Short video advertisement generation method and system

Abstract

The invention relates to the technical field of video advertisement generation, and particularly discloses a short video advertisement generation method and system, and the method comprises the steps: receiving an advertisement text, and carrying out the theme extraction and keyword extraction of the advertisement text; screening the media videos in a preset time range according to the extracted theme to obtain a case video containing exposure, and selecting a material in a preset material library according to the extracted keyword; calculating the individual exposure of the selected materials based on the case video, and combining the materials based on the individual exposure to generate a draft video; according to the method, the advertisement text is received, text recognition is carried out on the advertisement text, the keyword is extracted, the material is obtained through the keyword, combined evaluation is carried out based on the current popular video, some draft videos are rapidly generated, an editor only needs to carry out simple adjustment on the draft videos, even only needs to review the draft, and a video finished product can be obtained; and the working efficiency is greatly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of video advertisement generation, and in particular to a short video advertisement generation method and system. Background Art

[0002] Short video ads are a form of advertising that delivers brand information, product features or promotional content in a short period of time, usually between 5 and 30 seconds in length. With the rapid development of short video platforms, short video ads have become an important means of brand promotion due to their high frequency, short duration and high interactivity.

[0003] Existing short video advertisements are all video data manually generated by editors. When facing some relatively conventional needs, editors need to perform a lot of repetitive work, which affects their work efficiency. How to provide a simple short video generation solution to alleviate the work pressure of editors is the technical problem that the technical solution of the present invention aims to solve. Summary of the invention

[0004] The purpose of the present invention is to provide a short video advertisement generation method and system to solve the problems raised in the above-mentioned background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A method for generating a short video advertisement, the method comprising:

[0007] Receiving an advertisement text, and extracting topics and keywords from the advertisement text;

[0008] According to the extracted topics, media videos within a preset time range are screened to obtain case videos with exposure, and materials are selected from the preset material library according to the extracted keywords;

[0009] Calculate the individual exposure of the selected materials based on the case video, combine the materials based on the individual exposure, and generate a draft video;

[0010] Sending the draft video to the verification end, receiving evaluation information fed back by the verification end, and screening the draft video;

[0011] Among them, the material selection process is connected with the AI ​​module, and the application conditions of the AI ​​module are: comparing the individual exposure of the selected material with the preset exposure threshold, recording the number of materials whose individual exposure is greater than the preset exposure threshold, and when the number of materials is less than the preset number threshold, applying the AI ​​module to supplement the material.

[0012] As a further solution of the present invention: the step of receiving the advertisement text and extracting themes and keywords from the advertisement text comprises:

[0013] Receive the advertisement text input by the user, display the preset classification label, receive the selection instruction input by the user, and determine the text label of the advertisement text according to the selection result;

[0014] Querying a text set corresponding to the advertisement text based on the text tag; wherein the generation process of the text set is: clustering the advertisement texts within a preset time period based on the text tag to obtain a text set with the text tag as an index;

[0015] Read the LDA model calculation results of the text set to determine the topic distribution of the advertisement text and the word distribution of each topic; wherein, for each text set, the number of text changes in the text set is recorded in real time, and when the number of text changes reaches a preset threshold, an LDA model calculation process is executed;

[0016] Topics and keywords are extracted based on the topic distribution and word distribution of each topic.

[0017] As a further solution of the present invention: the step of screening the media videos within a preset time range according to the extracted theme to obtain case videos with exposure, and selecting materials from a preset material library according to the extracted keywords includes:

[0018] Establish a connection channel with the media video library to intercept media videos within a preset time range;

[0019] Calculate exposure according to browsing parameters of media videos, and select media videos whose exposure reaches a preset threshold as case videos;

[0020] Establish a connection channel with the material library and select materials from the preset material library based on the extracted keywords.

[0021] As a further solution of the present invention: the step of calculating the individual exposure of the selected materials based on the case video, combining the materials based on the individual exposure, and generating a draft video includes:

[0022] Split the case video into image sequences, and count all image sequences as image groups;

[0023] For any material, the material traverses the image group, and during the traversal process, calculates the individual exposure;

[0024] Determine the selection probability of each material based on individual exposure;

[0025] Determine the order of material selection according to the order of keywords in the ad text, determine the material corresponding to each keyword based on the selection probability and combine them in order to obtain a draft video;

[0026] The selection probability is proportional to the individual exposure.

[0027] As a further solution of the present invention: for any material, the step of traversing the image group by the material, during the traversal process, calculating the individual exposure includes:

[0028] For any material, calculate its similarity with each image in the image group;

[0029] Calculate the individual exposure of the material based on the calculated similarity and the exposure of the case video corresponding to the image;

[0030] The calculation process of individual exposure is:

[0031] ; In the formula, is the exposure of the material, For the The exposure of the case video corresponding to the image, For material and The similarity of images, is a preset integer not less than 1, is the preset coefficient, Indicates the number of images in the image group.

[0032] As a further solution of the present invention: the step of sending the draft video to the verification end, receiving the evaluation information fed back by the verification end, and screening the draft video includes:

[0033] Send the draft video to the verification end;

[0034] Receive evaluation information from the verification end and screen the draft videos;

[0035] The verification terminal includes an AI module for generating an advertisement text based on the draft video; the advertisement text serves as one type of evaluation information.

[0036] The technical solution of the present invention also provides a short video advertisement generation system, the system comprising:

[0037] A content extraction module, used for receiving an advertisement text, and performing subject extraction and keyword extraction on the advertisement text;

[0038] The data screening module is used to screen the media videos within a preset time range according to the extracted topics to obtain case videos with exposure, and select materials from the preset material library according to the extracted keywords;

[0039] A material combination module, used to calculate the individual exposure of the selected materials based on the case video, combine the materials based on the individual exposure, and generate a draft video;

[0040] A video screening module, used to send the draft video to the verification end, receive evaluation information fed back by the verification end, and screen the draft video;

[0041] Among them, the material selection process is connected with the AI ​​module, and the application conditions of the AI ​​module are: comparing the individual exposure of the selected material with the preset exposure threshold, recording the number of materials whose individual exposure is greater than the preset exposure threshold, and when the number of materials is less than the preset number threshold, applying the AI ​​module to supplement the material.

[0042] As a further solution of the present invention: the content extraction module includes:

[0043] A text label determination unit, used to receive the advertisement text input by the user, display the preset classification label, receive the selection instruction input by the user, and determine the text label of the advertisement text according to the selection result;

[0044] A text set query unit is used to query the text set corresponding to the advertisement text based on the text tag; wherein the text set is generated by clustering the advertisement texts within a preset time period based on the text tag to obtain a text set with the text tag as the index;

[0045] A distribution reading unit is used to read the LDA model calculation results of the text set, determine the topic distribution of the advertisement text and the word distribution of each topic; wherein, for each text set, the number of text changes in the text set is recorded in real time, and when the number of text changes reaches a preset number threshold, an LDA model calculation process is executed;

[0046] An execution unit is used to extract topics and keywords based on the topic distribution and the word distribution of each topic.

[0047] As a further solution of the present invention: the data screening module includes:

[0048] A data interception unit, used to establish a connection channel with a media video library and intercept media videos within a preset time range;

[0049] An exposure comparison unit, used to calculate the exposure according to the browsing parameters of the media video, and select the media video whose exposure reaches a preset threshold as the case video;

[0050] The material selection unit is used to establish a connection channel with the material library and select materials from the preset material library according to the extracted keywords.

[0051] As a further solution of the present invention: the material combination module includes:

[0052] A video splitting unit is used to split the case video into image sequences and count all image sequences as image groups;

[0053] A traversal unit, for traversing the image group from any material, and calculating individual exposures during the traversal process;

[0054] A probability calculation unit, used to determine the selection probability of each material based on individual exposure;

[0055] A sequential combination unit, used to determine the order of material selection according to the order of keywords in the advertisement text, determine the material corresponding to each keyword based on the selection probability and combine them in order to obtain a draft video;

[0056] The selection probability is proportional to the individual exposure.

[0057] Compared with the prior art, the beneficial effects of the present invention are: the present invention receives advertising text, performs text recognition on the advertising text, extracts keywords, obtains materials through keywords and performs combination evaluation based on current popular videos, and quickly generates some draft videos. The editor only needs to make simple adjustments to the draft video, or even only needs to review it, to get a finished video, which greatly improves work efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention.

[0059] Figure 1 A flowchart of a method for generating short video ads.

[0060] Figure 2 This is a structural block diagram of the short video advertisement generation system. DETAILED DESCRIPTION

[0061] In order to make the technical problems, technical solutions and beneficial effects to be solved by the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0062] Figure 1 The flowchart of the method for generating a short video advertisement is shown in the following figure. In an embodiment of the present invention, a method for generating a short video advertisement includes:

[0063] Step S100: receiving an advertisement text, and extracting topics and keywords from the advertisement text;

[0064] The advertisement text is the initial data that the user wants to convert into a video. After the user uploads the advertisement text, the execution platform of this method extracts the theme and keyword of the advertisement text to obtain the theme and keyword of the advertisement text. It should be noted that the obtained theme and keyword are actually words. The difference is that they belong to different word spaces. The theme is generally a word indicating a range, such as sports, entertainment and photography, which is a range label, and the keyword is a more specific word, such as a certain sports star.

[0065] Step S200: Screening media videos within a preset time range to obtain case videos with exposure based on the extracted theme, and selecting materials from a preset material library based on the extracted keywords;

[0066] After extracting the theme, some videos with high exposure are selected according to the theme, which are called case videos. In order to ensure the timeliness of the advertisement, the selection process occurs in the media videos within a preset time range. The media video is the video file in the software, and the time range indicates how long the video is obtained. In each media software, exposure is an inherent parameter of a media video. The media videos with high enough exposure are selected, which are called case videos.

[0067] At the same time, materials are selected from a preset material library according to the extracted keywords. The material library is an image library pre-constructed by the execution subject of the method.

[0068] Step S300: calculating the individual exposure of the selected materials based on the case video, combining the materials based on the individual exposure, and generating a draft video;

[0069] For the case videos and materials obtained, the case videos are regarded as excellent videos. The quality of the materials can be judged based on the case videos, which is represented by the parameter of individual exposure. The higher the individual exposure, the better it is considered. For advertising, this is an obvious logic. The materials are combined based on the individual exposure of each material, and the resulting material set is a video file, called a draft video.

[0070] Step S400: sending the draft video to the verification end, receiving evaluation information fed back by the verification end, and screening the draft video;

[0071] The generated draft video may not be unique, so there is also a selection process, which is to send the draft video to the verification end, receive the evaluation information (score) fed back by the verification end, and select a draft video; wherein, the verification end can be a manual end or some smart device with a built-in evaluation algorithm. The evaluation process is not something that needs to be considered in this application, so it will not be repeated.

[0072] In an example of the technical solution of the present invention, the material acquisition process is defined as follows:

[0073] The material selection process is connected to the AI ​​module. The application conditions of the AI ​​module are as follows: the individual exposure of the selected material is compared with the preset exposure threshold, and the number of materials whose individual exposure is greater than the preset exposure threshold is recorded. When the number of materials is less than the preset number threshold, the AI ​​module is used to supplement the material.

[0074] In the existing technology, the AI ​​module is already very mature. It can feedback an image by giving a text description. When the material library is insufficient, a more suitable material can be generated through the AI ​​module. Among them, there are two meanings of insufficient material in the material library. One is that there are really no photos corresponding to the keywords in the material library. The possibility of this situation is extremely low. The other is that the selected materials do not meet the conditions, that is, the individual exposure of most materials is not enough (in other words, the number of materials with sufficient individual exposure is small), which is more common.

[0075] Regarding step S100, the steps of receiving the advertisement text and extracting themes and keywords from the advertisement text include:

[0076] Receive the advertisement text input by the user, display the preset classification label, receive the selection instruction input by the user, and determine the text label of the advertisement text according to the selection result;

[0077] Querying a text set corresponding to the advertisement text based on the text tag; wherein the generation process of the text set is: clustering the advertisement texts within a preset time period based on the text tag to obtain a text set with the text tag as an index;

[0078] Read the LDA model calculation results of the text set to determine the topic distribution of the advertisement text and the word distribution of each topic; wherein, for each text set, the number of text changes in the text set is recorded in real time, and when the number of text changes reaches a preset threshold, an LDA model calculation process is executed;

[0079] Topics and keywords are extracted based on the topic distribution and word distribution of each topic.

[0080] The above content limits the topic extraction process and the keyword extraction process. The solution provided in this application is a content extraction process based on the LDA (Latent Dirichlet Allocation) model. LDA decomposes the text set into a set of topics and assigns a probability distribution of a topic to each document. After processing the text set using the LDA model, the main results obtained are the topic distribution of each document and the word distribution of each topic.

[0081] It should be noted that the LDA model processes text sets. Therefore, a text set must be constructed first. After receiving the advertising text entered by the user, the preset classification labels are immediately displayed. The user selects a label, and the selected label is used as the text label of the advertising text. The advertising text is clustered according to the text label to obtain a text set. When using it, the corresponding text set can also be queried according to the text label. Then, the LDA model can be used for topic extraction and keyword extraction.

[0082] It is worth mentioning that when the user selects the classification label, the theme of the advertising text has actually been determined. On this basis, this application introduces the calculation process of the LDA model, which means that on the basis of the large topic classification, further topic segmentation is carried out according to the text content, which greatly reduces the granularity of the topic, and the topic extraction process and the word extraction process are highly refined.

[0083] Regarding S200, the step of screening the media videos within a preset time range according to the extracted theme to obtain case videos with exposure, and selecting materials from a preset material library according to the extracted keywords includes:

[0084] Establish a connection channel with the media video library to intercept media videos within a preset time range;

[0085] Calculate exposure according to browsing parameters of media videos, and select media videos whose exposure reaches a preset threshold as case videos;

[0086] Establish a connection channel with the material library and select materials from the preset material library based on the extracted keywords.

[0087] The above content limits the data acquisition process. The data acquisition process is the preparatory work, including acquiring media videos and acquiring materials; establishing a connection channel with the media video library, intercepting media videos within a preset time range, which is generally a few days or a week, and calculating the exposure based on the browsing parameters of the media video. The media video whose exposure reaches the preset threshold is selected as the case video; then, the material can be obtained by searching in the preset material library based on the extracted keywords.

[0088] Among them, browsing parameters include views, likes, favorites and reposts, etc. Multiplying these data by preset weights and then accumulating them can get the exposure. In actual situations, exposure is often only related to the number of views. Simply multiplying the number of views by a coefficient can get the exposure.

[0089] Regarding step S300, the steps of calculating the individual exposure of the selected materials based on the case video, combining the materials based on the individual exposure, and generating a draft video include:

[0090] Split the case video into image sequences, and count all image sequences as image groups;

[0091] For any material, the material traverses the image group, and during the traversal process, calculates the individual exposure;

[0092] Determine the selection probability of each material based on individual exposure;

[0093] Determine the order of material selection according to the order of keywords in the ad text, determine the material corresponding to each keyword based on the selection probability and combine them in order to obtain a draft video;

[0094] The selection probability is proportional to the individual exposure.

[0095] In an example of the technical solution of the present invention, the process of combining materials is explained, and its core principle is to use case videos as excellent data to evaluate each material.

[0096] First, the case video is a collection of images, and the case video is converted into an image sequence; then, for any material, the material is compared with each image in the image sequence. After the comparison is completed, the quality of each material can be evaluated based on the case video, represented by the individual exposure. After the individual exposure is determined, the selection probability is determined based on the direct proportion of the individual exposure; finally, the materials can be combined based on the selection probability.

[0097] The process of combining materials based on selection probability is as follows:

[0098] The material acquisition process is as follows: select materials from the preset material library based on the extracted keywords. The selected materials are not unique, and there is a one-to-many relationship between keywords and materials. This also means that each material corresponds to a keyword. Keywords are words in the advertising text. The order of keywords in the advertising text is obtained as the processing order; keywords are selected as processing objects in turn according to the processing order, all materials corresponding to the keywords are queried, and one material is randomly selected from all materials based on the selection probability. They are spliced ​​in the processing order, and the resulting image set is called a draft video.

[0099] It should be noted that the process of randomly selecting materials and splicing them in the order of processing is a cyclic process, which is executed multiple times in a cycle. After the duplicate draft videos are eliminated, the loop is exited when there are enough remaining draft videos.

[0100] Specifically, for any material, the material traverses the image group, and during the traversal process, the step of calculating the individual exposure includes:

[0101] For any material, calculate its similarity with each image in the image group;

[0102] Calculate the individual exposure of the material based on the calculated similarity and the exposure of the case video corresponding to the image;

[0103] The calculation process of individual exposure is:

[0104] ; In the formula, is the exposure of the material, For the The exposure of the case video corresponding to the image, For material and The similarity of images, is a preset integer not less than 1, is the preset coefficient, Indicates the number of images in the image group.

[0105] In the above content, a specific calculation process of individual exposure is provided. The calculation principle is to multiply the similarity with the exposure corresponding to the corresponding image, and then add them up to get a total exposure value, which is then multiplied by a preset coefficient. , which can be used as individual exposure; its physical meaning is that the more similar it is to an image with a higher exposure, the higher the individual exposure.

[0106] Based on the above content, this application has compounded the similarity, specifically: Item, the similarity value is generally between zero and one. The item is actually , the greater the similarity, the smaller the denominator, The larger the term is, the higher the exposure is, and the greater the impact on the final individual exposure is. In other words, images with high exposure have greater weight.

[0107] There are many calculation schemes for the similarity between images, such as structural similarity, etc., and the existing calculation schemes can be used.

[0108] Regarding step S400, the step of sending the draft video to the verification end, receiving evaluation information fed back by the verification end, and screening the draft video includes:

[0109] Send the draft video to the verification end;

[0110] Receive evaluation information from the verification end and screen the draft videos;

[0111] The verification terminal includes an AI module for generating an advertisement text based on the draft video; the advertisement text serves as a type of evaluation information.

[0112] Since the present application involves an AI module, an AI module can actually be introduced in the verification end to evaluate the draft video. The specific evaluation method is to use the AI ​​module to identify the draft video and obtain a descriptive text. If the video is good enough, the obtained descriptive text should be similar enough to the original advertising text, which can be used as one of the evaluation information and is an expansion of the function of the verification end.

[0113] Figure 2 1 is a structural block diagram of a short video advertisement generation system. In an embodiment of the present invention, a short video advertisement generation system is provided. The system 10 includes:

[0114] The content extraction module 11 is used to receive the advertisement text and perform topic extraction and keyword extraction on the advertisement text;

[0115] The data screening module 12 is used to screen the case videos with exposure in the media videos within a preset time range according to the extracted topics, and select materials in the preset material library according to the extracted keywords;

[0116] A material combination module 13, used to calculate the individual exposure of the selected materials based on the case video, combine the materials based on the individual exposure, and generate a draft video;

[0117] The video screening module 14 is used to send the draft video to the verification end, receive the evaluation information fed back by the verification end, and screen the draft video;

[0118] Among them, the material selection process is connected with the AI ​​module, and the application conditions of the AI ​​module are: comparing the individual exposure of the selected material with the preset exposure threshold, recording the number of materials whose individual exposure is greater than the preset exposure threshold, and when the number of materials is less than the preset number threshold, applying the AI ​​module to supplement the material.

[0119] Furthermore, the content extraction module 11 includes:

[0120] A text label determination unit, used to receive the advertisement text input by the user, display the preset classification label, receive the selection instruction input by the user, and determine the text label of the advertisement text according to the selection result;

[0121] A text set query unit is used to query the text set corresponding to the advertisement text based on the text tag; wherein the text set is generated by clustering the advertisement texts within a preset time period based on the text tag to obtain a text set with the text tag as the index;

[0122] A distribution reading unit is used to read the LDA model calculation results of the text set, determine the topic distribution of the advertisement text and the word distribution of each topic; wherein, for each text set, the number of text changes in the text set is recorded in real time, and when the number of text changes reaches a preset number threshold, an LDA model calculation process is executed;

[0123] An execution unit is used to extract topics and keywords based on the topic distribution and the word distribution of each topic.

[0124] Specifically, the data screening module 12 includes:

[0125] A data interception unit, used to establish a connection channel with a media video library and intercept media videos within a preset time range;

[0126] An exposure comparison unit, used to calculate the exposure according to the browsing parameters of the media video, and select the media video whose exposure reaches a preset threshold as the case video;

[0127] The material selection unit is used to establish a connection channel with the material library and select materials from the preset material library according to the extracted keywords.

[0128] Furthermore, the material combination module 13 includes:

[0129] A video splitting unit is used to split the case video into image sequences and count all image sequences as image groups;

[0130] A traversal unit, for traversing the image group from any material, and calculating individual exposures during the traversal process;

[0131] A probability calculation unit, used to determine the selection probability of each material based on individual exposure;

[0132] A sequential combination unit, used to determine the order of material selection according to the order of keywords in the advertisement text, determine the material corresponding to each keyword based on the selection probability and combine them in order to obtain a draft video;

[0133] The selection probability is proportional to the individual exposure.

[0134] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for generating short video advertisements, characterized in that: The method comprises: Receiving an advertisement text, and extracting topics and keywords from the advertisement text; According to the extracted topics, media videos within a preset time range are screened to obtain case videos with exposure, and materials are selected from the preset material library according to the extracted keywords; Calculate the individual exposure of the selected materials based on the case video, combine the materials based on the individual exposure, and generate a draft video; Sending the draft video to the verification end, receiving evaluation information fed back by the verification end, and screening the draft video; Among them, the material selection process is connected with the AI ​​module, and the application conditions of the AI ​​module are: comparing the individual exposure of the selected material with the preset exposure threshold, recording the number of materials whose individual exposure is greater than the preset exposure threshold, and when the number of materials is less than the preset number threshold, applying the AI ​​module to supplement the material.

2. The short video advertisement generating method according to claim 1, characterized in that: The step of receiving the advertisement text and extracting themes and keywords from the advertisement text comprises: Receive the advertisement text input by the user, display the preset classification label, receive the selection instruction input by the user, and determine the text label of the advertisement text according to the selection result; Querying a text set corresponding to the advertisement text based on the text tag; wherein the generation process of the text set is: clustering the advertisement texts within a preset time period based on the text tag to obtain a text set with the text tag as an index; Read the LDA model calculation results of the text set to determine the topic distribution of the advertisement text and the word distribution of each topic; wherein, for each text set, the number of text changes in the text set is recorded in real time, and when the number of text changes reaches a preset threshold, an LDA model calculation process is executed; Topics and keywords are extracted based on the topic distribution and word distribution of each topic.

3. The short video advertisement generating method according to claim 1, characterized in that: The step of screening the media videos within a preset time range according to the extracted topics to obtain case videos with exposure, and selecting materials from a preset material library according to the extracted keywords includes: Establish a connection channel with the media video library to intercept media videos within a preset time range; Calculate exposure according to browsing parameters of media videos, and select media videos whose exposure reaches a preset threshold as case videos; Establish a connection channel with the material library and select materials from the preset material library based on the extracted keywords.

4. The short video advertisement generating method according to claim 1, characterized in that: The steps of calculating the individual exposure of the selected materials based on the case video, combining the materials based on the individual exposure, and generating a draft video include: Split the case video into image sequences, and count all image sequences as image groups; For any material, the material traverses the image group, and during the traversal process, calculates the individual exposure; Determine the selection probability of each material based on individual exposure; Determine the order of material selection according to the order of keywords in the ad text, determine the material corresponding to each keyword based on the selection probability and combine them in order to obtain a draft video; The selection probability is proportional to the individual exposure.

5. The short video advertisement generating method according to claim 4, characterized in that: For any material, the step of traversing the image group by the material, and calculating the individual exposure during the traversal process includes: For any material, calculate its similarity with each image in the image group; Calculate the individual exposure of the material based on the calculated similarity and the exposure of the case video corresponding to the image; The calculation process of individual exposure is: ; In the formula, is the exposure of the material, For the The exposure of the case video corresponding to the image, For material and The similarity of images, is a preset integer not less than 1, is the preset coefficient, Indicates the number of images in the image group.

6. The short video advertisement generating method according to claim 1, characterized in that: The step of sending the draft video to the verification end, receiving evaluation information fed back by the verification end, and screening the draft video comprises: Send the draft video to the verification end; Receive evaluation information from the verification end and screen the draft videos; The verification terminal includes an AI module for generating an advertisement text based on the draft video; the advertisement text serves as one type of evaluation information.

7. A short video advertisement generation system, characterized in that: The system comprises: A content extraction module, used for receiving an advertisement text, and performing subject extraction and keyword extraction on the advertisement text; The data screening module is used to screen the media videos within a preset time range according to the extracted topics to obtain case videos with exposure, and select materials from the preset material library according to the extracted keywords; A material combination module, used to calculate the individual exposure of the selected materials based on the case video, combine the materials based on the individual exposure, and generate a draft video; A video screening module, used to send the draft video to the verification end, receive evaluation information fed back by the verification end, and screen the draft video; Among them, the material selection process is connected with the AI ​​module, and the application conditions of the AI ​​module are: comparing the individual exposure of the selected material with the preset exposure threshold, recording the number of materials whose individual exposure is greater than the preset exposure threshold, and when the number of materials is less than the preset number threshold, applying the AI ​​module to supplement the material.

8. The short video advertisement generation system according to claim 7, characterized in that: The content extraction module comprises: A text label determination unit, used to receive the advertisement text input by the user, display the preset classification label, receive the selection instruction input by the user, and determine the text label of the advertisement text according to the selection result; A text set query unit is used to query the text set corresponding to the advertisement text based on the text tag; wherein the text set is generated by clustering the advertisement texts within a preset time period based on the text tag to obtain a text set with the text tag as the index; A distribution reading unit is used to read the LDA model calculation results of the text set, determine the topic distribution of the advertisement text and the word distribution of each topic; wherein, for each text set, the number of text changes in the text set is recorded in real time, and when the number of text changes reaches a preset number threshold, an LDA model calculation process is executed; An execution unit is used to extract topics and keywords based on the topic distribution and the word distribution of each topic.

9. The short video advertisement generating system according to claim 7, characterized in that: The data screening module comprises: A data interception unit, used to establish a connection channel with a media video library and intercept media videos within a preset time range; An exposure comparison unit, used to calculate the exposure according to the browsing parameters of the media video, and select the media video whose exposure reaches a preset threshold as the case video; The material selection unit is used to establish a connection channel with the material library and select materials from the preset material library according to the extracted keywords.

10. The short video advertisement generating system according to claim 7, characterized in that: The material combination module includes: A video splitting unit is used to split the case video into image sequences and count all image sequences as image groups; A traversal unit, for traversing the image group from any material, and calculating individual exposures during the traversal process; A probability calculation unit, used to determine the selection probability of each material based on individual exposure; A sequential combination unit, used to determine the order of material selection according to the order of keywords in the advertisement text, determine the material corresponding to each keyword based on the selection probability and combine them in order to obtain a draft video; The selection probability is proportional to the individual exposure.

Images