Video Generation Method and System Based on Artificial Intelligence and Hotspot Monitoring

Abstract

This invention discloses a video generation method and system based on artificial intelligence and hot topic monitoring, relating to the field of video generation technology. It includes a hot topic monitoring module, a compliance processing module, a video generation module, and a compliance verification module. By capturing and filtering hot topic information, and combining dissemination frequency and timeliness indices, it accurately identifies high-value content, achieving efficient topic selection for hot topic video creation. Through multi-layered compliance processing including false information identification, value orientation assessment, and portrait rights infringement detection, it ensures that the content is authentic, positive, and legal. By leveraging artificial intelligence to extract script elements and intelligently match or generate materials, it significantly improves video production efficiency and creative richness. Simultaneously, through multi-dimensional compliance verification and correction of the initial video, it ensures that the final video is fully compliant in terms of information authenticity, value orientation, and portrait rights protection. This provides users with high-quality video content that keeps up with hot topics and helps platforms efficiently produce compliant and high-quality works.

Description

Technical Field

[0001] This invention relates to the field of video generation technology, specifically to a video generation method and system based on artificial intelligence and hotspot monitoring. Background Technology

[0002] With the deep application of artificial intelligence technology in video generation, automated video generation based on trending topics has become an important way of information dissemination. However, while such videos are produced rapidly, they also face serious ethical and legal risks and challenges.

[0003] Regarding misinformation, trending topics online are inherently mixed, with some potentially false or exaggerated information. Current technology lacks an effective mechanism for verifying the authenticity of trending information. If videos are directly generated based on such information, it can easily lead to the secondary spread of misinformation, misleading public perception, damaging the credibility of information dissemination, and even causing social panic and other adverse consequences.

[0004] At the level of value orientation, trending information may contain negative values ​​such as violence, discrimination, and vulgarity. Current technology lacks systematic assessment methods for value orientation during video generation, making it difficult to proactively filter and adjust content with undesirable orientations. This results in generated videos that may spread negative values, violate social norms and public order, and negatively impact the formation of values, especially among teenagers.

[0005] Regarding infringements of portrait rights and reputation rights, current technology struggles to accurately detect whether a person's portrait rights or reputation rights have been infringed upon during video generation if the video involves their likeness. The lack of an effective facial feature database for comparison and authorization management mechanisms may lead to unauthorized use of another person's image or damage to their reputation through video content, potentially resulting in legal disputes and posing legal liability risks to the video generator.

[0006] In summary, existing video generation technologies based on artificial intelligence and trending topics lack effective mechanisms to avoid risks such as false information, negative value orientations, and infringements on portrait rights and reputation rights. They are unable to ensure both efficiency and compliance and social value of the content, which greatly limits the healthy and sustainable development of such technologies. Therefore, there is an urgent need for a video generation method that can systematically solve these risk problems. Summary of the Invention

[0007] (a) Technical problems to be solved

[0008] To address the shortcomings of existing technologies, this invention provides a video generation method and system based on artificial intelligence and hotspot monitoring, which solves the problems mentioned in the background section.

[0009] (II) Technical Solution

[0010] To achieve the above objectives, the present invention provides a video generation method based on artificial intelligence and hotspot monitoring, comprising the following steps:

[0011] Step 1: Monitoring and filtering of trending information:

[0012] Raw information is crawled from the network and formed into a raw information set. By determining the popularity value of each piece of raw information, a set of hot information is filtered out.

[0013] Step 2: Compliance processing of trending information:

[0014] Conducting functions such as identifying misinformation, assessing value orientations, and detecting portrait rights infringements in trending information;

[0015] Step 3: Artificial intelligence-assisted video generation:

[0016] Based on trending information, a video script is generated. By matching the descriptions of each material in the pre-built material library with the corresponding shots in the video script, existing materials are selected. At the same time, a generative artificial intelligence model is used to generate materials that are not in the material library but are required by the video script. Then, the selected and generated materials are combined according to the shot order of the script to form the initial video.

[0017] Step 4: Initial Video Compliance Verification

[0018] The initial video undergoes compliance verification, and the logic of the compliance verification process is consistent with that of the compliance processing steps for trending information.

[0019] As a further aspect of the present invention, the method for filtering hotspot information sets is as follows:

[0020] The original set of information captured is labeled as L={i1, i2, ..., i}. n}, where i k This represents the kth original message, where k = 1, 2, ..., n, and n represents the number of original messages retrieved.

[0021] At the same time, the number of times each piece of original information was disseminated and its timeliness index were extracted;

[0022] pass: Calculate the popularity value RD for each piece of original information. k ;

[0023] Among them, CB k SX represents the number of times the k-th original message has been propagated, where α is a preset weighting coefficient based on the number of propagations; k β represents the timeliness index of the k-th original information, and β is the preset weight coefficient based on the timeliness index;

[0024] Then, based on the preset heat threshold RDY, select those that meet the RD... k The set of hotspot information ≥ RDY, i.e. the set of information to be generated for the video, is denoted as L(R) = {i(R)1, i(R)2, ..., i(R)}. n0};

[0025] Where i(R) k0 Let k represent the k0th hotspot information, where k0 = 1, 2, ..., n0, and n0 represents the total number of hotspot information items in the video to be generated, where n0 ≤ n.

[0026] As a further aspect of the present invention: the number of disseminations is determined based on either the number of forwards or the number of clicks, or based on the sum of the number of forwards and the number of clicks, and the timeliness index is determined based on the interval between the information release time and the current time.

[0027] As a further aspect of the present invention, the method for identifying false information is as follows:

[0028] Extract from a pre-established database of real information T = {t1, t2, ..., t} m};

[0029] Through:

[0030] Calculate i(R) for each hotspot information. k0 Real information t in the real information database j Similarity S k0,j ;

[0031] Where j = 1, 2, ..., m, m represents the number of real information entries in the real information database. Let L(R) represent the number of elements in the intersection of sets L(R) and T. This represents the number of elements in the union of sets L(R) and T;

[0032] For each i(R) k0 :

[0033] If there exists a true piece of information t j Make S k0,j If the value is greater than or equal to SY, then the hotspot information i(R) is determined to be valid. k0 This is true information;

[0034] If all true information t j The corresponding S k0,j If <SY, then mark the hotspot information i(R). k0 This is suspected to be false information;

[0035] Where SY is the preset similarity threshold.

[0036] As a further aspect of the present invention, the value-oriented assessment method is as follows:

[0037] Construct a value keyword library V = {v1, v2, ... v} p The value keyword library includes both positive and negative value keywords.

[0038] For each hot topic information i(R) k0 Extract its keyword set W = {w1, w2, ... w} q}, where q represents hotspot information i(R) k0 The number of keywords in the text;

[0039] Subsequently passed:

[0040] Calculate the positive value matching degree P V Matching degree N with negative values V ;

[0041] Among them, Va is a subset of keywords representing positive values, and Vb is a subset of keywords representing negative values;

[0042] P V Used to measure the proportion of positive values ​​in video content; N V Used to measure the proportion of negative values ​​in video content; len(W) represents the total number of words in the word set W; Va is a subset of positive value keywords in the value keyword library; Vb is a subset of negative value keywords; the "1" in the formula is a counting marker, indicating that each time a word belonging to Va or Vb is matched in W, it is counted as 1.

[0043] This represents the total number of words in the word set W of the video content that belong to the subset Va of positive value keywords, i.e., 1 is added for each positive word matched. This represents the percentage of positive value keywords in the total vocabulary of the video; the higher the percentage, the stronger the positive value orientation of the content.

[0044] This represents the total number of words in the word set W of the video content that belong to the negative value keyword subset Vb, i.e., 1 is added for each negative word matched. This represents the percentage of negative value keywords in the total vocabulary of the video; the higher the percentage, the stronger the negative value orientation of the content, and the greater the possibility of value risks.

[0045] If N V >0.1 or P V If the value is less than 0.3, it is determined that there is a risk in the value orientation.

[0046] As a further aspect of the present invention, the method for detecting portrait rights infringement is as follows:

[0047] Establish a facial feature database F = {f1, f2, ..., f} r}, where f e These are the feature vectors of authorized portraits, e = 1, 2, ..., r;

[0048] Extract feature vector f from the face region in the generated video;

[0049] pass: Calculate the f values ​​that match all facial feature databases. e Similarity Sim(f,f) e );

[0050] Where d is the dimension of the feature vector, f z and f e,z It is the z-th component of the eigenvector;

[0051] If Sim(f, f) exists e If the value is greater than or equal to 0.7 and no authorization is obtained, there is a risk of infringement, so the face will be blurred or replaced.

[0052] As a further aspect of the present invention, the specific method for generating the initial video is as follows:

[0053] Based on hotspot information i(R) k0 Extract the core events, characters, and scene elements from the content, and generate a video script;

[0054] Simultaneously, the number of shots in the video script is marked as JT, and the description of each shot in the video script is marked as MS. x x = 1, 2, ..., JT;

[0055] For existing footage in the pre-built media library, based on the description of the video script shots, MS x The matching degree between each piece of footage and the corresponding description of the video script shot is calculated as follows:

[0056] Extract the features corresponding to the video script shots, and combine them into a feature set labeled MT={mt} x1 mt x2 ...mt xc}; and for each feature mt xu The assigned weight value is preset and marked as γ. xu Where u = 1, 2, ..., c, and c is the number of features described corresponding to the video script shot; where, ;

[0057] Simultaneously, extract the features of the relevant materials, and combine them into a feature set, which is then labeled as ST={st}. y1 st y2 ...st yz};

[0058] Where y = 1, 2, ..., SC, SC is the number of materials in the material library, and z is the number of features of the corresponding material;

[0059] For each video script shot, the corresponding feature mt is described xu :

[0060] If the corresponding material contains matching features, i.e., mt xu =st yv Then add a matching identifier PP to it. xu And set its value to 1;

[0061] If the corresponding material does not contain a matching feature, i.e., mt xu ≠st yv Then add a matching identifier PP to it. xu and set its value to 0;

[0062] Where v = 1, 2, ..., z;

[0063] Subsequently passed: Calculate the matching degree (PD) between the corresponding material and the description of the video script shot. x ;

[0064] Subsequently, PD was selected. x ≥PDY materials;

[0065] Wherein, PDY is the matching threshold between the source material and the corresponding description of the video script shot;

[0066] When the material library does not contain a PD-compliant item x When the material is ≥PDY, the material required for the video script shot is generated using a pre-built generative artificial intelligence model based on the corresponding description of the video script shot;

[0067] The selected and generated footage will then be composited according to the shot order in the script to form the initial video V0.

[0068] As a further aspect of this invention, the specific method for compliance verification is as follows:

[0069] The information in V0 is detected according to the false information identification steps. If there is suspected false information, a correction prompt is generated and fed back to the generation model to obtain the corrected video V1.

[0070] The revised video V1 was evaluated according to the value orientation assessment steps. If it was determined that there was a risk in the value orientation, the negative words in the revised video were replaced with positive words, and the visual style of the revised video was adjusted to a positive guidance style, resulting in the revised video V2.

[0071] Face detection was performed on the adjusted video V2 according to the portrait rights infringement detection steps. If there was a risk of infringement, the face was blurred or replaced to obtain the final compliant video V3.

[0072] A video generation system based on artificial intelligence and hotspot monitoring, which executes a video generation method based on artificial intelligence and hotspot monitoring, includes:

[0073] The hotspot monitoring module includes an information capture unit and a popularity calculation unit;

[0074] The information capture unit is used to capture raw information from the network and form a set of raw information;

[0075] The popularity calculation unit is used to calculate the popularity value of each piece of original information and filter out the set of hot information;

[0076] The compliance processing module includes a false information detection unit, a value orientation detection unit, and a portrait rights detection unit;

[0077] The fake information detection unit is used to identify fake information in trending information.

[0078] The value orientation detection unit is used to assess the value orientation of trending information.

[0079] The portrait rights detection unit is used to detect portrait rights infringements in trending information.

[0080] The video generation module includes a script generation unit, a material processing unit, and a video compositing unit;

[0081] The script generation unit is used to generate video scripts based on hotspot information;

[0082] The material processing unit is used to select existing materials from a pre-built material library based on the matching degree between each material and the corresponding description of the video script shots, and at the same time, it generates materials that are not in the material library but are required by the video script through a generative artificial intelligence model;

[0083] The video compositing unit is used to combine the selected and generated footage according to the shot order in the script to form the initial video;

[0084] The compliance verification module is used to perform compliance verification on the initial video in conjunction with the compliance processing module, and its verification method is consistent with the methods used for identifying false information, assessing value orientation, and detecting portrait rights infringements in hot information.

[0085] (III) Beneficial Effects

[0086] This invention provides a video generation method and system based on artificial intelligence and hotspot monitoring. Compared with existing technologies, it has the following advantages:

[0087] This invention utilizes a hot topic information monitoring and filtering mechanism, leveraging web crawlers to capture information from multiple platforms. It then combines dissemination frequency and timeliness indices to construct a popularity value calculation model, accurately identifying high-profile hot topics. This design ensures that video generation closely follows current public opinion trends, guaranteeing high attention and timeliness. It helps creators quickly grasp market demands, enhances the dissemination potential and appeal of videos, and solves the problems of delayed hot topic capture and lack of content attention in traditional video creation.

[0088] In the process of handling compliance issues related to trending information, a multi-dimensional compliance system is constructed, encompassing three aspects: false information identification, value orientation assessment, and portrait rights infringement detection. False information identification effectively mitigates the risk of spreading misinformation by comparing data with a database of authentic information; value orientation assessment utilizes quantitative analysis based on a keyword database to ensure that the content is positive and aligns with mainstream social values; and portrait rights infringement detection avoids disputes by comparing data with a facial feature database. This comprehensive approach enhances the compliance of video content and strengthens its credibility and security.

[0089] The AI-assisted video generation and initial video compliance verification process achieves both high efficiency and precision in video creation. The former, through material matching calculations and generative AI models, quickly integrates or generates script-compliant materials, improving video production efficiency. The latter performs multi-dimensional compliance verification and iterative corrections on the initial video, ensuring that the final output video is fully compliant in terms of information authenticity, value orientation, and portrait rights. This process improves video production efficiency while guaranteeing video quality, providing users with an efficient and compliant video generation solution. Attached Figure Description

[0090] Figure 1 This is a system block diagram of the video generation system based on artificial intelligence and hotspot monitoring of the present invention.

[0091] Figure 2 This is a flowchart illustrating the video generation method based on artificial intelligence and hotspot monitoring according to the present invention. Detailed Implementation

[0092] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0093] Please see Figure 1 and Figure 2 As shown, the embodiments of the present invention provide the following technical solutions:

[0094] As an embodiment of the present invention:

[0095] This invention relates to a video generation method based on artificial intelligence and hotspot monitoring, comprising the following steps:

[0096] Hot topic information monitoring and filtering steps:

[0097] Using web crawler technology to extract information from news platforms, social media and other platforms on the Internet;

[0098] The original set of information captured is then labeled as L={i1, i2, ..., i...} n}, where i k This represents the kth original message, where k = 1, 2, ..., n, and n represents the number of original messages retrieved.

[0099] Extract the number of times each piece of original information has been disseminated and its timeliness index;

[0100] The number of disseminations is determined by either the number of reposts or the number of clicks, or by the sum of the number of reposts and the number of clicks. The timeliness index is determined by the interval between the information release time and the current time, and the shorter the interval, the greater the timeliness index.

[0101] pass: Calculate the popularity value RD for each piece of original information. k ;

[0102] Among them, CB k SX represents the number of times the k-th original message has been propagated, where α is a preset weighting coefficient based on the number of propagations; k β represents the timeliness index of the k-th original information, and β is the preset weight coefficient based on the timeliness index;

[0103] Then, based on the preset heat threshold RDY, select those that meet the RD... k The set of hotspot information ≥ RDY, i.e. the set of information to be generated for the video, is denoted as L(R) = {i(R)1, i(R)2, ..., i(R)}. n0};

[0104] Where i(R) k0 This represents the k0th hotspot information, where k0 = 1, 2, ..., n0, and n0 represents the total number of hotspot information items in the video to be generated, where n0 ≤ n;

[0105] AI-assisted video generation steps:

[0106] For trending information i(R) that has passed compliance processing k0 To generate video;

[0107] Based on hotspot information i(R) k0 Extract the core events, characters, and scene elements from the content, and generate a video script;

[0108] Simultaneously, the number of shots in the video script is marked as JT, and the description of each shot in the video script is marked as MS. x x = 1, 2, ..., JT;

[0109] For existing footage in the pre-built media library, based on the description of the video script shots, MS x The matching degree between each piece of footage and the corresponding description of the video script shot is calculated as follows:

[0110] Extract the features corresponding to the video script shots, and combine them into a feature set labeled MT={mt} x1 mt x2 ...mt xc}; and for each feature mt xu The assigned weight value is preset and marked as γ. xu Where u = 1, 2, ..., c, and c is the number of features described corresponding to the video script shot; in this embodiment, ;

[0111] Simultaneously, extract the features of the relevant materials, and combine them into a feature set, which is then labeled as ST={st}. y1 st y2 ...st yz};

[0112] Where y = 1, 2, ..., SC, SC is the number of materials in the material library, and z is the number of features of the corresponding material;

[0113] For each video script shot, the corresponding feature mt is described xu :

[0114] If the corresponding material contains matching features, i.e., mt xu =st yv Then add a matching identifier PP to it. xu And set its value to 1;

[0115] If the corresponding material does not contain a matching feature, i.e., mt xu ≠st yv Then add a matching identifier PP to it. xu and set its value to 0;

[0116] Where v = 1, 2, ..., z;

[0117] Subsequently passed: Calculate the matching degree (PD) between the corresponding material and the description of the video script shot. x ;

[0118] In this embodiment, the weight value γ xu The feature mt corresponding to the video script shot description xu Importance weight, PP xu This indicates whether the feature matches in the data, and the weighted sum reflects the difference in the contribution of different features to the matching degree.

[0119] Subsequently, PD was selected. x ≥PDY materials;

[0120] Wherein, PDY is the matching threshold between the source material and the corresponding description of the video script shot;

[0121] When the material library does not contain a PD-compliant item x When the material is ≥PDY, the material required for the video script shot is generated using a pre-built generative artificial intelligence model based on the corresponding description of the video script shot;

[0122] The selected and generated footage will then be composited according to the shot order in the script to form the initial video V0.

[0123] This embodiment utilizes web crawling technology to extract information from news and social media platforms. It combines dissemination frequency and timeliness indices to calculate popularity scores and filter trending information, accurately capturing currently popular content. In the video generation stage, core elements of the trending information are extracted to generate scripts. Material is acquired and synthesized using feature matching and generative artificial intelligence models, achieving automated and intelligent generation of trending videos. This ensures the trending nature of the video content while leveraging artificial intelligence to improve material matching efficiency and video generation convenience, providing an effective method for rapidly producing trending-related videos.

[0124] As a second embodiment of the present invention:

[0125] In its specific implementation, compared to Embodiment 1, the technical solution of this embodiment differs only in that it further includes the step of: hot topic information compliance processing. This step is used after the hot topic information monitoring and screening step and before the artificial intelligence-assisted video generation step, as detailed below:

[0126] The system includes identifying misinformation, assessing values ​​and orientations, and detecting portrait rights infringements related to trending topics; specifically as follows:

[0127] Step A1, Identification of False Information:

[0128] Establish a real information database T = {t1, t2, ..., t} m};

[0129] Through:

[0130] Calculate i(R) for each hotspot information. k0 Real information t in the real information database j Similarity S k0,j ;

[0131] Where j = 1, 2, ..., m, m represents the number of real information entries in the real information database. Let L(R) represent the number of elements in the intersection of sets L(R) and T. This represents the number of elements in the union of sets L(R) and T;

[0132] For each i(R) k0 :

[0133] If there exists a true piece of information t j Make S k0,j If the value is greater than or equal to SY, then the hotspot information i(R) is determined to be valid. k0 This is true information;

[0134] If all true information t j The corresponding S k0,j If <SY, then mark the hotspot information i(R). k0 This is suspected to be false information;

[0135] Where SY is the preset similarity threshold;

[0136] Step A2, Values-Oriented Assessment:

[0137] Construct a value keyword library V = {v1, v2, ... v} p The value keyword library includes positive value keywords, such as "honesty" and "friendliness," and negative value keywords, such as "violence" and "discrimination."

[0138] For each hot topic information i(R) k0 Extract its keyword set W = {w1, w2, ... w} q}, where q represents hotspot information i(R) k0 The number of keywords in the text;

[0139] Subsequently passed:

[0140] Calculate the positive value matching degree P V Matching degree N with negative values V ;

[0141] Among them, Va is a subset of keywords representing positive values, and Vb is a subset of keywords representing negative values;

[0142] P V Used to measure the proportion of positive values ​​in video content; N V Used to measure the proportion of negative values ​​in video content; len(W) represents the total number of words in the word set W; Va is a subset of positive value keywords in the value keyword library; Vb is a subset of negative value keywords; the "1" in the formula is a counting marker, indicating that each time a word belonging to Va or Vb is matched in W, it is counted as 1.

[0143] In this embodiment:

[0144] This represents the total number of words in the word set W of the video content that belong to the subset Va of positive value keywords, i.e., 1 is added for each positive word matched. This represents the percentage of positive value keywords in the total vocabulary of the video; the higher the percentage, the stronger the positive value orientation of the content.

[0145] This represents the total number of words in the word set W of the video content that belong to the negative value keyword subset Vb, i.e., 1 is added for each negative word matched. This represents the percentage of negative value keywords in the total vocabulary of the video; the higher the percentage, the stronger the negative value orientation of the content, and the greater the possibility of value risks.

[0146] If N V >0.1 or P V If the value is less than 0.3, it is determined that there is a risk in the value orientation.

[0147] Step A3, Portrait Rights Infringement Detection:

[0148] Establish a facial feature database F = {f1, f2, ..., f} r}, where f eThese are the feature vectors of authorized portraits, e = 1, 2, ..., r;

[0149] Extract feature vector f from the face region in the generated video;

[0150] pass: Calculate the f values ​​that match all facial feature databases. e Similarity Sim(f,f) e );

[0151] Where d is the dimension of the feature vector, f z and f e,z It is the z-th component of the eigenvector;

[0152] If Sim(f, f) exists e If the value is greater than or equal to 0.7 and no authorization is obtained, there is a risk of infringement, so the face will be blurred or replaced.

[0153] This embodiment adds a compliance processing step for trending information based on Embodiment 1. Through false information identification, value orientation assessment, and portrait rights infringement detection, it effectively avoids risks such as the spread of false information, value deviations, and portrait rights infringement. False information identification enhances the authenticity of information by comparing similarity data from a database of real information; value orientation assessment quantifies the proportion of positive and negative values ​​using a keyword library to ensure that video content conforms to positive value orientations; and portrait rights infringement detection protects the portrait rights of others, making the generated trending videos not only trending but also more compliant and socially responsible, thus improving the credibility and standardization of the video content.

[0154] As an embodiment of the present invention:

[0155] In its specific implementation, compared to Embodiment 1 and Embodiment 2, the technical solution of this embodiment combines the solutions of Embodiment 1 and Embodiment 2. The difference between the technical solution of this embodiment and Embodiment 1 and Embodiment 2 lies only in that this embodiment also includes the step: initial video compliance verification.

[0156] The initial video undergoes compliance verification, with verification dimensions consistent with the compliance processing steps; the specific method is as follows:

[0157] The information in V0 is detected according to the false information identification steps. If there is suspected false information, a correction prompt is generated and fed back to the generation model to obtain the corrected video V1.

[0158] The revised video V1 was evaluated according to the value orientation assessment steps. If it was determined that there was a risk in the value orientation, the negative words in the revised video were replaced with positive words, and the visual style of the revised video was adjusted to a positive guidance style, resulting in the revised video V2.

[0159] Face detection was performed on the adjusted video V2 according to the portrait rights infringement detection steps. If there was a risk of infringement, the face was blurred or replaced to obtain the final compliant video V3.

[0160] This embodiment, combining embodiments one and two, adds an initial video compliance verification step. After video generation, it verifies and corrects the video again from the dimensions of false information, value orientation, and portrait rights infringement, forming a complete process from hot topic information screening and video generation to video compliance verification. This process further strengthens video compliance, correcting potential oversights in the initial video, such as generating correction prompts for suspected false information, replacing negative words and adjusting the visual style, and handling infringing faces. This ensures that the final output video is not only hot topic content but also a fully compliant, high-quality video with correct guidance, comprehensively improving video quality and compliance assurance.

[0161] As an embodiment of the present invention:

[0162] In specific implementation, compared with Embodiment 1, Embodiment 2 and Embodiment 3, the technical solution of this embodiment is to combine the solutions of Embodiment 1, Embodiment 2 and Embodiment 3.

[0163] Example 4 integrates all the solutions from the first three examples, achieving full-process coverage of hotspot monitoring, video generation, and compliance processing of both information and video. It possesses the advantages of Example 1 in rapidly capturing hotspots and intelligently generating videos, and through the compliance processing of hotspot information in Example 2 and the video compliance verification in Example 3, it controls compliance risks throughout the entire process from information source to finished video, forming a complete, closed-loop system for hotspot video generation and compliance assurance. This system can generate hotspot videos efficiently, accurately, and compliantly, ensuring timeliness and intelligence while maximizing the authenticity, positive value, and compliance of content, providing the industry with a comprehensive and reliable solution for hotspot video production.

[0164] This invention also provides a video generation system based on artificial intelligence and hotspot monitoring. This system is used to execute a video generation method based on artificial intelligence and hotspot monitoring. The system includes:

[0165] The hotspot monitoring module includes an information capture unit and a popularity calculation unit;

[0166] The information capture unit is used to capture raw information from the network and form a set of raw information;

[0167] The popularity calculation unit is used to calculate the popularity value of each piece of original information and filter out the set of hot information;

[0168] The compliance processing module includes a false information detection unit, a value orientation detection unit, and a portrait rights detection unit;

[0169] The fake information detection unit is used to identify fake information in trending information.

[0170] The value orientation detection unit is used to assess the value orientation of trending information.

[0171] The portrait rights detection unit is used to detect portrait rights infringements in trending information.

[0172] The video generation module includes a script generation unit, a material processing unit, and a video compositing unit;

[0173] The script generation unit is used to generate video scripts based on hotspot information;

[0174] The material processing unit is used to select existing materials from a pre-built material library based on the matching degree between each material and the corresponding description of the video script shots, and at the same time, it generates materials that are not in the material library but are required by the video script through a generative artificial intelligence model;

[0175] The video compositing unit is used to combine the selected and generated footage according to the shot order in the script to form the initial video;

[0176] The compliance verification module is used to perform compliance verification on the initial video in conjunction with the compliance processing module, and its verification method is consistent with the methods used for identifying false information, assessing value orientation, and detecting portrait rights infringements in hot information.

[0177] It should be stated that all user data collected in this application was collected with the user's consent and authorization, and the use of user data is legal and compliant, and the use and processing of user data comply with the relevant laws, regulations and standards of the relevant regions.

[0178] Furthermore, any content not described in detail in this specification is existing technology known to those skilled in the art.

[0179] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, 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 process, method, article, or apparatus.

[0180] The above formulas are all dimensionless calculations. Dimensionless calculation involves introducing a reference benchmark, such as the maximum, minimum, standard deviation, or theoretical extreme value of a physical quantity, to transform the original physical quantity into a dimensionless relative value. This value is usually mapped to a specific interval, such as [0,1] or [-1,1], which eliminates the influence of units while preserving the relative size relationship of the physical quantities. The formula is derived from software simulation based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0181] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

[0182] The various embodiments of this application have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical application, or improvement of the technology in the market, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. A video generation method based on artificial intelligence and hotspot monitoring, characterized in that, Includes the following steps: Step 1: Monitoring and filtering of trending information: Raw information is crawled from the network and formed into a raw information set. By determining the popularity value of each piece of raw information, a set of hot information is filtered out. The methods for filtering hot topic information collections are as follows: The original set of information captured is labeled as L={i1, i2, ..., i}. n }, where i k This represents the kth original message, where k = 1, 2, ..., n, and n represents the number of original messages retrieved. Simultaneously, the dissemination frequency and timeliness index of each original message are extracted. Then, the dissemination frequency of the k-th original message is multiplied by the corresponding preset weight coefficient, and this is added to the timeliness index of that message multiplied by the corresponding preset weight coefficient. The sum of these two values ​​yields the popularity value of each original message, denoted as RD. k ; Then, based on the preset heat threshold RDY, select those that meet the RD... k The set of hotspot information ≥ RDY, i.e. the set of information to be generated for the video, is denoted as L(R) = {i(R)1, i(R)2, ..., i(R)}. n0 }; where i(R) k0 This represents the k0th hotspot information, where k0 = 1, 2, ..., n0, and n0 represents the total number of hotspot information items in the video to be generated, where n0 ≤ n; Step 2: Compliance processing of trending information: Conducting functions such as identifying misinformation, assessing value orientations, and detecting portrait rights infringements in trending information; The methods for identifying false information are as follows: Extract from a pre-established database of real information T = {t1, t2, ..., t} m }; The number of elements in the intersection of the hot information set and the real information database is divided by the number of elements in the union of the hot information set and the real information database to obtain each hot information i(R). k0 Real information t in the real information database j The similarity is denoted as S. k0,j ; Where j = 1, 2, ..., m, m represents the number of real information entries in the real information database; For each i(R) k0 If there exists a true piece of information t j Make S k0,j If the value is greater than or equal to SY, then the hotspot information i(R) is determined to be valid. k0 For true information; if all true information t j The corresponding S k0,j If <SY, then mark the hotspot information i(R). k0 This is suspected to be false information; where SY is a preset similarity threshold. The value-oriented assessment method is as follows: Construct a value keyword library V = {v1, v2, ... v} p The value keyword library includes both positive and negative value keywords. For each hot topic information i(R) k0 Extract its keyword set W = {w1, w2, ... w} q }, where q represents hotspot information i(R) k0 The number of keywords in the text; The positive value relevance score is calculated by dividing the total number of words belonging to the positive value keyword subset within the keyword set of trending information by the total number of words in that keyword set. This positive value relevance score is denoted as P. V ; Simultaneously, the total number of words belonging to the negative value keyword subset within the keyword set of trending information is divided by the total number of words in that keyword set to obtain the negative value matching degree, denoted as N. V ; If N V >0.1 or P V If the value is less than 0.3, it is determined that there is a risk in the value orientation. The methods for detecting portrait rights infringement are as follows: Establish a facial feature database F = {f1, f2, ..., f} r }, where f e These are the feature vectors of authorized portraits, e = 1, 2, ..., r; Extract feature vector f from the face region in the generated video; pass: Calculate the f values ​​in the facial feature database. e Similarity Sim(f,f) e ); Where d is the dimension of the feature vector, z = 1, 2, ..., d, f z It is the z-th component of the feature vector f extracted from the face region in the generated video, f e,z It is the feature vector f of the authorized portrait in the face feature database. e The z-th component; If Sim(f, f) exists e If the value is ≥0.7 and no authorization is obtained, there is a risk of infringement, so the face will be blurred or replaced. Step 3: Artificial intelligence-assisted video generation: Based on trending information, a video script is generated. By matching the descriptions of each material in the pre-built material library with the corresponding shots in the video script, existing materials are selected. At the same time, a generative artificial intelligence model is used to generate materials that are not in the material library but are required by the video script. Then, the selected and generated materials are combined according to the shot order of the script to form the initial video. Step 4: Initial Video Compliance Verification The initial video undergoes compliance verification, and the specific verification method is as follows: The information in V0 is detected according to the false information identification steps. If there is suspected false information, a correction prompt is generated and fed back to the generation model to obtain the corrected video V1. The revised video V1 was evaluated according to the value orientation assessment steps. If it was determined that there was a risk in the value orientation, the negative words in the revised video were replaced with positive words, and the visual style of the revised video was adjusted to a positive guidance style, resulting in the revised video V2. Face detection was performed on the adjusted video V2 according to the portrait rights infringement detection steps. If there was a risk of infringement, the face was blurred or replaced to obtain the final compliant video V3.

2. The video generation method based on artificial intelligence and hotspot monitoring according to claim 1, characterized in that: in, The number of disseminations can be selected from either the number of reposts or the number of clicks, or the sum of the number of reposts and the number of clicks. The timeliness index is the interval between the information release time and the current time.

3. The video generation method based on artificial intelligence and hotspot monitoring according to claim 1, characterized in that: The initial video is generated in the following way: Based on hotspot information i(R) k0 Extract the core events, characters, and scene elements from the content, and generate a video script; Simultaneously, the number of shots in the video script is marked as JT, and the description of each shot in the video script is marked as MS. x x = 1, 2, ..., JT; For existing footage in the pre-built media library, based on the description of the video script shots, MS x Calculate the matching degree between each source material and the corresponding description of the video script shot, and record it as PD. x ; Subsequently, PD was selected. x Footage with a value of ≥PDY; where PDY is the matching threshold between the footage and the corresponding description of the video script shot; When the material library does not contain a PD-compliant item x When the material is ≥PDY, the material required for the video script shot is generated using a pre-built generative artificial intelligence model based on the corresponding description of the video script shot; The selected and generated footage will then be composited according to the shot order in the script to form the initial video V0.

4. The video generation method based on artificial intelligence and hotspot monitoring according to claim 3, characterized in that: The matching degree between the source material and the corresponding description of the video script shot is calculated as follows: First, assign a preset weight value to each feature described in the video script shot. At the same time, extract the features of the corresponding material. Then, determine whether the feature described in the video script shot exists in the corresponding material. If it exists, add a matching identifier to the feature and set its value to 1. Conversely, if the matching identifier is set to 0, then the weight value of each feature is multiplied by the corresponding judgment result, and finally all the multiplication results are added together. The sum is the matching degree (PD) between the corresponding material and the video script shot description. x .

5. A video generation system based on artificial intelligence and hotspot monitoring, the system being used to execute the video generation method based on artificial intelligence and hotspot monitoring as described in any one of claims 1-4, characterized in that, The system includes: The hotspot monitoring module is used to capture raw information from the network and form a raw information set. It then filters out the hot information set by determining the popularity value of each piece of raw information. The compliance processing module is used to identify false information, assess values, and detect infringements of portrait rights in trending information. The video generation module is used to generate video scripts based on hot topic information. It selects existing materials by matching the descriptions of the corresponding shots in the video script with the materials in the pre-built material library. At the same time, it generates materials that are not in the material library but are required by the video script through a generative artificial intelligence model. Then, the selected and generated materials are combined according to the shot order of the script to form the initial video. The compliance verification module is used to perform compliance verification on the initial video in conjunction with the compliance processing module, and its verification method is consistent with the methods used for identifying false information, assessing value orientation, and detecting portrait rights infringements in hot information.

Images