Message processing method and device, electronic equipment and medium

Abstract

The embodiment of the invention provides a message processing method and device, electronic equipment and a medium, and the message processing method specifically comprises the steps: determining plot summary data corresponding to a movie and television play according to description data corresponding to the movie and television play; the description data comprises diversity plot data corresponding to a television drama and / or explanation subtitle data corresponding to a movie; obtaining actor table data corresponding to the movie and television play; processing the plot summary data and actor table data corresponding to the movie and television play by using a first large language model to obtain push message content corresponding to the movie and television play; the actor name in the actor table data is used for replacing the role name in the plot summary data. According to the embodiment of the invention, the labor cost consumed by determining the content of the push message can be saved, the determination efficiency of the push message can be improved, and the richness of the content of the push message corresponding to a movie can be improved.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of communication technology, and in particular to a message processing method, device, electronic device and medium. Background Art

[0002] In the era of messaging, the server side of an application often pushes a large number of push messages to the client side. Push messages mainly refer to summary information of actual content such as news, videos, music, etc. Users can quickly browse the push messages on the client side to determine whether they are interested in the pushed messages they browse. If interested, they can directly enter the application interface by clicking on the push message to view the actual content.

[0003] In the related art, the content of the push message is usually determined by the operator. Taking the push message corresponding to the film and television drama as an example, the operator usually browses the film and television drama first, and then determines the content of the push message after understanding the film and television drama.

[0004] However, in actual applications, browsing and understanding movies and TV series takes a certain amount of time, which results in low efficiency in determining push messages in related technologies. Summary of the invention

[0005] The purpose of the embodiments of the present invention is to provide a message processing method, device, electronic device, and medium, which can save the manpower cost consumed in determining the content of push messages, improve the efficiency of determining push messages, and improve the richness of the push message content corresponding to movies.

[0006] The specific technical solutions are as follows:

[0007] In a first aspect of the implementation of the present invention, a message processing method is provided, the method comprising:

[0008] Determine the plot summary data corresponding to the TV series according to the description data corresponding to the TV series; the description data includes: the episode plot data corresponding to the TV series, and / or the commentary subtitle data corresponding to the movie;

[0009] Get the actor data corresponding to the film and television series;

[0010] The first language model is used to process the plot summary data and cast data corresponding to the film and television drama to obtain the push message content corresponding to the film and television drama; the actor names in the cast data are used to replace the character names in the plot summary data; the first language model is used to characterize the correspondence between the plot summary data and cast data corresponding to the film and television drama and the push message content corresponding to the film and television drama.

[0011] In a second aspect of the present invention, a message processing device is provided, the device comprising:

[0012] A plot summary determination module is used to determine the plot summary data corresponding to the TV series according to the description data corresponding to the TV series; the description data includes: the episode plot data corresponding to the TV series, and / or the commentary subtitle data corresponding to the movie;

[0013] Actor acquisition module, used to obtain actor data corresponding to a film or TV series;

[0014] The message content determination module is used to use the first large language model to process the plot summary data and cast data corresponding to the film and television drama to obtain the push message content corresponding to the film and television drama; the actor names in the cast data are used to replace the character names in the plot summary data; the first large language model is used to characterize the correspondence between the plot summary data and cast data corresponding to the film and television drama and the push message content corresponding to the film and television drama.

[0015] Optionally, the push message content includes: main title content and subtitle content; the message content determination module includes:

[0016] A template filling module, used for filling the plot summary data and the cast data into the prompt template to obtain a prompt text;

[0017] The model processing module is used to input the prompt text into the first large language model to obtain at least one main title content and at least one subtitle content output by the first large language model.

[0018] Optionally, the prompt template includes: task description information, first step information, second step information and precaution information;

[0019] The first step information corresponds to a generation step corresponding to the subtitle content, and the first step information includes: a subtitle case, a plot summary data item, a cast data item, a first character length range corresponding to the subtitle content, and first usage requirement information of the plot summary data item and the cast data item; the first usage requirement information is used to replace the character name in the plot summary data with the actor name in the cast data in the subtitle content;

[0020] The second step information corresponds to the generation step corresponding to the main title content, and the second step information includes: a main title case, a plot summary data item, a cast data item, a second character length range corresponding to the main title content, and second usage requirement information of the plot summary data item and the cast data item; the second usage requirement information is used to replace the character name in the plot summary data with the actor name in the cast data in the main title content;

[0021] The precaution information is used to detect the main title content and the subtitle content output by the first language model. If the detection result is that the main title content or the subtitle content contains both the actor name and the character name, the generation steps corresponding to the subtitle content and the generation steps corresponding to the main title content are executed.

[0022] Optionally, the first language model outputs a plurality of main title contents and a plurality of subtitle contents; and the message content determination module further includes:

[0023] A relevance score module, used to determine a relevance score between the main title content and the subtitle content;

[0024] The selection module is used to select a target main title content and a target subtitle content from the multiple main title contents and the multiple subtitle contents according to the relevance score as the push message content corresponding to the film and television drama.

[0025] Optionally, the training process of the first language model includes: pre-training and fine-tuning training;

[0026] The fine-tuning training includes:

[0027] Acquire a first training sample; the first training sample includes: an input sample and an output sample; the input sample includes: samples corresponding to the plot summary data and the cast data respectively; the output sample includes: a sample of the push message content;

[0028] According to the training samples, fine-tuning training is performed on the pre-trained first language model; the first language model that has completed fine-tuning training is used to characterize the correspondence between the plot summary data and cast data corresponding to the film and television drama and the push message content corresponding to the film and television drama.

[0029] Optionally, the plot summary determination module includes:

[0030] a first plot summary determination module, used to process the episode plot data corresponding to the TV series using the second largest language model to obtain the plot summary data corresponding to the TV series; the second largest language model is used to characterize the correspondence between the episode plot data corresponding to the TV series and the plot summary data corresponding to the TV series; and / or

[0031] The second plot summary determination module is used to use the third language model to process the commentary subtitle data corresponding to the movie to obtain the plot summary data corresponding to the movie; the commentary subtitle data is the subtitle text obtained by character recognition of the commentary video corresponding to the movie; the third language model is used to rewrite, polish and correct errors in the commentary subtitle data.

[0032] Optionally, the description data also includes: plot description data corresponding to the movie.

[0033] Optionally, the device further comprises:

[0034] The push module is used to send the push message content corresponding to the film and television drama to the target users corresponding to the film and television drama.

[0035] In a third aspect of the present invention, there is also provided an electronic device, comprising a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other via the communication bus;

[0036] Memory, used to store computer programs;

[0037] The processor is used to implement the aforementioned method steps when executing the program stored in the memory.

[0038] In a fourth aspect of the implementation of the present invention, a computer-readable storage medium is further provided, wherein the computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute any of the above-described methods.

[0039] The message processing method, device, electronic device, and medium provided in the embodiments of the present invention use the first language model to process the plot summary data and cast data corresponding to the film and television drama to obtain the push message content corresponding to the film and television drama. Since the embodiments of the present invention can generate the push message content based on the plot summary data and cast data corresponding to the film and television drama using the text generation capability of the first language model; therefore, the embodiments of the present invention can save the manpower cost consumed in determining the push message content and can improve the efficiency of determining the push message. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art are briefly introduced below.

[0041] Figure 1 A schematic diagram of the structure of a message processing system according to an embodiment of the present invention;

[0042] Figure 2 A flowchart of a message processing method according to an embodiment of the present invention;

[0043] Figure 3 A flowchart of a message processing method according to an embodiment of the present invention;

[0044] Figure 4A flowchart of a message processing method according to an embodiment of the present invention;

[0045] Figure 5 A schematic diagram of the structure of a message processing device according to an embodiment of the present invention;

[0046] Figure 6 The structure block diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0047] The technical solutions in the embodiments of the present invention will be described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0048] The embodiments of the present invention can be applied to the push scenario of push messages corresponding to movies and TV series, and can be used to increase the flow of application programs and the utilization rate of the application programs by users.

[0049] In the related art, the content of the push message is usually determined by the operator. Taking the push message corresponding to the film and television drama as an example, the operator usually browses the film and television drama first, and then determines the content of the push message after understanding the film and television drama. However, in actual applications, it takes a certain amount of time to browse and understand the film and television drama, which leads to low efficiency in determining the push message in the related art.

[0050] In view of the technical problem of low efficiency in determining push messages in related technologies, an embodiment of the present invention provides a message processing solution, which specifically includes: determining plot summary data corresponding to the film and television drama according to description data corresponding to the film and television drama; the above description data specifically includes: episode plot data corresponding to the above TV series, and / or commentary subtitle data corresponding to the above movie; obtaining cast data corresponding to the film and television drama; using a first large language model, processing the plot summary data and cast data corresponding to the above film and television drama to obtain push message content corresponding to the above film and television drama; the actor names in the above cast data are used to replace the character names in the above plot summary data; the above first large language model is used to characterize the correspondence between the plot summary data and cast data corresponding to the film and television drama and the push message content corresponding to the film and television drama.

[0051] The first language model, the second largest language model and the third largest language model in the embodiment of the present invention are all large language models. The large language model is a natural language processing technology based on deep neural networks, which first performs pre-training on large-scale text data to learn general language knowledge and representation; then, fine-tuning is performed on specific tasks to adapt to different downstream applications.

[0052] The big language model is an artificial intelligence model designed to understand and generate human language. The big language model can perform a wide range of tasks, such as the text generation task in the embodiment of the present invention, by training on a large amount of text data. The big language model is usually based on a deep learning architecture, such as a transformer structure. The embodiment of the present invention does not limit the specific type of the big language model. For example, the types of big language models may include: GPT (Generative Pre-Trained Transformer) model, GLM (Generative Language Model, General Language Model) and Baichuan big model, etc.

[0053] The embodiment of the present invention uses the first language model to process the plot summary data and cast data corresponding to the film and television drama to obtain the push message content corresponding to the film and television drama. Since the embodiment of the present invention can generate the push message content based on the plot summary data and cast data corresponding to the film and television drama using the text generation capability of the first language model; therefore, the embodiment of the present invention can save the manpower cost consumed in determining the push message content and can improve the efficiency of determining the push message.

[0054] Furthermore, the embodiment of the present invention uses the subtitle data corresponding to the movie to determine the plot summary data corresponding to the movie, which can improve the richness and completeness of the plot summary data corresponding to the movie. When the richness and completeness of the plot summary data corresponding to the movie are improved, the embodiment of the present invention can improve the richness of the push message content corresponding to the movie.

[0055] Furthermore, the embodiment of the present invention uses the actor names in the cast data to replace the character names in the plot summary data, so that the push message content contains the actor names of the film and television drama instead of the character names. Since the popularity of actor names is usually higher than that of character names, the embodiment of the present invention can increase the attractiveness of the push message content to users.

[0056] Reference Figure 1 , shows a schematic diagram of the structure of a message processing system according to an embodiment of the present invention, which may specifically include: a client 101 and a server 102. The client 101 and the server 102 may perform data interaction based on a wireless network or a wired network.

[0057] The server 102 can use the message processing solution of the embodiment of the present invention to determine the push message content corresponding to the film and television drama, and push the push message content corresponding to the film and television drama to the client 101 corresponding to the target user. The target user can be a user associated with the film and television drama. For example, the category corresponding to the film and television drama can be the category of film and television drama that the target user is interested in. It can be understood that the embodiment of the present invention does not limit the specific target user corresponding to the film and television drama.

[0058] In a specific implementation, the push message content may include a link. When a trigger operation is received from the target user for the link in the push message content, the link may be redirected to the video interface of the TV series corresponding to the link. Taking a TV series as an example, the link may be redirected to the first video interface corresponding to the TV series, which may be the TV series's behind-the-scenes video interface or the first episode video interface. Taking a movie as an example, the link may be redirected to the second video interface corresponding to the movie, which may be the movie's behind-the-scenes video interface or the movie video interface.

[0059] The embodiments of the present invention are described below through specific examples.

[0060] Reference Figure 2 , shows a flow chart of the steps of a message processing method according to an embodiment of the present invention, the method may specifically include the following steps:

[0061] Step 201: Determine the plot summary data corresponding to the TV series based on the description data corresponding to the TV series; the description data specifically includes: the episode plot data corresponding to the TV series, and / or the subtitle data corresponding to the movie;

[0062] Step 202: Obtain the actor data corresponding to the film and television drama;

[0063] Step 203: Use the first language model to process the plot summary data and cast data corresponding to the film and television drama to obtain the push message content corresponding to the film and television drama; the actor names in the cast data are used to replace the character names in the plot summary data; the first language model is used to characterize the correspondence between the plot summary data and cast data corresponding to the film and television drama and the push message content corresponding to the film and television drama.

[0064] Figure 2 The method embodiment shown can be executed by the server to determine the push message content corresponding to the film and television drama. It can be understood that the embodiment of the present invention is for Figure 2 The specific execution subject of the illustrated method embodiment is not limited.

[0065] In step 201 and step 202, the film or TV series may be any film or TV series for which the content of the push message needs to be determined. It is understood that the embodiment of the present invention does not limit the specific film or TV series.

[0066] In practical applications, the types of TV series include TV series and / or movies. The above description data include episode plot data corresponding to the TV series and / or subtitle data corresponding to the movie.

[0067] In a specific implementation, the episode plot data corresponding to the TV series can be obtained from video websites or video-related websites such as encyclopedia websites. The plot can refer to the plot of the TV series. The TV series usually includes multiple episodes, and the episode plot data can refer to the plot corresponding to a single episode of the TV series.

[0068] In a specific implementation, character recognition can be performed on the commentary video corresponding to the movie to obtain the commentary subtitle data corresponding to the movie. The commentary video corresponding to the movie may include multiple video frame images, each of which may include corresponding subtitles. The embodiment of the present invention can perform character recognition on the multiple video frame images included in the commentary video to obtain the subtitle text included in each video frame image; further, the subtitle texts included in the multiple video frame images can be combined to obtain the commentary subtitle data corresponding to the movie.

[0069] Character recognition technology can be a process of recognizing characters from an image using computer vision and machine learning technology. In one example, first, the character recognition technology can first use a text detection model to detect a bounding box around possible text; then, the processed bounding box is fed into the text recognition model to determine the specific characters within the bounding box. It can be understood that those skilled in the art can use the required character recognition technology according to actual application requirements, and the embodiments of the present invention do not limit the specific character recognition technology.

[0070] In actual applications, video-related websites contain movie plot data corresponding to movies, however, the completeness and richness of the movie plot data are usually not high. The embodiments of the present invention use the corresponding subtitle data of the movie to determine the corresponding plot summary data of the movie, which can improve the richness and completeness of the corresponding plot summary data of the movie.

[0071] In the embodiment of the present invention, in step 201, the process of determining the plot summary data corresponding to the film and television drama according to the description data corresponding to the film and television drama specifically includes:

[0072] Step A1: using the second largest language model, processing the episode plot data corresponding to the TV series to obtain the plot summary data corresponding to the TV series; the second largest language model is used to characterize the correspondence between the episode plot data corresponding to the TV series and the plot summary data corresponding to the TV series; and / or

[0073] Step A2: Using the third language model, the narration subtitle data corresponding to the above-mentioned movie is processed to obtain the plot summary data corresponding to the above-mentioned movie; the above-mentioned narration subtitle data is the subtitle text obtained by character recognition of the narration video corresponding to the movie; the above-mentioned third language model is used to rewrite, polish and correct errors in the above-mentioned narration subtitle data.

[0074] The second largest language model in step A1 may be a type of large language model, and the second largest language model and the first largest language model may be the same or different large language models. The second largest language model is used to characterize the correspondence between the episode plot data corresponding to the TV series and the plot summary data corresponding to the TV series. In other words, the second largest language model is used to generate the plot summary data corresponding to the TV series based on the episode plot data corresponding to the TV series, which can play a role in fusing the episode plot data corresponding to the TV series.

[0075] In one implementation of the present invention, the training process of the second largest language model may include: pre-training. The embodiment of the present invention may directly use the pre-trained second largest language model to process the episode plot data corresponding to the TV series to obtain the plot summary data corresponding to the TV series. For example, the episode plot data corresponding to the TV series may be input into the pre-trained second largest language model to obtain the plot summary data corresponding to the TV series output by the pre-trained second largest language model.

[0076] In another implementation of the present invention, the training process of the second language model may include: pre-training and fine-tuning training. The fine-tuning training is used to perform the task of generating plot summary data corresponding to the TV series, referred to as the first plot summary generation task.

[0077] The fine-tuning training process of the second language model includes:

[0078] Obtain a second training sample; the second training sample includes: a second input sample and a second output sample; the second input sample includes: an episode plot data sample; the second output sample includes: a first plot summary data sample; the first plot summary data sample is specifically a plot summary data sample corresponding to the TV series;

[0079] The second largest language model that has completed pre-training is fine-tuned according to the second training sample; the second largest language model that has completed fine-tuning is used to characterize the correspondence between the episode plot data corresponding to the TV series and the plot summary data corresponding to the TV series.

[0080] The episode plot data samples may be obtained from video websites or encyclopedia websites and other video-related websites. The first plot summary data samples corresponding to the episode plot data samples may be determined manually; or the episode plot data samples may be input into the fourth largest language model to obtain the first plot summary data samples output by the fourth largest language model. The fourth largest language model may be a large language model different from the second largest language model.

[0081] The third largest language model in step A2 may be a type of large language model, and the third largest language model and the first largest language model or the second largest language model may be the same or different large language models. The third largest language model is used to rewrite, polish and correct errors in the subtitle data. In other words, the third largest language model is used to generate plot summary data corresponding to the movie based on the subtitle data corresponding to the movie, which can play a role in rewriting, polishing and correcting errors in the subtitle data corresponding to the movie.

[0082] In an optional implementation of the present invention, the description data may further include: plot description data corresponding to the movie. The embodiment of the present invention may use the third language model to process the narration subtitle data corresponding to the movie and the plot description data corresponding to the movie to obtain the plot summary data corresponding to the movie. The plot description data corresponding to the movie may play a role in enriching the plot of the movie, so that the third language model enriches the plot of the movie based on the narration subtitle data.

[0083] In one implementation of the present invention, the training process of the third language model may include: pre-training. The embodiment of the present invention may directly use the pre-trained third language model to process the narration subtitle data corresponding to the movie to obtain the plot summary data corresponding to the movie. For example, the narration subtitle data corresponding to the movie may be input into the pre-trained third language model to obtain the plot summary data corresponding to the movie output by the pre-trained third language model.

[0084] In another implementation of the present invention, the training process of the third language model may include: pre-training and fine-tuning training. The fine-tuning training is used to perform the task of generating plot summary data corresponding to the movie, referred to as the second plot summary generation task.

[0085] The fine-tuning training process of the third language model includes:

[0086] Acquire a third training sample; the third training sample includes: a third input sample and a third output sample; the third input sample includes: a commentary subtitle data sample, or a commentary subtitle data sample and a plot description data sample corresponding to a movie; the third output sample includes: a second plot summary data sample; the second plot summary data sample is specifically a plot summary data sample corresponding to the movie;

[0087] According to the third training sample, fine-tuning training is performed on the pre-trained third language model; the fine-tuning training is used to characterize the correspondence between the commentary subtitle data corresponding to the movie and the plot summary data corresponding to the movie.

[0088] The movie commentary video can be obtained from a video-related website, and character recognition can be performed on the commentary video to obtain a commentary subtitle data sample. The second plot summary data sample corresponding to the commentary subtitle data sample can be determined manually; or the commentary subtitle data sample can be input into the fifth language model to obtain the second plot summary data sample output by the fifth language model. The fifth language model can be a large language model different from the third language model.

[0089] It should be noted that the first language model, the second largest language model, the third largest language model, the fourth largest language model and the fifth largest language model in the embodiment of the present invention are all large language models. The above five large language models may correspond to the same or different large language models.

[0090] In addition, it should be noted that when using any of the five large language models mentioned above, corresponding prompt information (Prompt) can be input into the large language model. The above prompt information can be used to guide the large language model to accurately generate the required output content. The above prompt information can provide specific instructions, background information of the question, expected answer format, etc., so that the large language model can better understand the task requirements of the text generation task, thereby improving the quality and pertinence of the output results.

[0091] For example, the prompt information of the second largest language model can guide the second largest language model to more effectively generate plot summary data based on the episode plot data corresponding to the TV series. Specifically, the prompt information of the second largest language model can indicate that the second largest language model needs to extract key plots, highlight the relationship between the main characters, maintain a specific language style, etc. It can also provide background information such as the type and theme of the TV series to help the second largest language model better understand the context of the episode plot data, so as to more accurately integrate and generate plot summaries. In addition, the prompt information of the second largest language model can also specify the output format and length requirements to obtain plot summary data that better meets the needs.

[0092] For another example, the functions of the prompt information of the third language model can include: clarifying the task direction, standardizing the output requirements, and assisting in error recognition. Among them, the task direction can include: emphasizing the wonderful plot twists, strengthening the characterization of the characters, making the language expression more concise and fluent, etc. The output requirements can specify the format, length, language style and other requirements of the plot summary data. Error recognition can list some common error types or issues that require special attention, such as grammatical errors, logical inconsistencies, information redundancy, etc., to help the third language model correct errors more efficiently and improve the accuracy and readability of the plot summary.

[0093] In step 202, the actor data corresponding to the film and television drama may be obtained from a video-related website. The actor data may include: a mapping relationship between actor names and role names.

[0094] In step 203, the text generation capability of the first language model may be used to process the plot summary data and cast data corresponding to the film and television series to obtain the push message content corresponding to the film and television series.

[0095] In one implementation of the present invention, the training process of the first language model may include: pre-training. The embodiment of the present invention may directly use the pre-trained first language model to process the plot summary data and cast data corresponding to the above-mentioned film and television drama to obtain the push message content corresponding to the film and television drama. For example, the plot summary data and cast data corresponding to the film and television drama may be input into the pre-trained first language model to obtain the push message content corresponding to the film and television drama output by the pre-trained first language model.

[0096] In another implementation of the present invention, the training process of the first language model may include: pre-training and fine-tuning training. The fine-tuning training of the first language model is used to generate push message content corresponding to the film and television drama. In other words, the fine-tuning training of the first language model is used to perform the push message content generation task.

[0097] The fine-tuning training process of the first language model includes:

[0098] Obtain a first training sample; the first training sample specifically includes: a first input sample and a first output sample; the first input sample includes: samples corresponding to the plot summary data and the cast data respectively; the first output sample includes: a sample of the push message content;

[0099] The pre-trained first language model is fine-tuned according to the first training sample; the fine-tuned first language model is used to characterize the correspondence between the plot summary data and cast data corresponding to the film and television drama and the push message content corresponding to the film and television drama.

[0100] In a specific implementation, the push message content specifically includes: main title content and subtitle content. The main title content can be a reflection of the overall content of the film or TV series. The subtitle content can serve as a supplementary explanation of the main title content. Usually, the first character length corresponding to the main title content can be less than the second character length corresponding to the subtitle content.

[0101] In one implementation of the present invention, plot summary data and cast data corresponding to a film or TV series may be input into a first language model to obtain at least one main title content and at least one subtitle content output by the first language model.

[0102] In one implementation of the present invention, step 203 utilizes the first large language model to process the plot summary data and cast data corresponding to the above-mentioned film and television drama, specifically including: filling the plot summary data and cast data into a prompt template to obtain a prompt text; inputting the prompt text into the first large language model to obtain at least one main title content and at least one subtitle content output by the first large language model.

[0103] The prompt text may be a prompt information in text form. The prompt text is a text fragment used to guide the first language model to generate a specific type of output. The prompt text can provide context and direction for the prompt text language model, and can guide the first language model to generate more accurate and targeted output content (at least one main title content and at least one subtitle content).

[0104] The embodiment of the present invention may provide a prompt template, and fill the plot summary data and cast data into the prompt template.

[0105] In a specific implementation, the above prompt template may specifically include: task description information, first step information, second step information and precaution information;

[0106] The first step information corresponds to a generation step corresponding to the subtitle content, and the first step information specifically includes: a subtitle case, a plot summary data item, a cast data item, a first character length range corresponding to the subtitle content, and first usage requirement information of the plot summary data item and the cast data item; the first usage requirement information is used to replace the character name in the plot summary data with the actor name in the cast data in the subtitle content;

[0107] The second step information corresponds to the generation step corresponding to the main title content, and the second step information specifically includes: a main title case, a plot summary data item, a cast data item, a second character length range corresponding to the main title content, and second usage requirement information of the plot summary data item and the cast data item; the second usage requirement information is used to replace the character name in the plot summary data with the actor name in the cast data in the main title content;

[0108] The precaution information is used to detect the main title content and the subtitle content output by the first language model. If the detection result is that the main title content or the subtitle content contains both the actor name and the character name, the generation steps corresponding to the subtitle content and the generation steps corresponding to the main title content are executed.

[0109] An example of a prompt template is provided here, which specifically includes: task description information, first step information, second step information and precaution information.

[0110] Among them, the task description information is specifically: Your task is to perform the following operations.

[0111] The first step information is as follows:

[0112] Step 1: Assuming you are a TV series operator, write 3 push messages for content promotion based on the <Plot Summary Data Item> of the TV series "TV Series Name A". Requirements: Choose the most exciting and attractive plot. The first character length range (the number of characters is greater than 25 Chinese characters and less than 30 Chinese characters). Imitate the format and writing style of <Subtitle Example>. Each push message has at least one preset emoticon. Use the TV series name in every push message as much as possible. Refer to the <Cast List Data Item>, and replace all character names with actor names. Character names and actor names cannot exist at the same time in each push message.

[0113] The second step information is as follows:

[0114] Step 2: Assuming you are a TV series operator, write 3 push messages for content promotion based on the <Plot Summary Data Item> of the TV series "TV Series Name A". Requirements: Select the plot corresponding to the first step. Second character length range (the number of characters is greater than 10 Chinese characters and less than 13 Chinese characters). Imitate the format and writing style of <Main Title Case>. Each push message has at least one preset emoticon. Use the TV series name in every push message as much as possible. Refer to <Cast List Data Item>, replace all character names with actor names, and character names and actor names cannot exist at the same time in each push message.

[0115] The specific precautions are as follows:

[0116] Note: After generating a push message, please check the push message again. If both the role name and the actor name exist in the push message, please repeat the first and second steps until the condition that the role name and the actor name do not exist in the push message at the same time is met.

[0117] Among them, the precaution information can avoid the situation where both the role name and the actor name exist in the push message, thereby increasing the attractiveness of the push message content to users.

[0118] Examples of main title cases include: "TV Series Name B" Actor A and Actor B's new drama is on air, etc.

[0119] Examples of subtitle cases include: Actor A, Actor B, Event A, etc.

[0120] In actual applications, the first language model can output a main title content and a subtitle content. In this case, the main title content and the subtitle content can be combined to obtain a push message content. In the push message content, the main title content is in front and the subtitle content is in the back. In addition, a main title content and a subtitle content can correspond to a typesetting format. The typesetting format can be: the main title content is in the first paragraph, the font is bold, the subtitle content is in the second paragraph, the font is normal font, etc. It can be understood that the embodiment of the present invention can determine the required typesetting format according to the actual application requirements, and the embodiment of the present invention does not limit the specific typesetting format.

[0121] The first language model may also output a plurality of main title contents and a plurality of sub-title contents; in this case, the process of processing the plot summary data and cast data corresponding to the film and television drama using the first language model in step 203 may also include: determining a correlation score between the main title content and the sub-title content; and selecting a target main title content and a target sub-title content from the plurality of main title contents and the plurality of sub-title contents as push message content corresponding to the film and television drama based on the correlation score.

[0122] The embodiment of the present invention can calculate the relevance score for any main title content and any subtitle content, and select a target main title content and a target subtitle content with the highest relevance score as the push message content corresponding to the film and television series. For example, the number of main title content is M, and the number of subtitle content is N, M and N can both be positive integers, and M and N can be the same or different. Then, M×N kinds of relevance scores can be calculated for any one of the M main title content and any one of the N subtitle content. Moreover, a target main title content and a target subtitle content with the highest relevance score can be selected from the M×N kinds of relevance scores.

[0123] Assuming that the correlation score between the i-th main title content and the j-th subtitle content is the highest, the i-th main title content and the j-th subtitle content can be combined to obtain the push message content corresponding to the film and television drama.

[0124] In practical applications, a first semantic vector corresponding to the main title content and a second semantic vector corresponding to the subtitle content may be determined; and a correlation score between the first semantic vector and the second semantic vector may be determined.

[0125] In one example, the reordering method of the semantic vector model can be used to determine the correlation score between the main title content and the subtitle content. Of course, the correlation score between two semantic vectors can also be determined by using a measurement method such as cosine similarity or Euclidean distance. It can be understood that the embodiment of the present invention does not limit the specific determination process of the correlation score.

[0126] The reordering method of the semantic vector model may include the steps of feature extraction, feature weighting and result sorting. Among them, feature extraction is used to represent the main title content as the corresponding first semantic vector, and the subtitle content as the corresponding second semantic vector; feature weighting is used to weight the first semantic vector or the second semantic vector according to the importance of the main title content or the subtitle content. Assuming that the importance of the main title content or the subtitle content is the same, the corresponding weight can be 1; result sorting is used to determine the correlation score between the main title content and the subtitle content, and sort the combination results between the main title content and the subtitle content according to the correlation score from high to low, and select the combination result with the first sorting result as the push message content corresponding to the film and television drama.

[0127] After determining the push message content corresponding to the film or TV series, the embodiment of the present invention may further send the push message content corresponding to the film or TV series to the target users corresponding to the film or TV series.

[0128] The target user may be a user associated with the film and television drama. Optionally, the category corresponding to the film and television drama may be a category of film and television drama that the target user is interested in. For example, if user A is interested in the film and television drama of the fairy-tale category, the push message content of the film and television drama corresponding to the fairy-tale category may be pushed to user A. For another example, if user B has watched the film and television drama A of the comedy category, the push message content of the film and television drama B of the comedy category may be pushed to user B. It can be understood that the embodiment of the present invention does not limit the specific target users corresponding to the film and television drama.

[0129] In a specific implementation, the push message content may include a link. When a trigger operation is received from the target user for the link in the push message content, the link may be redirected to the video interface of the TV series corresponding to the link. The video interface of the TV series corresponding to the link may be the actual content of the TV series. Taking a TV series as an example, the link may be redirected to the first video interface corresponding to the TV series. The first video interface may be the behind-the-scenes video interface or the first episode video interface of the TV series. Taking a movie as an example, the link may be redirected to the second video interface corresponding to the movie. The second video interface may be the behind-the-scenes video interface or the movie video interface.

[0130] In summary, the message processing method of the embodiment of the present invention uses the first language model to process the plot summary data and cast data corresponding to the film and television drama to obtain the push message content corresponding to the film and television drama. Since the embodiment of the present invention can generate the push message content based on the plot summary data and cast data corresponding to the film and television drama using the text generation capability of the first language model; therefore, the embodiment of the present invention can save the manpower cost consumed in determining the push message content and can improve the efficiency of determining the push message.

[0131] Furthermore, the embodiment of the present invention uses the subtitle data corresponding to the movie to determine the plot summary data corresponding to the movie, which can improve the richness and completeness of the plot summary data corresponding to the movie. When the richness and completeness of the plot summary data corresponding to the movie are improved, the embodiment of the present invention can improve the richness of the push message content corresponding to the movie.

[0132] Furthermore, the embodiment of the present invention uses the actor names in the cast data to replace the character names in the plot summary data, so that the push message content contains the actor names of the film and television drama instead of the character names. Since the popularity of actor names is usually higher than that of character names, the embodiment of the present invention can increase the attractiveness of the push message content to users.

[0133] Reference Figure 3 , shows a flow chart of the steps of a message processing method according to an embodiment of the present invention, the method can be used to determine the push message content corresponding to the TV series, and the method can specifically include the following steps:

[0134] Step 301, obtaining the episode plot data corresponding to the TV series;

[0135] Step 302: using the second largest language model, processing the episode plot data corresponding to the TV series to obtain plot summary data corresponding to the TV series; the second largest language model is used to characterize the corresponding relationship between the episode plot data corresponding to the TV series and the plot summary data corresponding to the TV series;

[0136] Step 303: Obtain the cast data corresponding to the TV series;

[0137] Step 304: input the plot summary data and cast data corresponding to the TV series into the first language model to obtain at least one main title content and at least one subtitle content output by the first language model;

[0138] Step 305: determine the correlation score between any main title content and any subtitle content, and select a target main title content and a target subtitle content with the highest correlation score as the push message content corresponding to the TV series according to the correlation score.

[0139] In practical applications, the character length of the plot summary data may be within a third character length range. Those skilled in the art may determine the third character length range according to practical application requirements, for example, the third character range may be [990, 1010] or the like.

[0140] In summary, the message processing method of the embodiment of the present invention uses the second largest language model to generate the plot summary data corresponding to the TV series, and uses the first largest language model to generate the push message content corresponding to the TV series. Since the large language model has been pre-trained corresponding to a large amount of text data, the output content of the large language model is diverse. Therefore, the embodiment of the present invention can automatically generate more abundant and diverse push message content for TV series, which can effectively save labor costs, and can use abundant and diverse push message content to bring more traffic to the application.

[0141] Reference Figure 4 , shows a flow chart of the steps of a message processing method according to an embodiment of the present invention, the method can be used to determine the push message content corresponding to a movie, and the method can specifically include the following steps:

[0142] Step 401, obtaining a commentary video corresponding to the movie;

[0143] Step 402: Perform character recognition on the above-mentioned commentary video to obtain commentary subtitle data corresponding to the movie;

[0144] Step 403: Using the third language model, the narration subtitle data corresponding to the movie is processed to obtain plot summary data corresponding to the movie; the narration subtitle data is a subtitle text obtained by character recognition of the narration video corresponding to the movie; the third language model is used to rewrite, polish and correct errors in the narration subtitle data;

[0145] Step 404, obtaining the actor data corresponding to the movie;

[0146] Step 405: input the plot summary data and cast data corresponding to the movie into the first language model to obtain at least one main title content and at least one subtitle content output by the first language model;

[0147] Step 406: Determine the correlation score between any main title content and any subtitle content, and select a target main title content and a target subtitle content with the highest correlation score as the push message content corresponding to the movie according to the correlation score.

[0148] In summary, the message processing method of the embodiment of the present invention uses the commentary subtitle data corresponding to the movie to determine the plot summary data corresponding to the movie, which can improve the richness and completeness of the plot summary data corresponding to the movie. When the richness and completeness of the plot summary data corresponding to the movie are improved, the embodiment of the present invention can improve the richness of the push message content corresponding to the movie.

[0149] Furthermore, the embodiment of the present invention uses the third language model to generate plot summary data corresponding to the movie, and uses the first language model to generate push message content corresponding to the movie. Since the large language model has been pre-trained corresponding to a large amount of text data, the output content of the large language model is diverse. Therefore, the embodiment of the present invention can automatically generate more abundant and diverse push message content for movies, which can effectively save labor costs, and can use abundant and diverse push message content to bring more traffic to the application.

[0150] It should be noted that, for the sake of simplicity, the method embodiments are described as a series of action combinations, but those skilled in the art should be aware that the embodiments of the present invention are not limited by the order of the actions described, because according to the embodiments of the present invention, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions involved are not necessarily required by the embodiments of the present invention.

[0151] Reference Figure 5 , shows a schematic diagram of the structure of a message processing device according to an embodiment of the present invention, and the message processing device may specifically include the following modules:

[0152] The plot summary determination module 501 is used to determine the plot summary data corresponding to the TV series according to the description data corresponding to the TV series; the description data specifically includes: the episode plot data corresponding to the TV series, and / or the commentary subtitle data corresponding to the movie;

[0153] A cast acquisition module 502 is used to acquire cast data corresponding to a film or TV series;

[0154] The message content determination module 503 is used to use the first language model to process the plot summary data and cast data corresponding to the above-mentioned film and television drama to obtain the push message content corresponding to the above-mentioned film and television drama; the actor names in the above-mentioned cast data are used to replace the character names in the above-mentioned plot summary data; the above-mentioned first language model is used to characterize the correspondence between the plot summary data and cast data corresponding to the film and television drama and the push message content corresponding to the film and television drama.

[0155] Optionally, the push message content specifically includes: main title content and subtitle content; and the message content determination module includes:

[0156] A template filling module is used to fill the above plot summary data and cast data into the prompt template to obtain a prompt text;

[0157] The model processing module is used to input the prompt text into the first language model to obtain at least one main title content and at least one subtitle content output by the first language model.

[0158] Optionally, the above-mentioned prompt template specifically includes: task description information, first step information, second step information and precaution information;

[0159] The first step information corresponds to the generation step corresponding to the subtitle content, and specifically includes: a subtitle case, a plot summary data item, a cast data item, a first character length range corresponding to the subtitle content, and first usage requirement information of the plot summary data item and the cast data item; the first usage requirement information is used to replace the character name in the plot summary data with the actor name in the cast data in the subtitle content;

[0160] The second step information corresponds to the generation step corresponding to the main title content, and specifically includes: a main title case, a plot summary data item, a cast data item, a second character length range corresponding to the main title content, and second usage requirement information of the plot summary data item and the cast data item; the second usage requirement information is used to replace the character name in the plot summary data with the actor name in the cast data in the main title content;

[0161] The above-mentioned precaution information is used to detect the main title content and sub-title content output by the above-mentioned first language model. If the above-mentioned detection result is that the above-mentioned main title content or the above-mentioned sub-title content contains both the actor name and the role name, then the generation steps corresponding to the above-mentioned sub-title content and the generation steps corresponding to the above-mentioned main title content are executed.

[0162] Optionally, the first language model outputs a plurality of main title contents and a plurality of subtitle contents; and the message content determination module may further include:

[0163] A relevance score module, used to determine the relevance score between the main title content and the subtitle content;

[0164] The selection module is used to select a target main title content and a target subtitle content from the above-mentioned multiple main title contents and multiple subtitle contents according to the above-mentioned relevance scores, as the push message content corresponding to the above-mentioned film and television drama.

[0165] Optionally, the training process of the first language model includes: pre-training and fine-tuning training;

[0166] The above-mentioned fine-tuning training specifically includes:

[0167] Obtaining a first training sample; the first training sample includes: an input sample and an output sample; the input sample includes: samples corresponding to the plot summary data and the cast data respectively; the output sample includes: a sample of the push message content;

[0168] According to the above training samples, the first language model that has completed pre-training is fine-tuned; the first language model that has completed fine-tuning training is used to represent the correspondence between the plot summary data and cast data corresponding to the film and television drama and the push message content corresponding to the film and television drama.

[0169] Optionally, the plot summary determination module specifically includes:

[0170] a first plot summary determination module, for processing the episode plot data corresponding to the TV series using the second largest language model to obtain the plot summary data corresponding to the TV series; the second largest language model is used to characterize the correspondence between the episode plot data corresponding to the TV series and the plot summary data corresponding to the TV series; and / or

[0171] The second plot summary determination module is used to use the third language model to process the commentary subtitle data corresponding to the above-mentioned movie to obtain the plot summary data corresponding to the above-mentioned movie; the above-mentioned commentary subtitle data is the subtitle text obtained by character recognition of the commentary video corresponding to the movie; the above-mentioned third language model is used to rewrite, polish and correct errors in the above-mentioned commentary subtitle data.

[0172] Optionally, the above-mentioned description data may also include: plot description data corresponding to the movie.

[0173] Optionally, the above device may further include:

[0174] The push module is used to send the push message content corresponding to the above-mentioned film and television drama to the target users corresponding to the above-mentioned film and television drama.

[0175] In summary, the message processing device of the embodiment of the present invention uses the first language model to process the plot summary data and cast data corresponding to the film and television drama to obtain the push message content corresponding to the film and television drama. Since the embodiment of the present invention can generate the push message content based on the plot summary data and cast data corresponding to the film and television drama using the text generation capability of the first language model; therefore, the embodiment of the present invention can save the manpower cost consumed in determining the push message content and can improve the efficiency of determining the push message.

[0176] Furthermore, the embodiment of the present invention uses the subtitle data corresponding to the movie to determine the plot summary data corresponding to the movie, which can improve the richness and completeness of the plot summary data corresponding to the movie. When the richness and completeness of the plot summary data corresponding to the movie are improved, the embodiment of the present invention can improve the richness of the push message content corresponding to the movie.

[0177] Furthermore, the embodiment of the present invention uses the actor names in the cast data to replace the character names in the plot summary data, so that the push message content contains the actor names of the film and television drama instead of the character names. Since the popularity of actor names is usually higher than that of character names, the embodiment of the present invention can increase the attractiveness of the push message content to users.

[0178] An embodiment of the present invention further provides an electronic device that can implement the functions of the aforementioned storage node or scheduling node or ledger node or management node.

[0179] like Figure 6 As shown, the electronic device may include a processor 1001, a communication interface 1002, a memory 1003 and a communication bus 1004, wherein the processor 1001, the communication interface 1002, and the memory 1003 communicate with each other via the communication bus 1004.

[0180] Memory 1003, used for storing computer programs;

[0181] The processor 1001 is used to execute the program stored in the memory 1003, and implements the following steps:

[0182] Storing the storage object;

[0183] Sending the storage object to a coupled second storage node, so that the second storage node stores and / or forwards the storage object;

[0184] After the storage object is successfully stored, the object information of the storage object is sent to the ledger node, so that the ledger node saves the ledger information of the first storage node; the ledger information includes: the object information stored in the corresponding storage node.

[0185] The communication bus mentioned in the above electronic device can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The communication bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus.

[0186] The communication interface is used for communication between the above electronic device and other devices.

[0187] The memory may include a random access memory (RAM) or a non-volatile memory, such as at least one disk memory. Optionally, the memory may also be at least one message processing device located away from the aforementioned processor.

[0188] The above-mentioned processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0189] In another embodiment of the present invention, a computer-readable storage medium is provided. The computer-readable storage medium stores instructions. When the computer-readable storage medium is executed on a computer, the computer executes the message processing method described in any one of the above embodiments.

[0190] In another embodiment of the present invention, a computer program product including instructions is provided. When the computer program product is run on a computer, the computer executes the message processing method described in any one of the above embodiments.

[0191] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented by software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from a website site, computer, server or data center to another website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more available media integrated. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state hard disk Solid State Disk (SSD)), etc.

[0192] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0193] Each embodiment in this specification is described in a related manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0194] The above description is only a preferred embodiment of the present invention and is not intended to limit the protection scope of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention are included in the protection scope of the present invention.

Claims

1. A message processing method, characterized in that: The method comprises: Determine the plot summary data corresponding to the TV series according to the description data corresponding to the TV series; the description data includes: the episode plot data corresponding to the TV series, and / or the commentary subtitle data corresponding to the movie; Get the actor data corresponding to the film and television series; The first language model is used to process the plot summary data and cast data corresponding to the film and television drama to obtain the push message content corresponding to the film and television drama; the actor names in the cast data are used to replace the character names in the plot summary data; the first language model is used to characterize the correspondence between the plot summary data and cast data corresponding to the film and television drama and the push message content corresponding to the film and television drama.

2. The method according to claim 1, characterized in that The push message content includes: main title content and subtitle content; the use of the first language model to process the plot summary data and cast data corresponding to the film and television drama includes: Filling the plot summary data and the cast data into a prompt template to obtain a prompt text; The prompt text is input into the first large language model to obtain at least one main title content and at least one subtitle content output by the first large language model.

3. The method according to claim 2, characterized in that The prompt template includes: task description information, first step information, second step information and precaution information; The first step information corresponds to a generation step corresponding to the subtitle content, and the first step information includes: a subtitle case, a plot summary data item, a cast data item, a first character length range corresponding to the subtitle content, and first usage requirement information of the plot summary data item and the cast data item; the first usage requirement information is used to replace the character name in the plot summary data with the actor name in the cast data in the subtitle content; The second step information corresponds to the generation step corresponding to the main title content, and the second step information includes: a main title case, a plot summary data item, a cast data item, a second character length range corresponding to the main title content, and second usage requirement information of the plot summary data item and the cast data item; the second usage requirement information is used to replace the character name in the plot summary data with the actor name in the cast data in the main title content; The precaution information is used to detect the main title content and the subtitle content output by the first language model. If the detection result is that the main title content or the subtitle content contains both the actor name and the character name, the generation steps corresponding to the subtitle content and the generation steps corresponding to the main title content are executed.

4. The method according to claim 2, characterized in that: The first language model outputs a plurality of main title contents and a plurality of subtitle contents; and the using the first language model to process the plot summary data and cast data corresponding to the film and television drama also includes: Determining a relevance score between the main title content and the subtitle content; According to the relevance score, a target main title content and a target subtitle content are selected from the multiple main title contents and the multiple subtitle contents as push message content corresponding to the film and television drama.

5. The method according to claim 2, characterized in that: The training process of the first language model includes: pre-training and fine-tuning training; The fine-tuning training includes: Acquire a first training sample; the first training sample includes: an input sample and an output sample; the input sample includes: samples corresponding to the plot summary data and the cast data respectively; the output sample includes: a sample of the push message content; According to the training samples, fine-tuning training is performed on the pre-trained first language model; the first language model that has completed fine-tuning training is used to characterize the correspondence between the plot summary data and cast data corresponding to the film and television drama and the push message content corresponding to the film and television drama.

6. The method according to claim 1, characterized in that Determining the plot summary data corresponding to the film and television drama according to the description data corresponding to the film and television drama includes: Using the second largest language model, the episode plot data corresponding to the TV series is processed to obtain the plot summary data corresponding to the TV series; the second largest language model is used to characterize the correspondence between the episode plot data corresponding to the TV series and the plot summary data corresponding to the TV series; and / or The third language model is used to process the commentary subtitle data corresponding to the movie to obtain the plot summary data corresponding to the movie; the commentary subtitle data is the subtitle text obtained by performing character recognition on the commentary video corresponding to the movie; the third language model is used to rewrite, polish and correct errors in the commentary subtitle data.

7. The method according to any one of claims 1 to 6, characterized in that: The method further comprises: Send push message content corresponding to the film or TV series to target users corresponding to the film or TV series.

8. A message processing device, characterized in that: The device comprises: A plot summary determination module is used to determine the plot summary data corresponding to the TV series according to the description data corresponding to the TV series; the description data includes: the episode plot data corresponding to the TV series, and / or the commentary subtitle data corresponding to the movie; Actor acquisition module, used to obtain actor data corresponding to a film or TV series; The message content determination module is used to use the first large language model to process the plot summary data and cast data corresponding to the film and television drama to obtain the push message content corresponding to the film and television drama; the actor names in the cast data are used to replace the character names in the plot summary data; the first large language model is used to characterize the correspondence between the plot summary data and cast data corresponding to the film and television drama and the push message content corresponding to the film and television drama.

9. An electronic device, characterized in that: It includes a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other through the communication bus; Memory, used to store computer programs; A processor, for implementing the method steps described in any one of claims 1 to 7 when executing a program stored in a memory.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images