Movie plot selection method, device and storage medium

Abstract

The application discloses a film plot selection method and device and a storage medium. The method comprises the following steps: obtaining a plot summary, each minute-level summary and user viewing data of a to-be-analyzed film; extracting all plots from the plot summary; matching each minute-level summary with all plots to determine plots contained in each minute-level summary; and obtaining target plots corresponding to each minute-level summary of the to-be-analyzed film based on the user viewing data and the plots corresponding to each minute-level summary. The application can improve the accuracy of film plot selection.

Description

Technical Field

[0001] This application relates to the field of video processing technology, and in particular to methods, devices and storage media for selecting video storylines. Background Technology

[0002] Minute-level summaries of a film can describe the plot of each minute of the film in text form, providing basic data for applications such as plot search and recommendation generation.

[0003] However, current minute-level summary generation technologies typically produce minute-level summaries that include all plot points, lacking a clear focus. This makes it difficult for users to quickly identify and understand the key content of the video, thus impacting their search and recommendation experience. Furthermore, when directly used for search or recommendation generation, unimportant plot points negatively affect the overall effect, resulting in poor search and recommendation quality and performance. Summary of the Invention

[0004] This application provides a method, device, and storage medium for selecting film storylines, which can improve the accuracy of film storyline selection.

[0005] Firstly, this application provides a method for selecting film storylines, the method comprising: Obtain the plot summary, minute-by-minute summaries, and user viewing data of the film to be analyzed; Extract all storylines from the plot summary; Each minute-level summary is matched with all storylines to determine the storylines contained in each minute-level summary; Based on user viewing data and the storylines corresponding to each minute-level summary, the target storylines corresponding to each minute-level summary of the film to be analyzed are obtained.

[0006] The further technical solution involves matching each minute-level summary with all storylines to determine the storylines contained in each minute-level summary, including: The first language model is used to pre-match each minute-level summary with all storylines to obtain a preset number of related storylines corresponding to each minute-level summary; A list of plots corresponding to each minute-level summary is generated based on a preset number of related storylines. A second language model is used to perform fine matching between each minute-level summary and the corresponding list of each plot, and the important plot corresponding to each minute-level summary is output. Based on user viewing data and the storylines corresponding to each minute-level summary, the target storylines corresponding to each minute-level summary are obtained, including: Based on user viewing data and the key plot points corresponding to each minute-level summary, the target storyline corresponding to each minute-level summary is obtained.

[0007] The further technical solution involves obtaining the target storyline corresponding to each minute-level summary based on user viewing data and the key plot points corresponding to each minute-level summary. The method includes: Obtain the storyline corresponding to each important plot point; According to the same storyline, the consecutive minute-level summaries are merged to obtain minute-level summaries and key plot points corresponding to each storyline; Based on the minute-level summaries, key plot points, and user viewing data corresponding to each storyline, the target minute-level summaries and target storylines corresponding to each target minute-level summary are determined.

[0008] Its further technical solution is that user viewing data includes peak user interaction times, determining the target minute-level summary corresponding to each storyline and the target storyline corresponding to the target minute-level summary, including: Identify the most important plot points that correspond one-to-one with each storyline; The minute-level summaries corresponding to the most important plot points are designated as important minute-level summaries of the storyline, while the remaining minute-level summaries are designated as candidate minute-level summaries. Determine whether the time period of each candidate minute summary contains the peak time of user interaction. If the time period of any candidate minute summary contains the peak time of user interaction, then both the candidate minute summary and the important minute summary are used as the target minute summary of the storyline. Among them, the most important plot is used as the target storyline of the important minute summary. Based on the key plot points corresponding to the candidate minute-level summaries, the target storyline corresponding to the candidate minute-level summaries is determined.

[0009] A further technical solution involves determining whether the time periods of each candidate minute-level summary include peak user interaction times. The method also includes: If none of the candidate minute summaries contain peak user interaction times, then the important minute summaries will be used as the target minute summaries for the storyline, and the most important plot points will be used as the target storylines for the important minute summaries.

[0010] Its further technical solution is to extract all storylines from the plot summary, including: Extract all storylines from the plot summary, as well as all the plot points corresponding to each storyline; Each minute-level summary is matched with all storylines to determine the storylines contained in each minute-level summary, including: Each minute-level summary is matched with all the story plots corresponding to each storyline to obtain the story plots contained in each minute-level summary, as well as the storylines corresponding to the story plots.

[0011] The further technical solution involves extracting all storylines and all corresponding plot elements from the plot summary, including: Obtain the first preset prompt, wherein the first preset prompt includes the judgment conditions of the plot and the judgment conditions of the same storyline; Input the first preset prompt and plot summary into the second language model to obtain all the story plots corresponding to each storyline.

[0012] The further technical solution is that the criteria for judging the same storyline include character consistency, plot coherence, and overall coordination between plots.

[0013] Secondly, this application provides a device for selecting film storylines, which includes a memory and a processor. The memory stores a computer program for performing the steps of any of the methods described above in this application.

[0014] Thirdly, this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, is used to implement the above-described method for selecting film storylines.

[0015] The beneficial effects of this application are as follows: Unlike existing technologies, this application obtains the plot summary, minute-level summaries, and user viewing data of the film to be analyzed; extracts all storylines from the plot summary; matches each minute-level summary with all storylines to determine the storylines contained in each minute-level summary; by matching the plot summary and the minute-level summaries, it ensures that the storylines contained in the minute-level summaries correspond to the actual storylines in the film, avoiding the omission of the most important plots and making the generated minute-level summaries more complete. Then, based on the matched minute-level summaries and user viewing data, the target storyline corresponding to each minute-level summary is obtained. In other words, this application also considers user viewing data, adjusting and optimizing the content and focus of the minute-level summaries according to the user's actual viewing behavior and preferences, making the generated minute-level summaries more targeted and better meeting user needs and expectations, thereby improving the quality and effectiveness of search and recommendation. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein: Figure 1 This is a flowchart illustrating the first embodiment of the method for selecting film storylines provided in this application; Figure 2 This is a flowchart illustrating the second embodiment of the method for selecting film storylines provided in this application; Figure 3 This is a graph showing the relationship between the time periods of each minute-level summary provided in this application and the peak times of user interaction; Figure 4 This is a schematic diagram of the structural process of the film storyline selection method provided in this application; Figure 5 This is another flowchart illustrating the method for selecting film storylines provided in this application; Figure 6 This is a structural diagram of the film story selection device provided in this application; Figure 7 This is a schematic diagram of an embodiment of the computer-readable storage medium provided in this application. Detailed Implementation

[0017] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. It is understood that the specific embodiments described herein are only for explaining this application and not for limiting it. Furthermore, it should be noted that, for ease of description, only the parts related to this application are shown in the accompanying drawings, not all structures. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0018] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0019] Minute-level summaries of a film can describe the plot of each minute of the film in text form, providing basic data for applications such as plot search and recommendation generation.

[0020] However, current minute-level summary generation technologies typically produce minute-level summaries that include all plot points, lacking a clear focus. This makes it difficult for users to quickly identify and understand the key content of the video, thus impacting their search and recommendation experience. Furthermore, when directly used for search or recommendation generation, unimportant plot points negatively affect the overall effect, resulting in poor search and recommendation quality and performance.

[0021] Therefore, to address the technical problem in existing technologies where the generated minute-level summaries typically contain all plot points, leading to poor search and recommendation quality and effectiveness, this application provides a method for selecting film storylines. See the following embodiments for details.

[0022] The following section details the method for selecting film storylines provided in this application. Please refer to [link / reference needed]. Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the method for selecting film storylines provided in this application. The method includes: Step 110: Obtain the plot summary, minute-by-minute summaries, and user viewing data of the film to be analyzed.

[0023] Among them, the minute-level summary is a summary written for each segment of the film to be analyzed, which is divided into multiple segments according to the visual scene transition points.

[0024] Plot summaries can be written by professional operations staff after watching the film. While these summaries may contain fewer details than minute-long summaries, they are not affected by editing style because they do not need to correspond chronologically to the film's plot. Their logic is clearer.

[0025] User viewing data can be information about the time points of videos that have received a lot of comments or have been repeatedly watched by most users.

[0026] Step 120: Extract all storylines from the plot summary.

[0027] Each specific event and development in the story constitutes a plot.

[0028] In some embodiments, all storylines can be extracted from the plot summary first, and then the plot elements contained in each storyline can be extracted based on the storylines.

[0029] Among them, multiple storylines with logical connections and causal relationships can form a storyline, and a story consists of one or more storylines.

[0030] In some embodiments, the plot summary can be extracted based on cue words using a large language model, and may specifically include the following process: 1) Obtain the first preset prompt word.

[0031] 2) Input the first preset prompt words and plot summary into the second language model to obtain all the plots corresponding to each storyline.

[0032] The first preset prompt may include the judgment conditions for the plot and the judgment conditions for the same storyline.

[0033] The criteria for determining whether a storyline is the same can include character consistency, plot coherence, and overall coordination between plot lines.

[0034] For example, the first preset prompt word could be as follows: “Each specific event and development in a story constitutes a plot point. A series of logically related and causally connected plot points form a storyline. A story consists of one or more storylines.”

[0035] The following points can be used to determine whether two storylines belong to the same storyline: Judgment criteria: 1. Character consistency: Characters in the storyline should be the same or related.

[0036] 2. Plot coherence: If there is a clear cause-and-effect relationship or logical connection between two plots, then they may belong to the same storyline.

[0037] 3. Overall Coherence: From a macro perspective, the two plots can be reasonably combined to form a story with a complete narrative structure, rather than independent fragments, and they may belong to the same storyline.

[0038] <Plot Summary> === Please summarize the storyline in the story above by following these steps, and give the theme of the storyline and the plot related to the theme.

[0039] 1. Extract all the plot points from the story.

[0040] 2. Organize the plot into storylines. If a plot does not belong to any storyline, treat it as an independent storyline.

[0041] The answers for each storyline are given in JSON format, where the key is the theme of the storyline and the value is the relevant plot point.

[0042] By first extracting the storyline, we can obtain the main narrative thread and overall structure of the film to be analyzed. Then, we can extract the plot based on the storyline, that is, associate the plot with the corresponding storyline. After matching each minute-level summary with all the plots corresponding to each storyline, the matching results can include the plots contained in each minute-level summary, as well as the storylines corresponding to each plot, making the matching results clearer and more organized.

[0043] Step 130: Match each minute summary with all storylines to determine the storylines contained in each minute summary.

[0044] This can be achieved by first performing text analysis on each minute-level summary to extract key information such as roles, events, and emotions.

[0045] The extracted key information is matched with storylines. By comparing the information in the minute-level summary with all storylines, the storylines relevant to that minute-level summary are identified.

[0046] Step 140: Based on user viewing data and the storylines corresponding to each minute-level summary, obtain the target storylines corresponding to each minute-level summary of the film to be analyzed.

[0047] This process allows for the analysis of the storylines and corresponding user viewing data within each minute-level summary. Based on user viewing behavior, an attention or importance score is calculated for each storyline. Storylines with attention or importance scores exceeding a preset threshold are identified as target storylines for that minute-level summary. Target storylines represent the most interesting and crucial plot points for users and are of significant importance for subsequent search and recommendation.

[0048] This embodiment acquires the plot summary, minute-level summaries, and user viewing data of the film to be analyzed; extracts all storylines from the plot summary; matches each minute-level summary with all storylines to determine the storylines contained in each minute-level summary; by matching the plot summary and the minute-level summaries, it ensures that the storylines contained in the minute-level summaries correspond to the actual storylines in the film, avoiding the omission of the most important plots and making the generated minute-level summaries more complete. Then, based on the matched minute-level summaries and user viewing data, the target storyline corresponding to each minute-level summary is obtained. That is, this application also considers user viewing data, adjusting and optimizing the content and focus of the minute-level summaries according to the user's actual viewing behavior and preferences, making the generated minute-level summaries more targeted and better in line with user needs and expectations, thereby improving the quality and effectiveness of search and recommendation.

[0049] See Figure 2 , Figure 2 This is a flowchart illustrating a second embodiment of the method for selecting film storylines provided in this application. The method includes: Step 210: Obtain the plot summary, minute-by-minute summaries, and user viewing data of the film to be analyzed.

[0050] Step 220: Extract all storylines from the plot summary.

[0051] Steps 210 and 220 have the same or similar technical solutions as those in the above embodiments, and will not be described in detail here.

[0052] Step 230: Use the first language model to pre-match each minute-level summary with all storylines to obtain a preset number of related storylines corresponding to each minute-level summary.

[0053] The first language model can be an embedding model, and the preset number can be 2, 3, 4, 5, 6, etc. The specific number can be selected according to the actual situation, and this application does not limit it.

[0054] For example, a specific implementation method could be to use an embedding model to encode each minute-level summary and each storyline into a vector, and then determine the relevant storyline based on the cosine distance between each minute-level summary and its embedding vector.

[0055] For example, if the preset number is 5, then the 5 episodes whose cosine distance between the minute-level summary and its embedding vector is closest can be selected as relevant episodes.

[0056] Step 240: Based on the preset number of related storylines corresponding to each minute-level summary, form a list of plots corresponding to each minute-level summary.

[0057] The plot list can be as follows: 1. Fang Duobing stood out in the martial arts test and infiltrated Baichuan Academy.

[0058] 2. Fang Duobing learned that the Lingshan Sect was searching for someone and prepared to go and join in the fun.

[0059] 3. Fang Duobing refused Li Lianhua's offer and prepared to personally investigate the true cause of Wang Qingshan's death.

[0060] 4. Fang Duobing learned that the Lingshan Sect was searching for someone and prepared to go and join in the fun.

[0061] 5. Li Xiangyi became the number one martial arts master in the world, established the Sigu Sect, and aspired to become the leader of the martial arts world.

[0062] Step 250: Use a second language model to perform fine matching between each minute-level summary and the corresponding plot list, and output the important plots corresponding to each minute-level summary.

[0063] The second language model can be a deep learning model trained with a large amount of text data, such as a large language model.

[0064] When using a second language model to perform precise matching between minute-level summaries and their corresponding episode lists, certain prompt words can be used, such as: Plot excerpt: ### <Plot List> ### A certain episode of a TV series contains several of the above plot segments (listed in order, numbered 1 to <number of plot segments>). A storyline is considered a match if it occurs within a plot segment or if the plot segment belongs to a plot segment. Please find the best-matching plot segment for the following storyline.

[0065] Plot: ### ### The answer is the number of the story segment. If no matching story segment exists, the answer is -1.

[0066] For example, the plot list might be: 1. Fang Duobing stood out in the martial arts test and infiltrated Baichuan Academy.

[0067] 2. Fang Duobing learned that the Lingshan Sect was searching for someone and prepared to go and join in the fun.

[0068] 3. Fang Duobing refused Li Lianhua's offer and prepared to personally investigate the true cause of Wang Qingshan's death.

[0069] 4. Fang Duobing learned that the Lingshan Sect was searching for someone and prepared to go and join in the fun.

[0070] 5. Li Xiangyi became the number one martial arts master in the world, established the Sigu Sect, and aspired to become the leader of the martial arts world.

[0071] : Baichuan Academy held a martial arts competition. Fang Duobing, using the alias Yuan Jiankang, participated and won first place, becoming a newly recruited detective at Baichuan Academy. However, his true identity was revealed: he was the sole heir of the Fang family and Tianji Manor. The academy's leaders, wanting to avoid trouble, planned to send Fang Duobing home. Fang Duobing produced the sword left to him by Li Xiangyi as evidence. After some debate, the academy's leaders finally agreed to take him in. Fang Duobing decided to travel to Jiaozhou to investigate the matter.

[0072] Output: 1, 3.

[0073] The outputs 1 (Fang Duobing stands out in the martial arts test and infiltrates Baichuan Academy) and 3 (Fang Duobing refuses Li Lianhua's kindness and prepares to personally investigate the true cause of Wang Qingshan's death) are important plot points output by the second language model after performing a fine match between the minute-level summaries listed above and the corresponding plot list.

[0074] Since the number of options significantly impacts the accuracy of large language models when handling multiple-choice tasks, this embodiment first utilizes a first language model to pre-match each minute-level summary with all storylines. This involves initially matching each minute-level summary with all relevant storylines, filtering out storylines less relevant to each minute-level summary, narrowing the selection range for subsequent fine-tuning, and thus reducing the computational load and complexity of fine-tuning. Simultaneously, pre-matching provides more accurate and targeted candidate storylines for fine-tuning, thereby improving the overall accuracy of the matching process.

[0075] Step 260: Based on user viewing data and the key plot points corresponding to each minute-level summary, obtain the target storyline corresponding to each minute-level summary.

[0076] In some embodiments, step 260 may include the following steps: Step 261: Obtain the storyline corresponding to each important plot point.

[0077] For example, the key plot points corresponding to each minute-level summary are shown in Table 1 below: Table 1:

[0078] In Table 1, the plot numbers are the numbers of important plot points.

[0079] Step 262: Merge consecutive minute summaries according to the same storyline to obtain minute summaries and key plot points corresponding to each storyline.

[0080] See Table 2: Table 2:

[0081] As shown in Table 2, Summary 1 and Summary 2 are consecutive summaries belonging to Storyline 1. Therefore, Summary 1 and Summary 2 can be merged. Although Summary 3 and Summary 7 both belong to Storyline 2, they are not consecutive. Therefore, Summary 3 and Summary 7 should not be merged.

[0082] Since minute-level summaries are divided based on visual transition points, there may be situations where a storyline is divided into multiple scenes. Therefore, in order to select important plot points within a complete storyline, this embodiment merges consecutive minute-level summaries belonging to the same storyline.

[0083] Step 263: Based on the minute-level summaries, key plot points, and user viewing data corresponding to each storyline, determine the target minute-level summaries and target storylines corresponding to each storyline.

[0084] The user viewing data includes peak user interaction times, and step 263 may include the following process: 1) Identify the most important plot points that correspond to each storyline.

[0085] For example, the minute-level summaries corresponding to storyline 2 are summaries 3, 4 and 7, which correspond to time periods t5-t6, t7-t8 and t13-t14 respectively. The alternative plots are 2-0, 2-1, 2-2, 2-3 and 2-4.

[0086] The candidate episodes 2-0, 2-1, 2-2, 2-3, and 2-4 are selected as a plot list. A large language model is used to filter the most important episodes from these plot lists based on relevant cue words. The relevant cue words can be the following: Plot list: ### <Plot List> ### Important plot points refer to the points in the story where conflict, turning points, climaxes, or endings occur. Please identify the most important plot point in the list above; its number is "".

[0087] 2) The minute-level summary corresponding to the most important plot is used as the important minute-level summary of the storyline, and the remaining minute-level summaries are used as candidate minute-level summaries.

[0088] For example, in the above storyline 2, there are several possible outcomes: 1) If the most important plot is 2-1, then the minute summary corresponding to 2-1 (i.e., summary 3) is the important minute summary of storyline 2, and the other minute summaries (such as summary 4 and summary 7) are all candidate minute summaries.

[0089] Similarly, 2) if the most important plot is 2-3, then the minute-level summary corresponding to 2-3 (i.e. summary 4 and summary 3) is used as the important minute-level summary of storyline 2, and the remaining minute-level summaries (i.e. summary 7) are used as candidate minute-level summaries; 3) if the most important plot is 2-0, then the minute-level summary corresponding to 2-0 (i.e. summary 7) is used as the important minute-level summary of storyline 2, and the remaining minute-level summaries (i.e. summary 4 and summary 3) are used as candidate minute-level summaries.

[0090] 3) Determine whether the time period of each candidate minute-level summary includes peak user interaction times. If yes, proceed to steps 4) and 5). If no, proceed to step 6).

[0091] For example, if the most important plot point selected above is 2-1, then the important minute-level summary corresponding to storyline 2 is summary 3, with a time period of t5-t6. Candidate minute-level summaries are summary 4 and summary 7, with corresponding time periods of t7-t8 and t13-t14, respectively. Assume the relationship between the time periods of the minute-level summaries and peak user interaction times is as follows... Figure 3 As shown, where, Figure 3 The five-pointed stars in the diagram represent peak user interaction times, and the boxes represent minute-level summaries for each time period. It can be seen that time period t13-t14 in summary 7 covers peak user interaction times, while time period t7-t8 in summary 4 does not.

[0092] 4) Both candidate minute summaries and important minute summaries are used as target minute summaries for the storyline, with the most important plot being used as the target storyline for the important minute summaries.

[0093] For example, if the time period t13-t14 of summary 7 in process 3) covers the peak time of user interaction, then the candidate minute-level summary t13-t14 and the important minute-level summary t5-t6 initially identified will both be used as the target minute-level summary of the storyline.

[0094] Among them, the target storyline for the important minute-level summary t5-t6 is 2-1.

[0095] 5) Based on the important plot points corresponding to the candidate minute-level summaries, determine the target storyline corresponding to the candidate minute-level summaries.

[0096] As listed above, the important plots corresponding to summary 7, which is also selected as the target minute-level summary, are 2-0 and 2-4. Then, 2-0 and 2-4 are used as a new <plot list> and input into the large language model to select the most important plot. If the output is 2-0, then 2-0 is used as the target story plot corresponding to the candidate minute-level summaries t13-t14.

[0097] 6) Use important minute summaries as target minute summaries for the storyline, and use the most important plot points as target storylines for important minute summaries.

[0098] Taking the most important plot point 2-1 as an example, if the time period t13-t14 of summary 7 and the time period t7-t8 of summary 4 do not cover the peak time points of user interaction, then the minute-level summary corresponding to 2-1, namely summary 3 (corresponding time period t5-t6), is directly used as the important minute-level summary of story line 2, and summary 3 (corresponding time period t5-t6) is used as the target minute-level summary of the story line. The most important plot point 2-1 is the target story plot of the important minute-level summary, namely summary 3.

[0099] After determining the most important plot point for each storyline, the most important minute-level summary within each storyline is identified based on the minute-level summary matched to that plot point. Then, it's determined whether the remaining minute-level summaries (i.e., candidate minute-level summaries) cover peak user interaction times. Candidate minute-level summaries covering peak user interaction times are also considered important minute-level summaries for that storyline, and these important minute-level summaries are designated as the target minute-level summary for that storyline. In other words, the target minute-level summary includes the important minute-level summary and the candidate minute-level summaries covering peak user interaction times; specifically, it can be one, two, or three.

[0100] By selecting the most important minute summaries for each storyline, and using these minute summaries as the basis for generating search results or recommendations, the generated content is made entirely of importance when using only minute summaries containing important plot points. Furthermore, using important plot points as supplementary information during the generation process makes the generated content more focused and clear.

[0101] When processing minute-level summaries within a storyline, selecting the most important summaries for each storyline ensures that the generated content is of importance, thus improving the overall quality of the generated content. When these minute-level summaries are used as the basis for generating search results or recommendations, users can more efficiently access the core content of the storyline because these summaries include the most important plot points or cover peak user interaction times. Furthermore, utilizing these important plot points during content generation makes the focus of the generated content clearer.

[0102] Corresponding to the method for selecting film storylines in the above embodiments, this application also provides a device for selecting film storylines, see below. Figure 4 and Figure 6 As shown, the film story selection device 100 includes a storyline extraction module 10, a plot matching module 20, and an important plot selection module 30.

[0103] The input and output content of each module can be as follows: 1) Storyline Extraction Module 10: Input plot summary and output its storyline and the plot contained in each storyline.

[0104] The storyline extraction module can extract information using a first preset prompt word.

[0105] For example, the input is: Plot Summary: Ten years ago, the fastest sword in the martial arts world belonged to Li Xiangyi. He became the number one master in the land at the age of 15, established the Sigu Sect at 17, and began his quest for the title of martial arts alliance leader at 20, thus ending the chaos in the martial arts world. In an instant, he became a legendary figure. Some saw him as the hope of the Central Plains martial arts world, but most people only dreamed of defeating him. (Summary of episode 1 of the TV series Lotus Tower).

[0106] Output: {“The Duel Between Li Xiangyi and Di Feisheng”: [“Li Xiangyi became the number one martial arts master in the world, established the Sigu Sect, and aspired to become the leader of the martial arts alliance.”, “Li Xiangyi and Di Feisheng engaged in a fierce battle on a stormy night.”, “Li Xiangyi and Di Feisheng both fell into deep water and disappeared.”] "Li Lianhua's Medical Skills and Wisdom": ["Li Lianhua deduced that the butcher was pregnant.", "Li Lianhua exposed the butcher's secrets and stole a piece of pork rib.", "Li Lianhua revived a dead man lying in a coffin at Baichuan Courtyard.", "Li Lianhua revealed information about the Lingshan Sect."] "Fang Duobing becomes a criminal investigator and accepts a mission": ["Baichuan Academy selects new criminal investigators", "Fang Duobing becomes a new criminal investigator", "Fang Duobing takes on the Cicada Shedding and Ascending to Immortality case"],}.

[0107] 2) Plot matching module 20: This module takes the output of the minute-level summary and storyline extraction module as input and outputs the plot matching for each minute-level summary.

[0108] For example, the input is: <Plot List>: 1. Baichuan Academy selects new criminal investigators.

[0109] 2. Fang Duobing becomes a newly promoted criminal investigator.

[0110] 3. Fang Duobing took over the case of the cicada shedding its skin and ascending to immortality.

[0111] <Minute-long summary>: Baichuan Academy held a martial arts competition. Fang Duobing, using the alias Yuan Jiankang, participated and won first place, becoming a newly recruited detective at Baichuan Academy. However, his true identity was revealed: he was the sole heir of the Fang family and Tianji Manor. The academy's leaders, wanting to avoid trouble, planned to send Fang Duobing home. Fang Duobing produced the sword left to him by Li Xiangyi as evidence. After some debate, the academy's leaders finally agreed to take him in. Fang Duobing decided to travel to Jiaozhou to investigate the matter.

[0112] Output: 1,2,3.

[0113] 3) Important Plot Selection Module 30: Input minute-level summaries matching the storyline and high-energy data of user viewing, and output important minute-level summaries and their important plots.

[0114] For example, the input is: <Plot List> 1. Baichuan Academy selects new criminal investigators.

[0115] 2. Fang Duobing becomes a newly promoted criminal investigator.

[0116] 3. Fang Duobing took over the case of the cicada shedding its skin and ascending to immortality.

[0117] Output: 3.

[0118] Based on the aforementioned storyline extraction module 10, plot matching module 20, and important plot selection module 30, and in conjunction with the above embodiments, see [reference needed]. Figure 4 and Figure 5 The method for selecting film storylines provided in this application mainly includes the following processes: S1: Obtain the plot summary, minute-level summaries, and user viewing data of the film to be analyzed.

[0119] S2: Use the storyline extraction module to extract all storylines from the plot summary, as well as all the plot points corresponding to each storyline.

[0120] S3: Episode pre-matching based on the embedding model.

[0121] The process involves inputting all storylines corresponding to each storyline from the minute-level summary and the storyline extraction module into a first language model (such as an embedding model) for pre-matching to obtain a preset number of related storylines corresponding to each minute-level summary.

[0122] This allows for the creation of a list of plot points corresponding to each minute-level summary by generating a preset number of related storylines.

[0123] S4: Episode matching based on a large language model.

[0124] This involves using a second language model (such as a large language model) to perform precise matching between each minute-level summary and its corresponding plot list, and outputting the important plots corresponding to each minute-level summary.

[0125] For example, if the plot list contains a preset number of 5 related plots, then the important plots corresponding to the minute-level summary output after fine matching can be between 0 and 5, usually less than 5.

[0126] S5: Minute-level summary merging.

[0127] This involves merging consecutive minute-level summaries along the same storyline to obtain minute-level summaries and key plot points corresponding to each storyline. Please refer to the relevant content in the above embodiments for details.

[0128] S6: Determine the most important plot point for each storyline, and use the minute-level summary corresponding to the most important plot point as the important minute-level summary, and use the remaining minute-level summaries as candidate minute-level summaries.

[0129] S7: Determine whether the time period of each candidate minute-level summary includes peak user interaction times. If yes, execute S8 and S9; otherwise, execute S10.

[0130] S8: Take both the candidate minute-level summary and the important minute-level summary as the target minute-level summary.

[0131] S9: Determine the target storyline corresponding to the candidate minute-level summary.

[0132] S10: Use important minute-level summaries as target minute-level summaries.

[0133] This application obtains the plot summary, minute-level summaries, and user viewing data of the film to be analyzed; extracts all storylines from the plot summary; matches each minute-level summary with all storylines to determine the storylines contained in each minute-level summary; by matching the plot summary and the minute-level summaries, it ensures that the storylines contained in the minute-level summaries correspond to the actual storylines in the film, avoiding the omission of the most important plots and making the generated minute-level summaries more complete. Then, based on the matched minute-level summaries and user viewing data, the target storyline corresponding to each minute-level summary is obtained. That is, this application also considers user viewing data, adjusting and optimizing the content and focus of the minute-level summaries according to the user's actual viewing behavior and preferences, making the generated minute-level summaries more targeted and better meeting the user's needs and expectations, thereby improving the quality and effectiveness of search and recommendation.

[0134] like Figure 6As shown, the film story selection device 100 provided in this application further includes a memory 50 and a processor 40, wherein the memory 50 stores a computer program; the processor 40 is used to implement the control method of the film story selection method provided in any of the aforementioned method embodiments when executing the computer program.

[0135] Please see Figure 7 , Figure 7 This is a schematic diagram of an embodiment of the computer-readable storage medium provided in this application. The computer-readable storage medium 90 is used to store a computer program 91. When the computer program 91 is executed by a processor, it is used to implement the following method steps: Obtain the plot summary, minute-by-minute summaries, and user viewing data of the film to be analyzed; Extract all storylines from the plot summary; Each minute-level summary is matched with all storylines to determine the storylines contained in each minute-level summary; Based on user viewing data and the storylines corresponding to each minute-level summary, the target storylines corresponding to each minute-level summary of the film to be analyzed are obtained.

[0136] It is understood that when computer program 91 is executed by the processor, it is also used to implement the technical solutions of any embodiment of this application.

[0137] In the several embodiments provided in this application, it should be understood that the disclosed methods and devices can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and there may be other division methods in actual implementation. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed.

[0138] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0139] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0140] If the integrated units in the other embodiments described above are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause an electronic device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the methods in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0141] The above are merely embodiments of this application and do not limit the scope of this patent application. Any equivalent structural or procedural changes made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of this application.

Claims

1. A method for selecting film storylines, characterized in that, The method includes: Obtain the plot summary, minute-by-minute summaries, and user viewing data of the film to be analyzed; Extract all storylines from the aforementioned plot summary; Each minute-level summary is matched with all the storylines to determine the storyline contained in each minute-level summary; Based on the user viewing data and the storylines corresponding to each minute-level summary, the target storylines corresponding to each minute-level summary of the film to be analyzed are obtained. The step of matching each minute-level summary with all the storylines to determine the storyline contained in each minute-level summary includes: The first language model is used to pre-match each minute-level summary with all the storylines to obtain a preset number of related storylines corresponding to each minute-level summary; A list of plots corresponding to each minute-level summary is formed based on a preset number of related storylines; A second language model is used to perform a fine match between each minute-level summary and its corresponding plot list, and the important plot corresponding to each minute-level summary is output. The step of obtaining the target storyline corresponding to each minute-level summary based on the user viewing data and the storyline corresponding to each minute-level summary includes: Based on the user viewing data and the key plot points corresponding to each minute-level summary, the target storyline corresponding to each minute-level summary is obtained.

2. The selection method according to claim 1, characterized in that, The method for obtaining the target storyline corresponding to each minute-level summary based on the user viewing data and the key plot points corresponding to each minute-level summary includes: Obtain the storyline corresponding to each of the important plot points; According to the same storyline, the minute-level summaries of each storyline are merged to obtain minute-level summaries and key plot points corresponding to each storyline; Based on the minute-level summaries corresponding to each storyline, the key plot points, and the user viewing data, the target minute-level summaries corresponding to each storyline and the target storylines corresponding to the target minute-level summaries are determined.

3. The selection method according to claim 2, characterized in that, The user viewing data includes peak user interaction times. Determining the target minute-level summary corresponding to each storyline and the target storyline corresponding to the target minute-level summary includes: Identify the most important plot points corresponding to each of the aforementioned storylines; The minute-level summary corresponding to the most important plot is used as the important minute-level summary of the storyline, and the remaining minute-level summaries are used as candidate minute-level summaries; Determine whether the time period of each of the candidate minute-level summaries contains the peak time point of user interaction. If the time period of any candidate minute-level summary contains the peak time point of user interaction, then both the candidate minute-level summary and the important minute-level summary are used as the target minute-level summary of the storyline, wherein the most important plot is used as the target storyline of the important minute-level summary. Based on the key plot points corresponding to the candidate minute-level summaries, the target storyline corresponding to the candidate minute-level summaries is determined.

4. The selection method according to claim 3, characterized in that, The method for determining whether the time periods of each of the candidate minute-level summaries contain peak user interaction times further includes: If none of the candidate minute-level summaries contain the peak user interaction time point, then the important minute-level summary is taken as the target minute-level summary of the storyline, and the most important plot is taken as the target storyline of the important minute-level summary.

5. The selection method according to claim 1, characterized in that, The extraction of all storylines from the plot summary includes: Extract all storylines from the plot summary, as well as all plot points corresponding to each storyline; The step of matching each minute-level summary with all the storylines to determine the storyline contained in each minute-level summary includes: Each minute-level summary is matched with all storylines corresponding to each storyline to obtain the storylines contained in each minute-level summary and the storylines corresponding to the storylines.

6. The selection method according to claim 5, characterized in that, The extraction of all storylines from the plot summary, and all plot elements corresponding to each storyline, includes: Obtain a first preset prompt word, wherein the first preset prompt word includes the judgment conditions of the story plot and the judgment conditions of the same story line; Input the first preset prompt word and the plot summary into the second language model to obtain all the story plots corresponding to each story line.

7. The selection method according to claim 6, characterized in that, The criteria for determining whether a storyline is the same include consistency of characters, coherence of plot, and overall coordination between plot lines.

8. A device for selecting film storylines, characterized in that, The device for selecting the film's storyline includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the method as described in any one of claims 1-7.

9. A computer-readable storage medium, characterized in that, The storage medium stores a computer program that, when executed by a processor, can implement the method as described in any one of claims 1-7.

Images