Risk identification method, electronic device, storage medium and program product
By analyzing the correlation between lyrics and descriptive text using a generative artificial intelligence model, the problem of inaccurate identification of public opinion risks in existing technologies has been solved. This enables accurate identification of complex semantics such as metaphors and homophones in songs, improving the accuracy and efficiency of risk identification.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-01
- Publication Date
- 2026-03-13
AI Technical Summary
Existing technologies cannot accurately identify complex semantic expressions such as metaphors, homophones, and irony when identifying public opinion risks in songs, resulting in inaccurate risk identification.
Using a generative artificial intelligence model, the system analyzes the correlation between descriptive text and media data through deep semantic understanding, and combines time conditions and prompt word templates to determine whether the lyrics have any violation risks.
It improves the accuracy of public opinion risk identification, enabling more accurate identification of complex semantic expressions such as metaphors, homophones, and irony in lyrics, reducing misjudgments, and improving the efficiency and timeliness of risk identification.
Smart Images
Figure CN121658675A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of computer technology, and in particular to a risk identification method, electronic device, storage medium, and program product. Background Technology
[0002] Against the backdrop of the explosive growth of digital music content, content platforms need to effectively identify risks associated with songs to prevent the spread of illegal content. Risk identification involves various types of violations, including general violations such as the use of uncivilized language, as well as public opinion-related violations such as negative guidance based on trending social events (e.g., distorted values).
[0003] Existing technologies typically use keyword retrieval for content review. If the lyrics of a song contain content that matches the keywords, it is determined that there is a risk of violation and can be marked for further manual review.
[0004] Because violations related to public opinion often employ complex semantic expressions such as metaphors, homophones, and irony—for example, spreading negative information through puns or allusions to social events—it is impossible to distinguish these contents using only keywords, leading to inaccurate risk identification of violations related to public opinion. Summary of the Invention
[0005] This application provides a risk identification method, electronic device, storage medium, and program product, which can solve the problem of poor accuracy in reviewing song-related public opinion risks. The technical solution is as follows: On the one hand, a risk identification method is provided, the method comprising: Obtain a set of descriptive texts, wherein the set of descriptive texts includes at least one descriptive text, and the at least one descriptive text is generated based on at least one social event information in a social event information database; Based on a generative artificial intelligence model, target media data that is associated with the target descriptive text in the descriptive text set is identified; the target media data is media data in a media database; the media database includes at least one type of media data. Based on a generative artificial intelligence model, the target media data and the target descriptive text are processed to obtain the violation risk analysis results of the target media data.
[0006] In one possible implementation, the step of determining target media data associated with the target descriptive text in the descriptive text set based on a generative artificial intelligence model; the target media data is media data in a media database; the media database includes at least one type of media data, including: From at least one descriptive text included in the descriptive text set and at least one media data included in the media database, obtain the descriptive text to be identified and the media data to be identified; The descriptive text to be identified and the media data to be identified are input into a generative artificial intelligence model to obtain the correlation analysis results output by the generative artificial intelligence model. If the correlation analysis results indicate that there is a correlation, the descriptive text to be identified and the media data to be identified are determined as target descriptive text and target media data that are correlated.
[0007] In another possible implementation, the descriptive text includes keywords related to the social event information; The step of obtaining the descriptive text to be identified and the media data to be identified from at least one descriptive text included in the descriptive text set and at least one media data included in the media database includes: In the at least one descriptive text included in the descriptive text set and at least one media data included in the media database, if a keyword in the first descriptive text matches the first media data that satisfies a first time condition, then the first descriptive text and the first media data are acquired as the descriptive text to be identified and the media data to be identified, wherein the first time condition is the time difference from the current time point within a first value range.
[0008] In another possible implementation, the descriptive text includes a summary of the corresponding social event information; The step of obtaining the descriptive text to be identified and the media data to be identified from at least one descriptive text included in the descriptive text set and at least one media data included in the media database includes: In the at least one descriptive text included in the descriptive text set and at least one media data included in the media database, if the summary in the second descriptive text matches the second media data that satisfies the second time condition, then the second descriptive text and the second media data are acquired as the descriptive text to be identified and the media data to be identified, wherein the second time condition is the time difference from the current time point within a second value range.
[0009] In another possible implementation, obtaining the descriptive text to be identified and the media data to be identified from at least one descriptive text included in the descriptive text set and at least one media data included in the media database includes: Among at least one descriptive text included in the descriptive text set and at least one media data included in the media database, a third descriptive text that satisfies a third time condition and a third media data that satisfies a fourth time condition are obtained as the descriptive text to be identified and the media data to be identified, wherein the third time condition is that the time difference with the current time point is within a third value range, and the fourth time condition is that the time difference with the current time point is within a fourth value range.
[0010] In another possible implementation, the process of processing the target media data and the target descriptive text based on a generative artificial intelligence model to obtain the violation risk analysis results of the target media data includes: The target media data and the target description text are inserted into the prompt word template to obtain model prompt words. The model prompt words are used to instruct the generative artificial intelligence model to output the violation risk analysis results for the target media data. The violation risk analysis results are used to indicate whether the target media data has any violation risk. The model prompts are input into the generative artificial intelligence model to obtain the violation risk analysis results output by the generative artificial intelligence model.
[0011] In another possible implementation, the media database includes a song library, and the media data includes lyrics text.
[0012] On the other hand, a risk identification device is provided, the device comprising: The acquisition module is configured to acquire a set of descriptive texts, wherein the set of descriptive texts includes at least one descriptive text, which is generated based on at least one social event information in a social event information database; The determination module is configured to determine target media data associated with target descriptive texts in the descriptive text set based on a generative artificial intelligence model; the target media data is media data in a media database; the media database includes at least one type of media data. The identification module is configured to process the target media data and the target descriptive text based on a generative artificial intelligence model to obtain the violation risk analysis results of the target media data.
[0013] In one possible implementation, the determining module is used to: From at least one descriptive text included in the descriptive text set and at least one media data included in the media database, obtain the descriptive text to be identified and the media data to be identified; The descriptive text to be identified and the media data to be identified are input into a generative artificial intelligence model to obtain the correlation analysis results output by the generative artificial intelligence model. If the correlation analysis results indicate that there is a correlation, the descriptive text to be identified and the media data to be identified are determined as target descriptive text and target media data that are correlated.
[0014] In another possible implementation, the determining module is further configured to: In the at least one descriptive text included in the descriptive text set and at least one media data included in the media database, if a keyword in the first descriptive text matches the first media data that satisfies a first time condition, then the first descriptive text and the first media data are acquired as the descriptive text to be identified and the media data to be identified, wherein the first time condition is the time difference from the current time point within a first value range.
[0015] In another possible implementation, the determining module is further configured to: In the at least one descriptive text included in the descriptive text set and at least one media data included in the media database, if the summary in the second descriptive text matches the second media data that satisfies the second time condition, then the second descriptive text and the second media data are acquired as the descriptive text to be identified and the media data to be identified, wherein the second time condition is the time difference from the current time point within a second value range.
[0016] In another possible implementation, the determining module is further configured to: Among at least one descriptive text included in the descriptive text set and at least one media data included in the media database, a third descriptive text that satisfies a third time condition and a third media data that satisfies a fourth time condition are obtained as the descriptive text to be identified and the media data to be identified, wherein the third time condition is that the time difference with the current time point is within a third value range, and the fourth time condition is that the time difference with the current time point is within a fourth value range.
[0017] In another possible implementation, the identification module is further used for: The target media data and the target description text are inserted into the prompt word template to obtain model prompt words. The model prompt words are used to instruct the generative artificial intelligence model to output the violation risk analysis results for the target media data. The violation risk analysis results are used to indicate whether the target media data has any violation risk. The model prompts are input into the generative artificial intelligence model to obtain the violation risk analysis results output by the generative artificial intelligence model.
[0018] On the other hand, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the program to implement the method described in any of the above.
[0019] On the other hand, a non-transitory computer-readable storage medium is provided, the non-transitory computer-readable storage medium storing computer instructions for causing a computer to perform the method described in any of the preceding claims.
[0020] On the other hand, a computer program product is provided, including computer program instructions that, when run on a computer, cause the computer to perform the method described in any of the preceding claims.
[0021] The beneficial effects of the technical solution provided in this application are: through a generative artificial intelligence model with deep semantic understanding capabilities, it is possible to identify complex semantic expressions such as metaphors, homophones, and irony in media data, and to more accurately determine whether media data is related to specific social events and whether it has a negative impact, thereby improving the accuracy of identifying public opinion risks in media data. Attached Figure Description
[0022] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0023] Figure 1 This is a schematic diagram of an implementation environment provided in an embodiment of this application; Figure 2 This is a flowchart of the risk identification method provided in the embodiments of this application; Figure 3 This is a flowchart of a risk identification method provided in another embodiment of this application; Figure 4 This is a flowchart of the risk identification method provided in the embodiments of this application; Figure 5 This is a timeline diagram of the risk identification method provided in the embodiments of this application; Figure 6 This is a schematic diagram of the risk identification device structure provided in the embodiments of this application; Figure 7 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Detailed Implementation
[0024] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.
[0025] This disclosure provides a method for risk identification. This method can be applied to a terminal. For example... Figure 1 As shown, the terminal may include a processor 110, a memory 120, and a communication component 130.
[0026] Processor 110 can be a central processing unit (CPU), graphics processing unit (GPU), microcontroller unit (MCU), accelerated processing unit (APU), neural processing unit (NPU), tensor processing unit (TPU), field-programmable gate array (FPGA), application-specific integrated circuit (ASIC), digital signal processor (DSP), etc. Processor 110 can be used to acquire descriptive text sets, run generative artificial intelligence models, and so on.
[0027] Memory 120 may include volatile memory and / or non-volatile memory. Volatile memory may include random access memory (RAM), static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), etc. Non-volatile memory may include read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), flash memory, non-volatile random access memory (NVRAM), etc. Memory 120 can be used to store descriptive text sets, social event information, and media data, etc.
[0028] The communication component 130 can be a wireless communication module (WCM), a subscriber identity module (SIM), a universal subscriber identity module (USIM), an optical network unit (ONU), etc. The communication component 130 can be used for communication between the terminal and a server, or other terminals, such as a server running a generative artificial intelligence model.
[0029] This application provides a risk identification method, such as Figure 2 As shown, in some embodiments, the method includes: S201. Obtain a set of descriptive texts, wherein the set of descriptive texts includes at least one descriptive text, and the at least one descriptive text is generated based on at least one social event information in a social event information database.
[0030] The social event information database refers to a collection storing information on social events, typically related to current social hot topics (such as news reports and commentaries), i.e., public opinion information. This information usually originates from public channels such as news websites, social media platforms, and forums, and includes diverse information such as event titles, detailed reports, occurrence times, and involved parties, ensuring the analysis foundation covers current social concerns. Furthermore, social event information can be manually summarized based on recent hot social events to ensure the database's completeness. Descriptive texts are generated based on social event information and include keywords and summaries. The process involves using natural language processing technology to extract keywords or phrases that best represent the core content of the social event information text, summarizing the text to form concise summaries that retain richer contextual semantics. The descriptive text set is a collection of one or more descriptive texts generated based on different social event information. The social event information database is typically dynamically updated, periodically (according to a preset time period, such as every hour) or in real-time, storing the latest hot event information, thus updating the descriptive text set accordingly.
[0031] S202. Based on a generative artificial intelligence (AI) model, determine target media data that is associated with the target description text in the description text set; the target media data is media data in a media database; the media database includes at least one type of media data.
[0032] The media database includes, but is not limited to, song libraries and video libraries. If the media database is a song library, the media data is lyrics text. If the media database is a video library, the media data can be subtitle text of videos in the video library, or text in each frame of the video, or text generated by a visual language model to describe the video content. Generative artificial intelligence models are based on deep learning architectures and learn complex patterns in data by training on massive datasets. They have deep semantic understanding capabilities, such as Large Language Models (LLM) or Transformer architecture models, which can understand the deep semantic relationships between descriptive text and lyrics text, thereby identifying related target descriptive text and target media data. In the at least one descriptive text included in the descriptive text set and at least one media data included in the media database, any one descriptive text and any one media data are input into the generative artificial intelligence model to obtain the correlation analysis result output by the generative artificial intelligence model, such as: "Lyrics summary: <100-word main idea>, Relevance: <Relevant / Irrelevant>, Reason for correlation: <Detailed reason or 'None'>". If the correlation analysis result indicates that there is a correlation, then the descriptive text and the media data are identified as target descriptive text and target media data that are correlated.
[0033] S203. Based on a generative artificial intelligence model, the target media data and the target descriptive text are processed to obtain the violation risk analysis results of the target media data.
[0034] In specific implementation, the target media data and the target description text are inserted into the prompt word template to obtain model prompt words. The model prompt words are used to instruct the generative artificial intelligence model to combine the target description text and the target media data to analyze the violation risks of the target media data from the perspectives of negative social impact, illegality and prohibition, etc., and thus output the violation risk analysis results for the target media data. After inputting the model prompt words into the generative artificial intelligence model, the violation risk analysis results output by the generative artificial intelligence model are obtained, such as: "Violation category: <negative social impact, illegality and prohibition, etc., violation level: <no risk / low risk / high risk>, original violation text: [violation lyrics fragment], violation reason: <detailed reason>".
[0035] In this embodiment, the generative AI model can understand the deeper meaning of text, going beyond mere word matching. It can semantically associate implicit allusions and metonymy in lyrics with descriptive texts of specific social events, thereby identifying content that deliberately circumvents traditional review processes. The generative AI model possesses strong contextual understanding capabilities; when analyzing a lyric, it considers the preceding and following lyrics, as well as the overall tone of the song, to make a comprehensive judgment, reducing misjudgments caused by misinterpretation. By establishing a dynamically updated social event information database and generating corresponding descriptive text sets, it can quickly respond to constantly changing social hotspots, keeping risk identification synchronized with public opinion development, and promptly detecting and preventing song content that uses the latest hot events for illegal allusions.
[0036] like Figure 3 As shown, in some embodiments, the step of determining target media data associated with the target descriptive text in the descriptive text set based on a generative artificial intelligence model; the target media data is media data in a media database; the media database includes at least one type of media data, including: S301. Obtain the description text to be identified and the media data to be identified from at least one description text included in the description text set and at least one media data included in the media database.
[0037] In practice, the descriptive text to be identified and the media data to be identified are first obtained. The descriptive text to be identified can be any descriptive text in the descriptive text set, and the media data to be identified can be any media data in the media database. If there are N descriptive texts in the descriptive text set and M media data in the media database, then there are N×M combinations of descriptive text and media data to be identified.
[0038] S302. Input the description text to be identified and the media data to be identified into the generative artificial intelligence model to obtain the correlation analysis result output by the generative artificial intelligence model. If the correlation analysis result indicates that there is a correlation, then the description text to be identified and the media data to be identified are determined as target description text and target media data that are correlated.
[0039] In practice, the descriptive text and media data to be identified are inserted into corresponding prompt word templates to obtain model prompt words. These prompt words instruct the generative AI model to output a correlation analysis result between the descriptive text and the media data. Specifically, the model is instructed to evaluate the deep semantic correlation between the descriptive text and the media data from six dimensions: event, metaphor, emotional resonance, detail mapping, theme, and main idea. If at least four dimensions are correlated, the descriptive text and the media data are considered correlated. If there is a clear semantic opposition or conflict in the theme or main idea dimension, even if other dimensions are correlated, the descriptive text and the media data are directly determined to be uncorrelated to avoid misjudgment. After inputting the model prompt words into the generative AI model, the correlation analysis result output by the model is obtained. If the correlation analysis result indicates a correlation, the descriptive text and the media data are identified as correlated target descriptive text and target media data.
[0040] In this embodiment, by traversing all possible combinations of descriptive text and media data, all possible relevances can be covered, avoiding missed judgments due to the selection of specific combinations. Generative AI models possess deep semantic understanding capabilities, enabling them to analyze the relationship between descriptive text and media data from multiple dimensions (such as events, metaphors, emotional resonance, etc.). By comprehensively evaluating these dimensions, the model can more accurately determine whether a relationship exists, reducing the possibility of misjudgment.
[0041] In some embodiments, such as Figure 4 As shown, the descriptive text includes keywords corresponding to social event information; the step of obtaining the descriptive text to be identified and the media data to be identified from at least one descriptive text included in the descriptive text set and at least one media data included in the media database includes: In the at least one descriptive text included in the descriptive text set and at least one media data included in the media database, if a keyword in the first descriptive text matches the first media data that satisfies a first time condition, then the first descriptive text and the first media data are acquired as the descriptive text to be identified and the media data to be identified, wherein the first time condition is the time difference from the current time point within a first value range.
[0042] In practical implementation, the first time condition is that the time difference with the current time point is within a first range, such as... Figure 5As shown on the timeline, the first value range can be set to 1 to 12 months. Therefore, the first media data meeting the first time condition refers to media data whose entry time is more than one month but less than 12 months from the current time. The first value range can also be determined based on the entry time of the earliest media data in the media database. For example, if the time difference between the earliest entry time and the current time is two years, the first value range is 1-24 months. If keywords in the first description text are found in the first media data (lyrics text) meeting the first time condition, then the keywords in the first description text match the first media data meeting the first time condition. Therefore, the first description text and the first media data are obtained as the description text and media data to be identified. Furthermore, for the song library, before keyword matching, the library can be filtered by song language and release source. For example, Chinese songs and songs from high-risk releasers (such as personally uploaded songs) can be filtered out, as these songs are more likely to contain negative content. Then, keyword matching can be performed, reducing computational resource consumption.
[0043] In this embodiment, by setting a first-time condition, all media data can be analyzed, thereby responding more comprehensively to changes in public opinion. When the descriptive text contains keywords, matching it with media data under the first-time condition can more quickly find content related to current hot topics. This effectively improves the efficiency, accuracy, and timeliness of risk identification, while reducing computational resource consumption and the possibility of misjudgment.
[0044] In some embodiments, such as Figure 4 As shown, the descriptive text includes a summary of the corresponding social event information; the step of obtaining the descriptive text to be identified and the media data to be identified from at least one descriptive text included in the descriptive text set and at least one media data included in the media database includes: In the at least one descriptive text included in the descriptive text set and at least one media data included in the media database, if the summary in the second descriptive text matches the second media data that satisfies the second time condition, then the second descriptive text and the second media data are acquired as the descriptive text to be identified and the media data to be identified, wherein the second time condition is the time difference from the current time point within a second value range.
[0045] In practical implementation, the second time condition is the time difference from the current time point within a second value range, such as... Figure 5As shown on the timeline, the second value range can be set to 1 to 30 days. Therefore, the second media data meeting the second time condition refers to media data whose entry time is more than 1 day but less than 30 days from the current time. If the summary in the second description text has a high text similarity to the text of the second media data meeting the second time condition, then the summary in the second description text matches the second media data meeting the second time condition. Specifically, if the second media data is lyrics text, the text vectors of the summary and the lyrics text can be determined, and the cosine similarity between the text vectors of the summary and the lyrics text can be calculated. When the cosine similarity is greater than or equal to a preset cosine similarity (e.g., 0.8), it can be considered that the summary in the second description text has a high text similarity to the lyrics text meeting the second time condition, and the summary matches the lyrics text. Therefore, the second description text and the second media data (lyrics text) are obtained as the description text and media data to be identified. In addition, for the song library, before matching by keywords, the song library can be filtered by language and release source. For example, Chinese songs and songs released by high-risk publishers (such as songs uploaded by individuals) can be filtered out, as these songs are more likely to contain negative content. Then, matching can be performed by summary, which can reduce the consumption of computing resources.
[0046] In this embodiment, by setting a second time condition, priority can be given to media data very close to the current time, avoiding the processing of outdated data and thus responding more promptly to changes in public opinion. Using cosine similarity between the summary and lyrics text for matching can find highly relevant media data based on semantic relevance, capturing deep semantic connections, reducing the possibility of misjudgment, and improving the accuracy of association. By limiting the time range of the media data, the amount of media data requiring similarity calculation and analysis is reduced, improving processing efficiency and reducing the consumption of computing resources.
[0047] In some embodiments, such as Figure 4 As shown, the step of obtaining the descriptive text to be identified and the media data to be identified from at least one descriptive text included in the descriptive text set and at least one media data included in the media database includes: In the at least one descriptive text included in the descriptive text set and the at least one media data included in the media database, a third descriptive text that satisfies a third time condition and a third media data that satisfies a fourth time condition are obtained as the descriptive text to be identified and the media data to be identified. The third time condition is that the time difference with the current time point is within a third value range, and the fourth time condition is that the time difference with the current time point is within a fourth value range. The third value range and the fourth value range overlap.
[0048] In specific implementation, the third time condition is that the time difference with the current time point is within the third value range, the third value range overlaps with the fourth value range, and the lower limit of both the third value range and the fourth value range is 0. For example... Figure 5 As shown on the timeline, the third value range can be set to 0-30 days. Therefore, the third descriptive text satisfying the third time condition refers to descriptive text whose entry time is within 30 days of the current time. The fourth time condition is that the time difference from the current time is within the fourth value range. The fourth time condition can be set to 0-1 day. Therefore, the third media data satisfying the fourth time condition refers to media data whose entry time is within 1 day of the current time, such as lyrics text entered on the same day. From at least one descriptive text included in the descriptive text set and at least one media data included in the media database, the third descriptive text satisfying the third time condition and the third media data satisfying the fourth time condition are obtained. Each third descriptive text is matched with each third media data. For example, if there are A third descriptive texts and B third media data, there are A×B combinations. The matched third descriptive texts and third media data are used as the descriptive text and media data to be identified. Alternatively, one can match A third-party descriptive texts with each third-party media data point, meaning one third-party media data point corresponds to A third-party descriptive texts, resulting in B combinations. A third-party descriptive texts and one matching third-party media data point can be used as the descriptive text and media data to be identified. Furthermore, for the song library, before keyword matching, it can be filtered by language and distribution source. For example, Chinese songs and songs from high-risk distributors (such as personally uploaded songs) can be filtered out, as these songs are more likely to contain negative content. Matching these songs after filtering can reduce computational resource consumption.
[0049] In this embodiment, the third and fourth time conditions limit the time range of the descriptive text and media data, respectively. The third descriptive text is limited to the last 30 days, ensuring that the acquired data represents current or recently occurring trending events. The third media data is limited to the last day, ensuring that the acquired data represents the latest published content. This makes the matching process more focused on real-time performance, enabling timely capture of the latest media content related to current trending events. By limiting the time range, the amount of data that needs to be processed is reduced, making the matching process more efficient and targeted. By combining generative artificial intelligence models, a deeper semantic understanding and correlation analysis can be performed on the matched content, enabling more accurate identification of media content related to current trending events. This enhances the responsiveness to the latest trending events and provides strong support for application scenarios such as content review and public opinion management.
[0050] All of the above-mentioned optional technical solutions can be combined in any way to form optional embodiments of this disclosure, and will not be described in detail here.
[0051] Based on the same inventive concept, corresponding to the risk identification method provided in the embodiments of this application, this application also provides a risk identification method apparatus.
[0052] refer to Figure 6 The risk identification method apparatus includes: The acquisition module 601 is configured to acquire a set of descriptive texts, wherein the set of descriptive texts includes at least one descriptive text, which is generated based on at least one social event information in a social event information database; The determination module 602 is configured to determine, based on a generative artificial intelligence model, target media data that is associated with the target descriptive text in the descriptive text set; the target media data is media data in a media database; the media database includes at least one type of media data. The identification module 603 is configured to process the target media data and the target descriptive text based on a generative artificial intelligence model to obtain the violation risk analysis results of the target media data.
[0053] In one possible implementation, the determining module 602 is used to: From at least one descriptive text included in the descriptive text set and at least one media data included in the media database, obtain the descriptive text to be identified and the media data to be identified; The descriptive text to be identified and the media data to be identified are input into a generative artificial intelligence model to obtain the correlation analysis results output by the generative artificial intelligence model. If the correlation analysis results indicate that there is a correlation, the descriptive text to be identified and the media data to be identified are determined as target descriptive text and target media data that are correlated.
[0054] In another possible implementation, the determining module 602 is further configured to: In the at least one descriptive text included in the descriptive text set and at least one media data included in the media database, if a keyword in the first descriptive text matches the first media data that satisfies a first time condition, then the first descriptive text and the first media data are acquired as the descriptive text to be identified and the media data to be identified, wherein the first time condition is the time difference from the current time point within a first value range.
[0055] In another possible implementation, the determining module 602 is further configured to: In the at least one descriptive text included in the descriptive text set and at least one media data included in the media database, if the summary in the second descriptive text matches the second media data that satisfies the second time condition, then the second descriptive text and the second media data are acquired as the descriptive text to be identified and the media data to be identified, wherein the second time condition is the time difference from the current time point within a second value range.
[0056] In another possible implementation, the determining module 602 is further configured to: Among at least one descriptive text included in the descriptive text set and at least one media data included in the media database, a third descriptive text that satisfies a third time condition and a third media data that satisfies a fourth time condition are obtained as the descriptive text to be identified and the media data to be identified, wherein the third time condition is that the time difference with the current time point is within a third value range, and the fourth time condition is that the time difference with the current time point is within a fourth value range.
[0057] In another possible implementation, the identification module 603 is further used for: The target media data and the target description text are inserted into the prompt word template to obtain model prompt words. The model prompt words are used to instruct the generative artificial intelligence model to output the violation risk analysis results for the target media data. The violation risk analysis results are used to indicate whether the target media data has any violation risk. The model prompts are input into the generative artificial intelligence model to obtain the violation risk analysis results output by the generative artificial intelligence model.
[0058] It should be noted that the risk identification device provided in the above embodiments is only illustrated by the division of the above functional modules when performing risk identification. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the risk identification device and the risk identification method embodiments provided in the above embodiments belong to the same concept, and their specific implementation process can be found in the method embodiments, which will not be repeated here.
[0059] Based on the same inventive concept, corresponding to the risk identification method provided in the embodiments of this application, this application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the risk identification method described in the above embodiments.
[0060] Figure 7This embodiment illustrates a more specific hardware structure of an electronic device, which may include a processor 1010, a memory 1020, an input / output interface 1030, a communication interface 1040, and a bus 1050. The processor 1010, memory 1020, input / output interface 1030, and communication interface 1040 are interconnected internally via the bus 1050.
[0061] The processor 1010 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this specification.
[0062] The memory 1020 can be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 1020 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented by software or firmware, the relevant program code is stored in the memory 1020 and is called and executed by the processor 1010.
[0063] The input / output interface 1030 is used to connect input / output modules to realize information input and output. Input / output modules can be configured as components within the device (not shown in the figure) or externally connected to the device to provide corresponding functions. Input devices may include keyboards, mice, touchscreens, microphones, various sensors, etc., while output devices may include displays, speakers, vibrators, indicator lights, etc.
[0064] The communication interface 1040 is used to connect a communication module (not shown in the figure) to enable communication between this device and other devices. The communication module can communicate via wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).
[0065] Bus 1050 includes a pathway for transmitting information between various components of the device, such as processor 1010, memory 1020, input / output interface 1030, and communication interface 1040.
[0066] It should be noted that although the above-described device only shows the processor 1010, memory 1020, input / output interface 1030, communication interface 1040, and bus 1050, in specific implementations, the device may also include other components necessary for normal operation. Furthermore, those skilled in the art will understand that the above-described device may only include the components necessary for implementing the embodiments of this specification, and not necessarily all the components shown in the figures.
[0067] The electronic devices described above are used to implement the corresponding risk identification methods in the foregoing embodiments and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.
[0068] In an exemplary embodiment, a computer-readable storage medium is also provided, such as a memory including instructions that can be executed by a processor in a terminal to perform the risk identification method described above. This computer-readable storage medium may be non-transitory. For example, the computer-readable storage medium may be ROM (Read-Only Memory), RAM (Random Access Memory), CD-ROM (Compact Disc Read-Only Memory), magnetic tape, floppy disk, and optical data storage devices, etc.
[0069] In an exemplary embodiment, a computer program product is also provided, including computer program instructions that, when executed on a computer, cause the computer to perform the risk identification method described above.
[0070] It should be noted that the information (including but not limited to user equipment information, user personal information, etc.), data (including but not limited to data used for analysis, data stored, data displayed, etc.) and signals (including but not limited to signals transmitted between user terminals and other devices, etc.) involved in this application are all authorized by the user or fully authorized by all parties, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions.
[0071] Those skilled in the art will understand that all or part of the steps of the above embodiments can be implemented by hardware or by a program instructing related hardware. The program can be stored in a computer-readable storage medium, such as a read-only memory, a disk, or an optical disk.
[0072] It should be understood that "multiple" as used herein refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. Furthermore, the step numbers described herein are merely illustrative of one possible execution order. In some other embodiments, the steps may not be executed in numerical order, such as two steps with different numbers being executed simultaneously, or two steps with different numbers being executed in the reverse order of the illustration. This application does not limit this.
[0073] The above description is merely an optional embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.
Claims
1. A risk identification method, characterized in that, include: Obtain a set of descriptive texts, wherein the set of descriptive texts includes at least one descriptive text, and the at least one descriptive text is generated based on at least one social event information in a social event information database; Based on a generative artificial intelligence model, target media data that is associated with the target descriptive text in the descriptive text set is identified; the target media data is media data in a media database; the media database includes at least one type of media data. Based on a generative artificial intelligence model, the target media data and the target descriptive text are processed to obtain the violation risk analysis results of the target media data.
2. The risk identification method according to claim 1, characterized in that, The method for determining target media data associated with target descriptive texts in the descriptive text set based on a generative artificial intelligence model includes: From at least one descriptive text included in the descriptive text set and at least one media data included in the media database, obtain the descriptive text to be identified and the media data to be identified; The descriptive text to be identified and the media data to be identified are input into a generative artificial intelligence model to obtain the correlation analysis results output by the generative artificial intelligence model. If the correlation analysis results indicate that there is a correlation, the descriptive text to be identified and the media data to be identified are determined as target descriptive text and target media data that are correlated.
3. The risk identification method according to claim 2, characterized in that, The descriptive text includes keywords related to the social event information; The step of obtaining the descriptive text to be identified and the media data to be identified from at least one descriptive text included in the descriptive text set and at least one media data included in the media database includes: In the at least one descriptive text included in the descriptive text set and at least one media data included in the media database, if a keyword in the first descriptive text matches the first media data that satisfies a first time condition, then the first descriptive text and the first media data are acquired as the descriptive text to be identified and the media data to be identified, wherein the first time condition is the time difference from the current time point within a first value range.
4. The risk identification method according to claim 2, characterized in that, The descriptive text includes a summary of the corresponding social event information; The step of obtaining the descriptive text to be identified and the media data to be identified from at least one descriptive text included in the descriptive text set and at least one media data included in the media database includes: In the at least one descriptive text included in the descriptive text set and at least one media data included in the media database, if the summary in the second descriptive text matches the second media data that satisfies the second time condition, then the second descriptive text and the second media data are acquired as the descriptive text to be identified and the media data to be identified, wherein the second time condition is the time difference from the current time point within a second value range.
5. The risk identification method according to claim 2, characterized in that, The step of obtaining the descriptive text to be identified and the media data to be identified from at least one descriptive text included in the descriptive text set and at least one media data included in the media database includes: In the at least one descriptive text included in the descriptive text set and the at least one media data included in the media database, a third descriptive text that satisfies a third time condition and a third media data that satisfies a fourth time condition are obtained as the descriptive text to be identified and the media data to be identified. The third time condition is that the time difference with the current time point is within a third value range, and the fourth time condition is that the time difference with the current time point is within a fourth value range. The third value range and the fourth value range overlap.
6. The risk identification method according to claim 1, characterized in that, The generative artificial intelligence model processes the target media data and the target descriptive text to obtain the violation risk analysis results of the target media data, including: The target media data and the target description text are inserted into the prompt word template to obtain model prompt words. The model prompt words are used to instruct the generative artificial intelligence model to output the violation risk analysis results for the target media data. The violation risk analysis results are used to indicate whether the target media data has any violation risk. The model prompts are input into the generative artificial intelligence model to obtain the violation risk analysis results output by the generative artificial intelligence model.
7. The risk identification method according to any one of claims 1-6, characterized in that, The media database includes a song library, and the media data includes lyrics text.
8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the program, it implements the method as described in any one of claims 1 to 7.
9. A non-transitory computer-readable storage medium storing computer instructions, characterized in that, The computer instructions are used to cause the computer to perform the method described in any one of claims 1 to 7.
10. A computer program product comprising computer program instructions, characterized in that, When the computer program instructions are executed on a computer, the computer causes the computer to perform the method as described in any one of claims 1 to 7.