mechanism 510. The communication interface 502 enables device 500 to communicate with other devices, access networks, and transmission networks via analog or digital modulation. For example, the communication interface 502 may include a chipset and antenna for wireless communication with a radio access network or access point. Furthermore, the communication interface 502 can be a wired interface such as Ethernet, Token Ring, or a USB port, or a wireless interface such as Wi-Fi, Bluetooth, Global Positioning System (GPS), or a wide-area wireless interface (e.g., WiMAX or LTE). Of course, the communication interface 502 can also support other forms of physical layer interfaces and standard or proprietary communication protocols. The communication interface 502 may also include multiple physical communication interfaces, such as Wi-Fi, Bluetooth, and wide-area wireless interfaces.

[0095] User interface 504 includes receiving user input and providing output to the user. Therefore, user interface 504 may include input components such as a keypad, keyboard, touch-sensitive or presence-sensitive panel, computer mouse, trackball, joystick, microphone, still camera, and video camera, and output components such as a display screen (which may be combined with a touch-sensitive panel), CRT, LCD, LED, display using DLP technology, printer, and other similar devices known or developed in the future. User interface 504 may also generate auditory output via speakers, speaker jacks, audio output ports, audio output devices, headphones, and other similar devices known or developed in the future. In some embodiments, user interface 504 may include software, circuitry, or other forms of logic capable of transmitting and receiving data from external user input / output devices. Additionally or alternatively, device 500 may support remote access from other devices via communication interface 502 or another physical interface (not shown). User interface 504 may be configured to receive user input, the position and movement of which may be indicated by indicators or cursors described herein. User interface 504 may also be configured as a display device for rendering or displaying text fragments.

[0096] Processor 506 may contain one or more general-purpose processors and / or special-purpose processors.

[0097] Data storage 508 may include one or more volatile and / or non-volatile storage components and may be integrated wholly or partially with processor 506. Data storage 508 may include removable and non-removable components.

[0098] Processor 506 is capable of executing program instructions 518 (e.g., compiled or uncompiled program logic and / or machine code) stored in data storage 508 to perform the various functions described herein. Data storage 508 may contain a non-transitory computer-readable medium on which program instructions are stored, which, when executed by device 500, enable device 500 to perform any methods, processes, or functions disclosed in this specification and / or the accompanying drawings. Execution of program instructions 518 by processor 506 may result in processor 506 using data 512.

[0099] For example, program instructions 518 may include an operating system 522 (e.g., an operating system kernel, device drivers, and / or other modules) installed on device 500 and one or more applications 520 (e.g., a browser, social application, or game application). Similarly, data 512 may include operating system data 516 and application data 514. Operating system data 516 is primarily accessible to the operating system 522, while application data 514 is primarily accessible to one or more applications 520. Application data 514 may reside in a file system visible or hidden from the user of device 500.

[0100] Application 520 can communicate with operating system 522 through one or more application programming interfaces (APIs). These APIs help application 520 read and / or write application data 514, transmit or receive information via communication interface 502, receive or display information on user interface 504, etc.

[0101] In some terminology, application 520 may be simply referred to as "app". Furthermore, application 520 can be downloaded to device 500 through one or more online app stores or app markets. However, applications can also be installed on device 500 in other ways, such as through a web browser or a physical interface on device 500 (e.g., a USB port). Please refer to [link / reference]. Figure 6 Content detection devices that can be used across business scenarios can be applied to, for example... Figure 5 The device shown implements the technical solution described in this specification. The cross-business scenario content detection device may include: The acquisition module 600 is used to acquire business trigger information submitted by the user to any one of the multiple business modules, as well as the business output result presented by the business module on the corresponding page; wherein, the business output result includes the target presentation content of the business scenario corresponding to the business module and other presentation content of the business scenario corresponding to other business modules. The construction module 602 is used to determine the cross-scenario information verification logic uniformly formulated for the business scenarios of the multiple business modules, and to construct prompt words by combining the business trigger information and the business output results; The detection module 604 is used to input the prompt word into a preset verification engine based on a generative language model and obtain the verification result output by the verification engine. The verification result includes: the result of consistency detection of the business trigger information and the business output result of the target presentation content based on the cross-scene information verification logic, and / or the result of correlation detection of the target presentation content and the other presentation content based on the cross-scene information verification logic.

[0102] Optionally, the acquisition module 600 is specifically used to: collect data from the graphic and text data displayed on the front end of the business module and the graphic and text data loaded in the data layer of the business module; determine the business trigger information and the business output result in each graphic and text data according to the status identifier carried by each graphic and text data collected; and determine the presentation content corresponding to different business scenarios according to the scene identifier carried by each graphic and text data corresponding to the business output result.

[0103] Optionally, the plurality of business modules are business modules within the target application; The acquisition module 600 is specifically used to: determine the application type corresponding to the target application and construct an information presentation environment that matches the application type; determine the first page source data on which the business trigger information is generated on the user terminal; generate the business trigger information in the information presentation environment based on the first page source data; and determine the second page source data associated with the first page source data and generate the business output result in the information presentation environment based on the second page source data.

[0104] Optionally, the acquisition module 600 is further configured to: detect information windows in the user terminal that are unrelated to the target application based on a preset information interception protocol, and close the detected information windows; and filter dynamic elements in the presentation page corresponding to the business trigger information and / or the business output result based on preset element filtering rules.

[0105] Optionally, the detection module 604 is specifically configured to: determine a text consistency detection result based on the semantic similarity between the text information corresponding to the service trigger information and the text information corresponding to the target presentation content; determine an image-text consistency detection result based on the matching degree between the text information and image information corresponding to the service trigger information and the matching degree between the text information and image information corresponding to the target presentation content; determine an image consistency detection result based on the similarity between the image information corresponding to the service trigger information and the image information corresponding to the target presentation content; and determine the consistency detection result for the service trigger information and the service output result based on the text consistency detection result, the image-text consistency detection result, and the image consistency detection result.

[0106] Optionally, the verification result may also include: the result of quality detection of the information content of the business triggering information and the business output result based on the cross-scenario information verification logic; Optionally, the detection module 604 is specifically used to: determine whether the text content corresponding to the service triggering information and the service output result meets a preset text quality, and obtain a text quality detection result; determine whether the image content corresponding to the service triggering information and the service output result meets a preset image quality, and obtain an image quality detection result; and determine the result of quality detection of the information content of the service triggering information and the service output result based on the text quality detection result and the image quality detection result.

[0107] Optionally, the device further includes: The processing module 606 is used to determine the risk level corresponding to the business module based on the consistency detection results of the business trigger information and the business output results, and the information quality detection results of the business trigger information and the business output results; determine the target risk handling strategy based on the risk level, and perform risk handling on the business module through the target risk handling strategy.

[0108] Optionally, the device further includes: The adjustment module 608 is used to acquire misjudged sample data of the content detection model and determine the actual tag information corresponding to the misjudged sample data; wherein, the misjudged sample data includes: business trigger information and business output results that actually conform to the information verification logic but are judged not to conform to the information verification logic, and / or business trigger information and business output results that actually do not conform to the information verification logic but are judged to conform to the information verification logic; constructing prompt words based on the misjudged sample data and the information verification logic, inputting the prompt words into the verification engine, and obtaining the verification result to be corrected output by the verification engine; determining the loss value based on the deviation between the detection result to be corrected and the actual tag information, and adjusting the parameters of the verification engine based on the loss value.

[0109] For ease of description, the above devices are described by dividing them into various modules or units based on their functions. Of course, when implementing one or more of these specifications, the functions of each module or unit can be implemented in the same or different software and / or hardware, or a module that performs the same function can be implemented by a combination of multiple sub-modules or sub-units, etc. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division; in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed.

[0110] Based on the same concept as the methods described above, this specification also provides an electronic device, including: a processor; a memory for storing processor-executable instructions; wherein the processor performs the steps of the method as described in any of the above embodiments by executing the executable instructions.

[0111] Based on the same concept as the methods described above, this specification also provides a computer-readable storage medium having computer instructions stored thereon that, when executed by a processor, implement the steps of the methods as described in any of the above embodiments.

[0112] Based on the same concept as the methods described above, this specification also provides a computer program product, including a computer program / instructions that, when executed by a processor, implement the steps of the methods as described in any of the above embodiments.

[0113] What those skilled in the art will understand is: In this specification, the terms "comprising," "including," or any other variations thereof are intended to cover a non-exclusive inclusion, such that a process, method, product, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, product, or apparatus. Without further limitation, the presence of additional identical or equivalent elements in a process, method, product, or apparatus that includes said elements is not excluded.

[0114] In this specification, “a,” “an,” and “the” do not specifically refer to the singular, but may also include the plural.

[0115] In this specification, ordinal numbers such as "first," "second," etc., do not necessarily indicate order; they are often used to distinguish between objects. For example, "first server" and "second server" usually refer to two servers. To differentiate between these two servers, they are described as "first server" and "second server." Of course, sometimes these two servers may be the same server.

[0116] In this specification, unless explicitly stated otherwise, "receiving and sending data" does not necessarily mean direct receiving and sending; it can also mean indirect receiving and sending. For example, A receiving data sent by B can be understood as A directly receiving the data sent by B, or it can be understood as A indirectly receiving the data sent by B through other entities such as C. Similarly, B sending data to A can be understood as B sending the data directly to A, or it can be understood as B indirectly sending the data to A through other entities such as C. Here, C can be one entity, or it can be two or more entities.

[0117] In this specification, unless explicitly stated otherwise, the relationships between structures can be direct or indirect. For example, when describing "A is connected to B," unless it is explicitly stated that A and B are directly connected, it should be understood that A can be directly connected to B or indirectly connected to B. Similarly, when describing "A is on top of B," unless it is explicitly stated that A is directly above B (AB is adjacent and A is above B), it should be understood that A can be directly above B or indirectly above B (AB is separated by other elements, and A is above B). And so on.

[0118] This specification uses specific terms to describe embodiments thereof. Terms such as "an embodiment," "one embodiment," and / or "some embodiments" refer to a particular feature, structure, or characteristic associated with at least one embodiment of this specification. Therefore, it should be emphasized and noted that references to "an embodiment," "one embodiment," or "an alternative embodiment" in different locations throughout this specification do not necessarily refer to the same embodiment. Furthermore, those skilled in the art can combine and integrate the different embodiments or examples described herein, as well as the features of those different embodiments or examples, without contradiction.

[0119] Although one or more embodiments of this specification provide method steps as described in the embodiments or flowcharts, it is understood that the order of steps listed in the embodiments or flowcharts is only one of many possible execution orders and does not represent the only execution order. Therefore, when the claims involve method steps, any changes or adjustments to the order of such steps, or the parallelism between steps, are also within the scope of protection of the claims.

Claims

1. A content detection method for cross-business scenarios, comprising: Obtain business trigger information submitted by a user to any one of multiple business modules, and the business output result presented by the business module on the corresponding page; wherein, the business output result includes the target presentation content of the business scenario corresponding to the business module and other presentation content of the business scenario corresponding to other business modules; Determine a unified cross-scenario information verification logic for the business scenarios of the multiple business modules, and construct prompt words by combining the business trigger information and the business output results; The prompt words are input into a preset validation engine based on a generative language model, and the validation results output by the validation engine are obtained. The validation results include: the result of consistency detection of the business trigger information and the target presentation content based on the cross-scenario information validation logic, and / or the result of correlation detection of the target presentation content and the other presentation content based on the cross-scenario information validation logic.

2. The method as described in claim 1, obtaining business trigger information submitted by a user to any one of multiple business modules, and the business output result presented by the any one business module on the corresponding page, specifically includes: Data collection is performed on the graphic and text data displayed on the front end of this business module, as well as the graphic and text data loaded in the data layer of this business module. Based on the status identifiers carried by each collected graphic and text data, the service trigger information and the service output results are determined in each graphic and text data. Furthermore, based on the scene identifiers carried by each graphic and text data corresponding to the service output results, the presentation content corresponding to different service scenarios is determined.

3. The method as described in claim 1, wherein the plurality of business modules are business modules within the target application; Before obtaining the service trigger information submitted by the user to any of the multiple service modules, the method further includes: Determine the application type corresponding to the target application, and construct an information presentation environment that matches the application type; Obtaining business trigger information submitted by a user to any one of multiple business modules, and the business output result presented by that business module on the corresponding page, specifically including: The system determines the first page source data on which the service trigger information is generated on the user's end, generates the service trigger information in the information presentation environment based on the first page source data, and determines the second page source data associated with the first page source data, generates the service output result in the information presentation environment based on the second page source data.

4. The method as described in claim 3, further comprising, before obtaining the service trigger information submitted by the user to any one of the multiple service modules: Based on a preset information interception protocol, information windows unrelated to the target application in the user terminal are detected and closed. Also, based on preset element filtering rules, dynamic elements in the presentation page corresponding to the business trigger information and / or the business output result are filtered.

5. The method as described in claim 1, wherein determining the consistency detection result of the cross-scenario information verification logic on the business trigger information and the business output result specifically includes: The text consistency detection result is determined based on the semantic similarity between the text information corresponding to the business trigger information and the text information corresponding to the target presentation content; The image-text consistency detection result is determined based on the matching degree between the text information and image information corresponding to the business trigger information and the matching degree between the text information and image information corresponding to the target presentation content; The image consistency detection result is determined based on the similarity between the image information corresponding to the service trigger information and the image information corresponding to the target presentation content. Based on the text consistency detection result, the image-text consistency detection result, and the image consistency detection result, the consistency detection result of the service trigger information and the service output result is determined.

6. The method as described in claim 1, wherein the verification result further includes: The result of quality detection of the information content of the business trigger information and the business output result based on the cross-scenario information verification logic; The determination of the quality inspection results of the cross-scenario information verification logic on the information content of the business trigger information and the business output result specifically includes: Determine whether the text content corresponding to the business triggering information and the business output result meets the preset text quality, and obtain the text quality detection result; Determine whether the image content corresponding to the service triggering information and the service output result meets the preset image quality, and obtain the image quality detection result; Based on the text quality detection results and the image quality detection results, the results of quality detection of the information content of the service triggering information and the service output results are determined.

7. The method of claim 6, further comprising: Based on the consistency detection results of the business triggering information and the business output results, and the information quality detection results of the business triggering information and the business output results, the risk level corresponding to the business module is determined. Based on the risk level, a target risk management strategy is determined, and the business module is then managed using the target risk management strategy.

8. The method of claim 1, further comprising: Obtain the misjudged sample data of the content detection model and determine the actual tag information corresponding to the misjudged sample data; wherein, the misjudged sample data includes: business trigger information and business output results that actually conform to the information verification logic but are judged not to conform to the information verification logic, and / or business trigger information and business output results that actually do not conform to the information verification logic but are judged to conform to the information verification logic; Based on the misjudged sample data and the information verification logic, a prompt word is constructed, the prompt word is input into the verification engine, and the verification result to be corrected is obtained from the output of the verification engine; The loss value is determined based on the deviation between the detection result to be corrected and the actual label information, and the parameters of the verification engine are adjusted based on the loss value.

9. A content detection device for cross-business scenarios, comprising: The acquisition module is used to acquire business trigger information submitted by the user to any one of the multiple business modules, as well as the business output result presented by the business module on the corresponding page; wherein, the business output result includes the target presentation content of the business scenario corresponding to the business module and other presentation content of the business scenario corresponding to other business modules. The module is used to determine the cross-scenario information verification logic uniformly formulated for the business scenarios of the multiple business modules, and to construct prompt words by combining the business trigger information and the business output results; The detection module is used to input the prompt words into a preset validation engine based on a generative language model and obtain the validation results output by the validation engine. The validation results include: the result of consistency detection of the business trigger information and the target presentation content based on the cross-scenario information validation logic, and / or the result of correlation detection of the target presentation content and the other presentation content based on the cross-scenario information validation logic.

10. An electronic device, comprising: processor; A memory for storing processor-executable instructions; wherein the processor implements the steps of the method as described in any one of claims 1-8 by executing the executable instructions.

11. A computer-readable storage medium having stored thereon computer instructions that, when executed by a processor, implement the steps of the method as claimed in any one of claims 1-8.

12. A computer program product comprising a computer program / instructions that, when executed by a processor, implement the steps of the method as claimed in any one of claims 1-8.

Images