Video highlight generation method and device, equipment and storage medium

Abstract

The application provides a video highlight generation method and device, equipment and a storage medium, and the method comprises the following steps: determining barrage theme information, the barrage theme information comprises a barrage theme word and a time period corresponding to the barrage theme word, the barrage theme word is a phrase determined based on a plurality of barrage texts with an association relationship; taking a plot summary as a target plot summary, the plot summary has an intersection with the time period corresponding to the barrage theme word; determining a standard character name in the barrage theme word based on a character name in the barrage theme word; obtaining character information corresponding to the standard character name from a character information library based on the standard character name; and generating highlight information of a video segment based on the barrage theme word, the target plot summary and the character information corresponding to the standard character name. The method avoids the step of manually analyzing the barrage content, improves the plot highlight generation efficiency, avoids the influence of human subjective factors on the video highlights, and improves the quality of the plot highlights.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a method, apparatus, device and storage medium for generating video highlights. Background Technology

[0002] With the continuous development of internet technology, video websites have become an important platform for public leisure and entertainment. For example, users can play videos by accessing video websites and video applications.

[0003] In related technologies, staff members need to manually summarize and generalize the bullet comments (danmu) based on their understanding of them while watching TV series to identify the highlights of the video.

[0004] However, manually watching TV series and reading bullet comments takes a long time and requires a lot of manpower and time, resulting in low efficiency in generating video highlights. Summary of the Invention

[0005] To address the aforementioned technical issues, this application provides a method, apparatus, device, and storage medium for generating video highlights, which avoids the step of manually analyzing bullet screen content, improves the efficiency of generating plot highlights, avoids the influence of human subjective factors on video highlights, and improves the quality of plot highlights.

[0006] Firstly, this application provides a method for generating video highlights, which includes: determining bullet screen theme information, wherein the bullet screen theme information includes bullet screen theme words and the time period corresponding to the bullet screen theme words, wherein the bullet screen theme words are phrases determined based on multiple bullet screen texts with related relationships, and the time period corresponding to the bullet screen theme words is the time period during which the multiple bullet screen texts appear in the initial video, and the initial video is the video watched when the user inputs the bullet screen text; taking the plot summary whose time period corresponding to the plot summary intersects with the time period corresponding to the bullet screen theme words as the target plot summary, wherein the time period corresponding to the plot summary is the time period of the video segment corresponding to the plot summary in the initial video; determining standard character names in the bullet screen theme words based on character names in the bullet screen theme words, wherein character names refer to the names that identify a specific character appearing in the bullet screen theme words, and character names refer to the names that identify the characters shown in the initial video, and standard character names refer to character names that conform to preset naming rules; obtaining character information of standard character names from a character information database based on standard character names, wherein the character information database includes multiple character information, and character information refers to descriptive information related to the characters in the initial video; and generating highlight information of the video segment based on the bullet screen theme words, the target plot summary, and the character information of standard character names.

[0007] Secondly, this application provides a video highlight generation device, which includes: a theme information determination module, used to determine bullet screen theme information, the bullet screen theme information including bullet screen theme words and the time period corresponding to the bullet screen theme words, the bullet screen theme words being phrases determined based on multiple related bullet screen texts, the time period corresponding to the bullet screen theme words being the time period during which the multiple bullet screen texts appear in the initial video, the initial video being the video watched when the user inputs the bullet screen text; and a target plot summary acquisition module, used to obtain plot summaries whose time periods intersect with the time periods corresponding to the plot summaries as target plot summaries, the time periods corresponding to the plot summaries being the time periods of the video segments corresponding to the plot summaries in the initial video; the target plot summary is the plot summary obtained by obtaining the target plot summary. The character name extraction module is used to determine the standard character names in the bullet screen keywords based on the character names in the bullet screen keywords. The character name refers to the name that identifies a specific character in the bullet screen keywords, and the role name refers to the name that identifies the character shown in the initial video. The standard role name refers to the role name that conforms to the preset naming rules. The character information acquisition module is used to obtain the character information of the standard role name from the character information database. The character information database includes multiple character information, which refers to the descriptive information related to the character in the initial video. The highlight information generation module is used to generate the highlight information of the video clip based on the bullet screen keywords, the target plot summary, and the character information of the standard role name.

[0008] Thirdly, this application provides an electronic device comprising: one or more processors; a storage device for storing one or more programs; and, when the one or more programs are executed by the one or more processors, causing the one or more processors to implement the video highlight generation method as described in the first aspect above.

[0009] Fourthly, this application provides a storage medium, which may be a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the video highlight generation method as described in the first aspect above.

[0010] Fifthly, embodiments of this application provide a computer program product, which includes a computer program or instructions that, when executed by a processor, implement the video highlight generation method as described in any of the first aspects above.

[0011] The technical solution provided in this application has the following advantages compared with the prior art:

[0012] This application provides a method, apparatus, device, and storage medium for generating video highlights. The method includes: determining bullet screen theme information, which includes bullet screen theme words and the corresponding time periods. The bullet screen theme words are phrases determined based on multiple related bullet screen texts, and the corresponding time periods are the time periods during which the multiple bullet screen texts appear in an initial video, which is the video watched by the user when inputting the bullet screen text; and selecting plot summaries whose corresponding time periods overlap with the time periods corresponding to the plot summary words as target plot summaries, where the corresponding time periods are the video segments in the initial video. The time period in the video; based on the character names in the bullet screen keywords, determine the standard character names in the bullet screen keywords. The character name refers to the name that identifies a specific character in the bullet screen keywords. The role name refers to the name that identifies the character shown in the initial video. The standard role name refers to the role name that conforms to the preset naming rules; based on the standard role name, obtain the role information of the standard role name from the role information database. The role information database includes multiple role information. The role information refers to the descriptive information related to the character in the initial video; generate the highlights information of the video clip based on the bullet screen keywords, the target plot summary, and the role information of the standard role name.

[0013] In this embodiment, the plot summary and character information in the video are analyzed starting with the bullet comments (danmu) to generate the final video highlights. Since the bullet comment themes are derived from the analysis of user-input bullet comment text, the generated video highlights are more in line with public interest, improving the quality of the plot highlights. Furthermore, the introduction of character information during the generation of video highlights makes the final highlights more accurate, further improving their quality. In addition, this process avoids the step of manually analyzing bullet comment content, improving the efficiency of plot highlight generation and avoiding the influence of subjective human factors on the plot highlights, further enhancing their quality. Attached Figure Description

[0014] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0015] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 A flowchart illustrating a video highlight generation method provided in this application embodiment;

[0017] Figure 2 A flowchart illustrating the optimized video highlight generation method provided in this application embodiment;

[0018] Figure 3 A flowchart illustrating the standard role name determination method provided in this application embodiment;

[0019] Figure 4 A schematic diagram of a video highlight generation device provided in an embodiment of this application;

[0020] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0021] To better understand the above-mentioned objectives, features, and advantages of this application, the solution of this application will be further described below. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0022] Many specific details are set forth in the following description in order to provide a full understanding of this application, but this application may also be implemented in other ways different from those described herein; obviously, the embodiments in the specification are only some embodiments of this application, and not all embodiments.

[0023] The term "comprising" and its variations as used herein are open-ended inclusion, meaning "including but not limited to". The term "based on" means "at least partially based on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". Definitions of other terms will be given in the description below.

[0024] It should be noted that the concepts of "first" and "second" mentioned in this application are only used to distinguish different devices, modules or units, and are not used to limit the order of functions performed by these devices, modules or units or their interdependencies.

[0025] It should be noted that the terms "a" and "a plurality of" used in this application are illustrative rather than restrictive, and those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0026] The video highlight generation method provided in this application will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0027] Figure 1This is a flowchart illustrating a video highlight generation method according to an embodiment of this application. This embodiment is applicable to situations involving the extraction of video highlights. The method can be executed by a video highlight generation device, which can be implemented using software and / or hardware. This video highlight generation device can be configured in an electronic device, such as a smartphone, laptop, desktop computer, portable computer, or server.

[0028] like Figure 1 As shown, the video highlight generation method provided in this application embodiment mainly includes steps S101-S105.

[0029] S101. Determine the bullet screen theme information. The bullet screen theme information includes the bullet screen theme words and the time period corresponding to the bullet screen theme words. The bullet screen theme words are phrases determined based on multiple bullet screen texts with related relationships. The time period corresponding to the bullet screen theme words is the time period during which multiple bullet screen texts appear in the initial video. The initial video is the video watched when the user enters the bullet screen text.

[0030] In this context, "bullet comments" (danmu) are a feature where real-time comments sent by users float across the screen as subtitles during video playback. Bullet comment text refers to the content of these user-sent comments displayed in bullet comment format. In other words, bullet comment text can refer to real-time comments sent by users, which can consist of text and / or special symbols. Furthermore, users can send voice bullet comments while watching the video; the voice in these voice bullet comments is then processed through audio recognition to obtain the bullet comment text.

[0031] Bullet screen keywords are phrases determined based on multiple related bullet screen texts. These related texts can be understood as multiple bullet screen texts discussing the same or similar topics. In other words, bullet screen keywords are the common content or theme of bullet screen texts within a certain period. Bullet screen keywords are usually obtained by performing semantic analysis on multiple bullet screen texts using a pre-defined algorithm. Specifically, by collecting a certain number of bullet screen texts or within a certain time period, the Latent Dirichlet Allocation (LDA) algorithm or Latent Semantic Analysis (LSA) algorithm can be used to perform semantic analysis on the collected bullet screen texts to obtain bullet screen keywords.

[0032] The initial video is the video watched when the user enters the bullet comment text. This initial video can refer to an episode of a TV series, a movie, or a clip from a TV series or movie. During the playback of the initial video, multiple bullet comment texts sent by the user are received, and semantic analysis is performed on these texts to obtain one or more bullet comment keywords.

[0033] In an initial video, different bullet comment keywords may appear at different stages of the story as the plot unfolds. Therefore, multiple bullet comment keywords may exist in an initial video. For each bullet comment keyword, the video highlight generation method provided in this application embodiment can be used to generate its corresponding video highlight.

[0034] Since bullet comment keywords are obtained by aggregating multiple bullet comment texts, and these texts may have been sent by one or more viewers at different times, resulting in varying sending times, a single bullet comment keyword is formed when multiple viewers send bullet comment texts on the same topic or discussion point within the same initial video. Semantic analysis determines the time interval within which the multiple bullet comment texts containing the bullet comment keyword appear in the initial video; this time interval corresponds to the bullet comment keyword. The time interval corresponding to the bullet comment keyword can be the time range between the earliest and latest sending times of the multiple bullet comment texts that formed the keyword.

[0035] For example, the topic of the bullet screen is obtained by semantic analysis of bullet screen text 1, bullet screen text 2, ..., bullet screen text N. The topics or themes discussed by bullet screen text 1, bullet screen text 2, ..., bullet screen text N are the same. Among the N bullet screen texts, the i-th bullet screen text is sent earliest, when the initial video plays to 10:35, and the j-th bullet screen text is sent latest, when the initial video plays to 15:48. All other bullet screen texts in the N bullet screen texts are sent within the time interval [10:35, 15:48]. Therefore, the time interval corresponding to this bullet screen topic is [10:35, 15:48].

[0036] S102. The plot summary whose time period overlaps with the time period of the bullet screen keyword is taken as the target plot summary. The time period corresponding to the plot summary is the time period of the video segment corresponding to the plot summary in the initial video.

[0037] A plot summary is a textual summary that outlines the plot of a film or television work. In this application's embodiments, the plot summary may include a minute-level plot summary. A minute-level plot summary is a detailed description of the episode's content in chronological order, typically pinpointing key events or dialogues every minute. Minute-level plot summaries are primarily used to help people quickly understand or review the main plot of a movie, TV series, or video.

[0038] In practical applications, a plot summary typically refers to a minute-level summary of the plot of an episode of a TV series or a movie. The aforementioned target plot summary can be understood as a plot summary related to the keywords in the bullet comments. In other words, the plot described in the target plot summary is the same as or similar to the theme expressed by the bullet comment keywords, or the theme expressed by the bullet comment keywords is a part of the plot content in the target plot summary.

[0039] Because each important plot point or dialogue in the minute-by-minute plot summary is recorded and labeled with its exact appearance time in the initial video (e.g., "minute 5", "minute 23 45 seconds"), this helps to quickly locate specific scenes. The time period corresponding to the target plot summary is the time period in which the important plot points or dialogues included in the target plot summary appear in the initial video.

[0040] In one possible implementation, a minute-level plot summary of the initial video is obtained. The minute-level plot summaries whose time periods overlap with the time periods corresponding to the bullet comment keywords are used as the target plot summary. In other words, the time periods corresponding to the bullet comment keywords overlap with the time periods corresponding to the target plot summary.

[0041] In one possible implementation, a plot summary whose time period intersects with the time period corresponding to the bullet comment theme is used as the target plot summary. This includes: obtaining the start and end times of the time period corresponding to the plot summary; obtaining the start and end times of the time period corresponding to the bullet comment theme; taking the minimum value between the end time of the plot summary and the end time of the bullet comment theme as the first time point; taking the maximum value between the start time of the plot summary and the start time of the bullet comment theme as the second time point; if the difference between the first time point and the second time point is greater than a set time threshold, then the plot summary is used as the target plot summary.

[0042] The initial video's plot summary can be segmented to obtain multiple plot summaries. This segmentation can be based on set time intervals, with each plot summary corresponding to a specific video segment within that time interval. For example, an initial video of 45 minutes can be divided into 9 time intervals of 5 minutes each, with each interval corresponding to a plot summary. Alternatively, the segmentation can be based on scene transitions within the initial video. If a transition occurs, the plot summary is segmented at the transition point, with the summary before the transition forming one segment and the summary after the transition forming another. It should be noted that this embodiment only provides a brief overview of plot summary segmentation methods and is not intended to limit the scope of the analysis.

[0043] As mentioned above, each plot summary is derived from the overall plot summary of the initial video. Therefore, each plot summary has its corresponding time segment. This time segment is the period during which the corresponding video clip in the initial video appears.

[0044] As shown in Table 2, the time period corresponding to the first plot summary is The time period corresponding to the second plot summary is

[0045] Table 2

[0046]

[0047] If the time period corresponding to the topic of the bullet comments is The time period corresponding to the topic of the bullet comments Time periods corresponding to each plot summary Compare the time periods corresponding to the keywords in the bullet comments. Time period corresponding to the plot summary If there is an overlap, then that plot summary will be used as the target plot summary.

[0048] Furthermore, if multiple plot summaries correspond to the same time period as the bullet comment keywords, then the multiple plot summaries will be merged in chronological order to form the target plot summary.

[0049] In one possible implementation, taking one plot summary as an example, we introduce a method to determine whether the plot summary is the target plot summary. Each plot summary can be determined to be the target plot summary using the following method.

[0050] Get the start time of the time period corresponding to the topic of the bullet comments. The start time point in the time period corresponding to the plot summary The start time point of the time period corresponding to the topic of the bullet comments The start time point in the time period corresponding to the plot summary The maximum value in the time frame is used as the second time point.

[0051] Get the end time of the time period corresponding to the topic of the bullet comments. The end time point in the time period corresponding to the plot summary Take the end time point of the time period corresponding to the theme words of the bullet comments. The end time point in the time period corresponding to the plot summary Perform a size comparison and select the minimum value from the two end time points as the first time point.

[0052] Then, the difference between the first time point and the second time point is calculated. If the difference is greater than 0, it indicates that the time period corresponding to the plot summary and the time period corresponding to the bullet screen theme words overlap. If the difference is less than or equal to 0, it indicates that the time period corresponding to the plot summary and the time period corresponding to the bullet screen theme words do not overlap.

[0053] Specifically, formula (1) can be used to determine whether there is an overlap between the time period corresponding to a certain plot summary and the time period corresponding to the topic words in the bullet comments.

[0054]

[0055] in, This indicates the end time of the time period corresponding to the topic of the bullet comments. This indicates the end time of a specific time period corresponding to a plot summary. This indicates the start time of the time period corresponding to the topic of the bullet comments. This indicates the start time of a specific time period corresponding to a plot summary. This represents the minimum value between the two end times. This represents the maximum value between the two start times.

[0056] If formula (1) is greater than 0, meaning the difference between the smaller end time and the larger start time is greater than 0, then the time period corresponding to the bullet comment theme and the time period corresponding to the plot summary partially overlap. If formula (1) is less than or equal to 0, meaning the difference between the smaller end time and the larger start time is less than or equal to 0, then the time period corresponding to the plot summary begins after the time period corresponding to the bullet comment theme ends, or the time period corresponding to the plot summary ends before the time period corresponding to the bullet comment theme begins. In this case, the time period corresponding to the bullet comment theme and the time period corresponding to the plot summary do not overlap.

[0057] In this embodiment, the method of determining whether a plot summary is related to a bullet screen keyword is determined by whether there is an overlap between the time period corresponding to the bullet screen keyword and the time period corresponding to the plot summary. This provides a method for identifying target plot summaries, making the obtained target plot summaries more accurate and thus improving the accuracy of video highlights.

[0058] S103. Based on the character names in the bullet screen keywords, determine the standard character names in the bullet screen keywords. The character name refers to the name that identifies a specific character in the bullet screen keywords. The character name refers to the name that identifies the character shown in the initial video. The standard character name refers to the character name that conforms to the preset naming rules.

[0059] In this context, "character name" refers to the way a specific character is identified in the bullet screen's theme. This character name can be a standard name or a non-standard name. A standard name conforms to established naming rules and formats; specifically, a standard name follows the rule of surname plus given name, for example, "Zhang San" is a standard name, with the surname "Zhang" and the given name "San". A non-standard name refers to a character name that is not a standard name; it can also be understood as any form of address other than a standard name.

[0060] A character name refers to the name that identifies the character shown in the initial video. For example, the names of the male and female protagonists in a series are character names. Character names can include both non-standard and standard character names. For example, non-standard character names can include nicknames used by users for actors, nicknames used by users for characters, the character's identity in the series, etc. For example, a virtual character might be a princess of a dynasty, and users might also refer to this virtual character by the identity "Princess."

[0061] Furthermore, the standard name also includes the standard actor name, which refers to the actor's name that conforms to the established naming rules and formats. The actor name refers to the name of the performer who plays the role. For example, if the standard actor name is "Zhang San", the standard character name of the virtual character he plays in the initial video is "Zhao Wu".

[0062] Character names are extracted from the bullet comment keywords. These names are then matched against a standard character name database. If a match is found, the character name is considered a standard character name and is directly used as the standard character name in the bullet comment keywords. The standard character name database includes the standard character names of all characters from the initial video.

[0063] The character names are extracted from the bullet screen's theme keywords. If the character name is not a standard character name, it is matched against a standard actor name database. If a standard actor name is found, the corresponding standard character name is queried from the actor-role relationship database, and this queried standard character name is used as the standard character name in the bullet screen's theme keywords. The standard actor name database includes the standard actor names of all the actors who played the characters in the initial video. The actor-role relationship refers to the correspondence between virtual characters and the performers who played them, including the correspondence between standard character names. This correspondence can be stored in list form or data form; this embodiment does not limit the storage.

[0064] In this embodiment of the application, when the standard actor name is extracted from the bullet screen theme words, the actor role relationship is queried to determine the standard role name corresponding to the standard actor name, so as to obtain the role information of the standard role name, thereby improving the accuracy of the highlights information.

[0065] When the character name is a non-standard name, extract the standard character name of all characters from the target plot summary, perform semantic analysis on the non-standard name and the target plot summary, and obtain the standard character name corresponding to the character name.

[0066] When character names are non-standard, the reasons users use these names are highly personalized. In such cases, it's impossible to obtain standard character names simply by processing the bullet comment keywords or character names alone. Therefore, using a summary of the target plot to help determine the standard character names within the bullet comment keywords improves the accuracy of the key information.

[0067] S104. Based on the standard character name, obtain the character information of the standard character name from the character information database. The character information database includes multiple character information, which refers to the descriptive information related to the character in the initial video.

[0068] Character information, also known as character introduction information, can be understood as the description of the name, background, personality, characteristics, abilities, etc. of any virtual character in the initial video. Character information can help the large model to better understand the character.

[0069] In another possible implementation, since the character information stored in the character information database is recorded using standard character names, after extracting the standard character name from the bullet screen keywords, a query is performed in the character information database based on the standard character name. If a standard character name that matches the standard character name can be found, the character information corresponding to the standard character name is read from the character information database.

[0070] S105. Generate highlights information for video clips based on the keywords of the bullet comments, the target plot summary, and the character information of the standard character names.

[0071] Highlights of a video clip refer to information about segments that viewers might find interesting. Highlights include the reasons users post comments and / or plot phrases. Plot phrases are concise and powerful expressions that summarize or emphasize key points or turning points in the plot.

[0072] In one possible implementation, the character information, including the bullet comment keywords, the target plot summary, and the standard character names, is input into a large text-generating model. This model then outputs the reason why the user sent the bullet comment and the plot phrase corresponding to the bullet comment keyword.

[0073] In one possible implementation, pre-set prompts are retrieved. These prompts are primarily used to obtain specific information or answers when interacting with a large text-generating model. The prompts help the large text-generating model understand the user's needs and generate relevant answers or content.

[0074] Specifically, in the process of generating large-scale models for input text including bullet screen keywords, target plot summaries, and character information corresponding to standard character names, the prompt words corresponding to bullet screen keywords can be: "The viewer sent a bullet screen while watching a certain plot segment: []". The prompt words corresponding to the target plot summary can be: "The plot is as follows: []". The prompt words corresponding to character information can be: "Related character introduction: []".

[0075] For example, in a bullet screen with the keyword "aaa", the target plot summary is "bbbbb", and the character information is "Zhang San is ccccc, Li Si is ddddd". Here, "ccccc" represents Zhang San's character information, and "ddddd" represents Li Si's character information.

[0076] The data input into the text-generating large-scale model is: "A viewer sent a bullet comment while watching a certain scene: [aaa]; the scene is as follows: [bbbbb]; the relevant character introductions are: [Zhang San is ccccc, Li Si is ddddd]". Please analyze the reasons why the viewer sent the bullet comment and summarize the relevant plot phrases based on the given scene and dialogue. The text-generating large-scale model analyzes and processes the input data, and can output the reasons why users sent bullet comment comments and the brief plot phrases corresponding to the bullet comment keywords.

[0077] In one possible implementation, generating highlight information for video clips based on bullet screen keywords, target plot summaries, and character information of standard character names includes: obtaining pre-set prompt words, which indicate the type of plot highlights output by the text-generating large model; generating input information from the prompt words, bullet screen keywords, target plot summaries, and character information of standard character names, and inputting the input information into the generative large model; and receiving highlight information of video clips output from the generative large model, wherein the type of highlight information is consistent with the type indicated by the prompt words.

[0078] The cue words indicate the type of compelling information output by the text-generating large model. This type primarily refers to the tendency of the plot phrases. For example, the types indicated by the cue words can include: "exciting plot," "emotional resonance," "social issues," "educational value," "visual effects," etc. For example, "Phrase type: [Happy]," etc.

[0079] In this embodiment of the application, the prompt words, bullet screen theme words, target plot summary, and target object character information are simultaneously input into the text generation big model to instruct the text generation big model to mine the highlights information in the bullet screen theme words.

[0080] For example, the prompt word is "happiness" in the emotional resonance. The data input into the text-generating large-scale model is: "A viewer sent a bullet comment while watching a certain plot: [aaa]; the plot is as follows: [bbbbb]; the relevant character introductions are: [Zhang San is ccccc, Li Si is ddddd]"; phrase type: [happiness]; please analyze the reasons why the viewer sent the bullet comment and summarize the plot related to the bullet comment based on the given plot and dialogue." The text-generating large-scale model analyzes the reasons for the user's emotional resonance based on the input data and generates relevant plot phrases.

[0081] For example, the first prompt word is XXXX in the social issue. The data input into the large model is: "A viewer sent a bullet comment while watching a certain plot: [aaa]; the plot is as follows: [bbbbb]; the characters' dialogue is as follows: [Zhang San: "*******", Li Si: "*******"]; "Related character introductions are: [Zhang San is ccccc, Li Si is ddddd]; "Phrase type: [XXXX]"; please analyze the reasons why the viewer sent the bullet comment and summarize the plot phrases related to the bullet comment based on the given plot and dialogue." The text generation large model analyzes the reasons why the viewer raised the issue based on the input data and generates relevant plot phrases.

[0082] Based on the above embodiments, the data input into the text-generating large model may also include step prompts, which are used to instruct the large model on the steps of analyzing bullet screen keywords, target plot summaries, and character information such as standard character names.

[0083] For example, the step prompts could include: "Step 1: Extract characters from the bullet comment keywords; Step 2: Extract characters from the plot summary; Step 3: If there are characters in the bullet comment keywords who do not appear in the plot, try to identify them based on their character descriptions and replace them in the bullet comment keywords, providing the replaced bullet comment keywords; Step 4: Analyze the reasons for the bullet comment keywords obtained in Step 3; Step 5: If there is a plot related to the bullet comment keywords, extract the relevant plot phrases from the plot summary and dialogue. The answer should be given in JSON format, with the key being the bullet comment reason, the relevant plot, and a summary of the relevant events. If the bullet comment is not based on the plot, please answer: {"Bullet Comment Reason":"None","Related Plot":"None"}.

[0084] This application provides a method for generating video highlights. The method includes: determining bullet screen theme information, which includes bullet screen theme words and the time period corresponding to the bullet screen theme words. The bullet screen theme words are phrases determined based on multiple related bullet screen texts, and the time period corresponding to the bullet screen theme words is the time period during which the multiple bullet screen texts appear in the initial video. The initial video is the video watched when the user inputs the bullet screen text. Taking the plot summary whose time period intersects with the time period corresponding to the plot summary as the target plot summary, the time period corresponding to the plot summary is the time period of the video segment corresponding to the plot summary in the initial video. Based on the character names in the bullet screen theme words, determining the standard character names in the bullet screen theme words, where the character name refers to the name that identifies a specific character in the bullet screen theme words, and the character name refers to the name that identifies the character shown in the initial video. The standard character name refers to the character name that conforms to a preset naming rule. Based on the standard character name, obtaining the character information of the standard character name from a character information database, where the character information database includes multiple character information, and the character information refers to descriptive information related to the character in the initial video. Generating highlight information of the video segment based on the bullet screen theme words, the target plot summary, and the character information of the standard character name.

[0085] In this embodiment, the plot summary and character information in the video are analyzed starting with the bullet comments (danmu) to generate the final video highlights. Since the bullet comment themes are derived from the analysis of user-input bullet comment text, the generated video highlights are more in line with public interest, improving the quality of the plot highlights. Furthermore, the introduction of character information during the generation of video highlights makes the final highlights more accurate, further improving their quality. In addition, this process avoids the step of manually analyzing bullet comment content, improving the efficiency of plot highlight generation and avoiding the influence of subjective human factors on the plot highlights, further enhancing their quality.

[0086] Based on the above embodiments, the embodiments of this application further optimize the video highlight generation method, such as... Figure 2 As shown, the optimized video highlight generation method provided in this application embodiment mainly includes the following steps S201 to S206.

[0087] S201. Determine the bullet screen theme information. The bullet screen theme information includes the bullet screen theme words and the time period corresponding to the bullet screen theme words. The bullet screen theme words are phrases determined based on multiple bullet screen texts with related relationships. The time period corresponding to the bullet screen theme words is the time period during which multiple bullet screen texts appear in the initial video. The initial video is the video watched when the user enters the bullet screen text.

[0088] S202. The plot summary whose time period overlaps with the time period of the bullet screen keyword is taken as the target plot summary. The time period corresponding to the plot summary is the time period of the video segment corresponding to the plot summary in the initial video.

[0089] S203. Based on the character names in the bullet screen keywords, determine the standard character names in the bullet screen keywords. The character name refers to the name that identifies a specific character in the bullet screen keywords. The character name refers to the name that identifies the character shown in the initial video. The standard character name refers to the character name that conforms to the preset naming rules.

[0090] S204. Based on the standard character name, obtain the character information of the standard character name from the character information database. The character information database includes multiple character information, which refers to the descriptive information related to the character in the initial video.

[0091] Steps S201-S204 provided in this embodiment have the same execution flow as steps S101-S104 provided in the above embodiment. For details, please refer to the description in the above embodiment. They will not be repeated in this embodiment.

[0092] S205. Take dialogue content whose time period overlaps with the time period of the bullet screen keywords as target dialogue content. The time period corresponding to the dialogue content is the time period of the video segment corresponding to the dialogue content in the initial video.

[0093] The dialogue content refers to the spoken words of the various characters in the initial video. Multiple dialogue segments are obtained by segmenting the dialogue content of the initial video. The target dialogue content can be understood as the intersection of its corresponding time period and the time period corresponding to the bullet comment theme. In other words, the target dialogue content refers to the dialogue content of the characters within the video segment corresponding to the bullet comment theme. The dialogue content includes both the speaking character and the dialogue itself. For example, the dialogue content might be: "Zhang San: '*******'; Li Si: '*******'".

[0094] The segmentation method for dialogue content is the same as that for plot summary, and can be referred to the description in the above embodiments. It will not be repeated in the embodiments of this application.

[0095] As mentioned above, each dialogue segment is derived from the overall dialogue content corresponding to the initial video. Therefore, each dialogue segment has its corresponding time period. The time period corresponding to the dialogue segment is the time period of the corresponding video clip within the initial video.

[0096] As shown in Table 3, the time period corresponding to dialogue content 1 is The time period corresponding to dialogue content 2 is

[0097] Table 3

[0098]

[0099] If the time period corresponding to the topic of the bullet comments is Then the time period corresponding to the topic of the bullet comments. Time periods corresponding to each dialogue content Compare the time periods corresponding to the keywords in the bullet comments. Time period corresponding to the content of the conversation If there is an overlap, then the dialogue content is taken as the target dialogue content.

[0100] Furthermore, if multiple dialogues overlap with the time periods corresponding to the bullet screen keywords, the multiple dialogues will be merged in chronological order to form the target dialogue.

[0101] Taking one of the dialogue contents as an example, this article introduces a method to determine whether the dialogue content is the target dialogue content. Each dialogue content can be determined to be the target dialogue content using the following method.

[0102] Dialogue content whose time period intersects with the time period corresponding to the bullet comment theme is selected as target dialogue content. This includes: obtaining the start and end times of the time period corresponding to the dialogue content; obtaining the start and end times of the time period corresponding to the bullet comment theme; using the minimum of the end times of the dialogue content and the end times of the bullet comment theme as the third time point; using the maximum of the start times of the dialogue content and the start times of the bullet comment theme as the fourth time point; if the difference between the third and fourth time points is greater than a set time threshold, then the dialogue content is selected as the target dialogue content.

[0103] Get the start time of the time period corresponding to the topic of the bullet comments. and the start time of the time period corresponding to the dialogue content. The maximum value between the two start time points is selected as the fourth time point.

[0104] Get the end time of the time period corresponding to the topic of the bullet comments. and the end time of the time period corresponding to the dialogue content. The minimum value between the two end time points is selected as the third time point.

[0105] Then, the difference between the third and fourth time points is calculated. If the difference is greater than 0, it indicates that the time period corresponding to the target content overlaps with the time period corresponding to the bullet screen theme. If the difference is less than or equal to 0, it indicates that the time period corresponding to the target content does not overlap with the time period corresponding to the bullet screen theme.

[0106] Specifically, formula (2) can be used to determine whether there is an overlap between the time period corresponding to a target content and the time period corresponding to the topic word of the bullet screen.

[0107]

[0108] in, This indicates the end time of the time period corresponding to the topic of the bullet comments. This indicates the end time of a time period corresponding to a specific target content. The start time of the time period corresponding to the topic of the bullet comments. This indicates the start time of a time period corresponding to a specific target content. This represents the minimum value between the two end times. This represents the maximum value between the two start times.

[0109] If formula (2) is greater than 0, that is, the difference between the smaller end time and the larger start time is greater than 0, then the time period corresponding to the bullet comment theme and the time period corresponding to the target content partially overlap. If formula (1) is less than or equal to 0, that is, the difference between the smaller end time and the larger start time is less than or equal to 0, then it means that the time period corresponding to the target content starts after the time period corresponding to the bullet comment theme ends, or the time period corresponding to the target content ends before the time period corresponding to the bullet comment theme starts, then the time period corresponding to the bullet comment theme and the time period corresponding to the target content do not overlap.

[0110] S206. Based on the bullet screen keywords, target plot summary, character information of standard character names, and target dialogue content, generate the highlights information of the video clip.

[0111] In one possible implementation, the bullet comment keywords, target plot summary, character information of standard character names, and target dialogue content are input into a large text generation model, which then outputs the reason why the user sent the bullet comment and the plot phrase corresponding to the bullet comment keyword.

[0112] Specifically, in the process of generating the large-scale model for input text including bullet screen keywords, target plot summary, character information such as standard character names, and target dialogue content, the prompt words corresponding to the bullet screen keywords can be: "The viewer sent a bullet screen while watching a certain plot segment: []". The prompt words corresponding to the target plot summary can be: "The plot is as follows: []". The prompt words corresponding to the character information can be: "Related character introduction: []; the dialogue content is 'Zhang San: "*******", Li Si: "*******"'".

[0113] The S206 provided in this embodiment is the same as the S105 provided in the above embodiment, except that relevant information about the dialogue content is added to the prompt words and the data input into the text-generating large model. Other processing methods are the same, and you can refer to the description in the above embodiment for details.

[0114] In this embodiment, starting with the bullet comments, the system analyzes the character information in the plot summary and dialogue content video to generate the final video highlights. Since the bullet comment themes are derived from user-input text analysis, the generated highlights are more aligned with public interest, improving the quality of the plot highlights. Furthermore, the inclusion of character information in the highlight generation process makes the final highlights more accurate, further enhancing their quality. The inclusion of dialogue content in the highlight generation process also makes the final highlights more accurate, further improving their quality.

[0115] Based on the above embodiments, this application optimizes the step of "determining the standard character name in the bullet screen theme based on the character name in the bullet screen theme" in the above embodiments, such as... Figure 3 As shown, the optimized standard role name determination method provided in this application embodiment mainly includes the following steps S301 to S306.

[0116] S301. Extract character names from the bullet screen keywords.

[0117] Use natural language processing (NLP) tools to extract the names of characters from the keywords in the bullet comments, or use a trained deep learning model or machine learning model to extract the names of characters from the keywords in the bullet comments.

[0118] S302. Determine if the character's name is a standard name. If yes, proceed to S303; otherwise, proceed to S306. A standard name refers to a name that conforms to the preset naming rules.

[0119] Using a large number of standard name samples, an entity naming judgment model is trained. Character names in the bullet screen keywords are input into the entity naming judgment model, and the entity naming judgment model is used to determine whether the character name is a standard name.

[0120] S303. Determine if the character name is a standard character name. If yes, execute S304; otherwise, execute S305.

[0121] A standard character name database is pre-built. Character names extracted from bullet screen keywords are searched within this database. If a matching standard character name is found, the character name is considered a standard character name. If no matching standard character name is found, the character name is not a standard character name and further identification of the target character name is required.

[0122] S304. Use character names as standard character names in the bullet screen theme words.

[0123] If a standard character name matching the character name can be found in the standard character name database, it indicates that the character name is a standard character name, and this character name can be directly used as the standard character name in the bullet screen theme.

[0124] S305. Determine if the character name is a standard actor name. If so, execute S306.

[0125] A standard actor name database is pre-built. The database is searched based on the character name. If a standard actor name that matches the character name can be found, or if no standard actor name that matches the character name can be found in the database, it indicates that the character name is a standard actor name, and S306 is executed.

[0126] If no standard actor name matching the character name is found in the standard actor name database, it means that the character name is neither a standard actor name nor a standard character name.

[0127] S306. Query the standard character name corresponding to the standard actor name in the actor-role relationship; use the queried standard character name as the standard character name in the bullet screen theme words.

[0128] Based on this standard actor name, a search is performed in the actor role relationship, and the retrieved standard role name is used as the standard role name in the bullet screen theme keywords. For example, if the standard actor name extracted from the bullet screen theme keywords is "Zhang San", by searching in the actor role relationship, the standard role name played by "Zhang San" is found to be "Zhao Wu".

[0129] S307. Extract the standard character names included in the target plot summary; obtain the character information of the standard character names included in the target plot summary from the character information database; perform semantic analysis on the character names and character information to determine the standard character names corresponding to the character names.

[0130] If a character's name is not a standard name, the reasons users use non-standard names are highly personalized. In this case, it's impossible to obtain a standard character name simply by processing the bullet comment keywords or character names. Instead, the standard character name should be determined using the target plot summary and character information.

[0131] In one implementation, for each standard character name extracted from the target plot summary, the corresponding character information is obtained. Semantic analysis is then performed on the character names and their corresponding character information to determine the standard character names that correspond to each character name.

[0132] If a character's name is a non-standard name, such as "Princess," then semantic analysis is performed on each character's information to determine which character's information describes "Princess." Then, the standard character name from that information is used as the standard character name in the bullet comment keywords.

[0133] Based on the above embodiments, when the character name is a standard actor name or the character name is not a standard name, the character name in the bullet screen theme is replaced with the standard character name corresponding to the character name to obtain a new bullet screen theme; the highlights information of the video clip is generated based on the bullet screen theme, the target plot summary and the character information of the standard character name, including: the highlights information of the video clip is generated based on the new bullet screen theme, the target plot summary and the character information of the standard character name.

[0134] Furthermore, when the character name is the standard actor name, the standard character name is used to replace the character name to obtain new bullet screen keywords.

[0135] For example, if the bullet screen theme is "Zhang San returns on horseback," and the standard actor name extracted from the bullet screen theme is "Zhang San," then by querying the actor's role relationships, we find that the standard role name played by "Zhang San" is "Zhao Wu." Therefore, we replace the standard actor name "Zhang San" in the bullet screen theme with the standard role name "Zhao Wu," forming a new bullet screen theme: "Zhao Wu returns on horseback."

[0136] Since standard character names are usually used in plot summaries and dialogue information in practical applications, replacing the standard actor names in the bullet screen keywords with standard character names will generate new bullet screen topics. This will help the text-generating large model understand the bullet screen keywords and improve the accuracy of the highlights information.

[0137] Furthermore, when the character name is not a standard name, the character name is replaced with the corresponding standard character name to obtain new bullet screen keywords.

[0138] For example, if the bullet screen's theme is "General Zhao returns on horseback," and the non-standard name extracted from the theme is "General Zhao," then by searching the actor's role relationships, we find that the standard character name played by "General Zhao" is "Zhao Wu." Therefore, we replace the non-standard name "General Zhao" with the standard character name "Zhao Wu" to form a new bullet screen theme: "Zhao Wu returns on horseback."

[0139] In practical applications, plot information and dialogue information usually use standard character names to describe characters. In this embodiment, standard character names are used to replace non-standard names to obtain new bullet screen topics, which helps the large model understand the bullet screen keywords and improves the accuracy of the highlights information.

[0140] Figure 4 This is a schematic diagram of the structure of a video highlight generation device according to an embodiment of this application, such as... Figure 4 As shown, the video highlight generation device 30 provided in this application embodiment mainly includes: a theme information determination module 41, a target plot summary acquisition module 42, a target character name extraction module 43, a character information acquisition module 44, and a highlight information generation module 45.

[0141] The module 41, which determines the topic information of the bullet comments, includes bullet comment keywords and the corresponding time periods. The bullet comment keywords are phrases determined based on multiple related bullet comment texts, and the corresponding time periods are the time periods during which the multiple bullet comment texts appear in the initial video. The initial video is the video watched when the user inputs the bullet comment text. The module 42, which obtains the target plot summary, uses plot summaries whose time periods intersect with the time periods corresponding to the bullet comment keywords as target plot summaries. The time periods corresponding to the plot summaries are the time periods of the video segments corresponding to the plot summaries in the initial video. The module 43, which extracts the target character names, is used for... Based on the character names in the bullet screen keywords, the standard character names in the bullet screen keywords are determined. The character name refers to the name that identifies a specific character in the bullet screen keywords, and the role name refers to the name that identifies the character shown in the initial video. The standard role name refers to the role name that conforms to the preset naming rules. The role information acquisition module 44 is used to obtain the role information of the standard role name from the role information database. The role information database includes multiple role information, and the role information refers to the descriptive information related to the character in the initial video. The highlight information generation module 45 is used to generate the highlight information of the video clip based on the bullet screen keywords, the target plot summary, and the role information of the standard role name.

[0142] This application provides a video highlight generation device, which performs the following process: determining bullet screen theme information, including bullet screen theme words and the time period corresponding to the bullet screen theme words, wherein the bullet screen theme words are phrases determined based on multiple related bullet screen texts, and the time period corresponding to the bullet screen theme words is the time period during which the multiple bullet screen texts appear in the initial video, and the initial video is the video watched when the user inputs the bullet screen text; taking the plot summary whose time period corresponds to the plot summary and whose time period corresponds to the bullet screen theme words as the target plot summary, and the time period corresponding to the plot summary is the time period of the video segment corresponding to the plot summary in the initial video. The process involves: determining standard character names within the bullet screen keywords based on the character names in the bullet screen keywords; identifying a specific character name from the bullet screen keywords; identifying the character name from the character name displayed in the initial video; and defining standard character names as those conforming to preset naming rules. Based on these standard character names, the process retrieves character information from a character information database, which includes multiple character information entries. Character information refers to descriptive information related to the character in the initial video. Finally, the process generates highlight information for the video clip based on the bullet screen keywords, the target plot summary, and the character information from the standard character names.

[0143] In this embodiment, the plot summary and character information in the video are analyzed starting with the bullet comments (danmu) to generate the final video highlights. Since the bullet comment themes are derived from the analysis of user-input bullet comment text, the generated video highlights are more in line with public interest, improving the quality of the plot highlights. Furthermore, the introduction of character information during the generation of video highlights makes the final highlights more accurate, further improving their quality. In addition, this process avoids the step of manually analyzing bullet comment content, improving the efficiency of plot highlight generation and avoiding the influence of subjective human factors on the plot highlights, further enhancing their quality.

[0144] In one possible implementation, the target plot summary acquisition module 42 is specifically used to acquire the start and end times of the time period corresponding to the plot summary; acquire the start and end times of the time period corresponding to the bullet comment theme; take the minimum value between the end time of the plot summary and the end time of the bullet comment theme as the first time point; take the maximum value between the start time of the plot summary and the start time of the bullet comment theme as the second time point; if the difference between the first time point and the second time point is greater than a set time threshold, then the plot summary is taken as the target plot summary.

[0145] In one possible implementation, the device further includes: a target dialogue content acquisition module, used to take dialogue content whose time period corresponds to the dialogue content and whose time period corresponds to the bullet screen theme words as target dialogue content, and the time period corresponding to the dialogue content is the time period of the video segment corresponding to the dialogue content in the initial video; and a highlight information generation module 45, specifically used to generate highlight information of the video segment based on bullet screen theme words, target plot summary, character information of standard character names and target dialogue content.

[0146] In one possible implementation, the target dialogue content acquisition module is specifically used to acquire the start and end times of the time period corresponding to the dialogue content; acquire the start and end times of the time period corresponding to the bullet comment theme; take the minimum value between the end time of the dialogue content and the end time of the bullet comment theme as the third time point; take the maximum value between the start time of the dialogue content and the start time of the bullet comment theme as the fourth time point; if the difference between the third time point and the fourth time point is greater than a set time threshold, then the dialogue content is taken as the target dialogue content.

[0147] In one possible implementation, the target character name extraction module 43 is specifically used to extract character names from the bullet screen keywords; determine whether the character name is a standard name, where a standard name refers to a name that conforms to the set specifications and format naming rules; if the character name is a standard name, then determine whether the character name is a standard character name; if the character name is a standard character name, then use the character name as the standard character name in the bullet screen keywords.

[0148] In one possible implementation, the target character name extraction module 43 is specifically used to determine whether the character name is a standard actor name if the character name is not a standard character name. A standard actor name refers to an actor name that conforms to the set naming rules of specifications and format. An actor name refers to the name of the performer who plays the role. If the character name is a standard actor name, the module queries the standard character name corresponding to the standard actor name in the actor-role relationship. The queried standard character name is used as the standard character name in the bullet screen theme words.

[0149] In one possible implementation, the target character name extraction module 43 is specifically used to extract the standard character name included in the target plot summary if the character name is not a standard name; obtain the character information of the standard character name included in the target plot summary from the character information database; and perform semantic analysis on the character name and character information to determine the standard character name corresponding to the character name.

[0150] One possible implementation also includes: a name replacement module, used to replace the character name in the bullet screen theme words with the standard character name corresponding to the character name when the character name is a standard actor name or the character name is not a standard name, to obtain new bullet screen theme words; and a highlight information generation module 45, specifically used to generate highlight information of video clips based on the new bullet screen theme words, the target plot summary and the character information of the standard character name.

[0151] In one possible implementation, the highlight information generation module 45 is specifically used to obtain pre-set prompt words, which indicate the type of plot highlights output by the text-generating large model; generate input information by combining prompt words, bullet screen theme words, target plot summary, and standard character names, and input the input information into the generative large model; receive highlight information of video clips output from the generative large model, where the type of highlight information is consistent with the type indicated by the prompt words. The video highlight generation device provided in this application embodiment can execute the video highlight generation method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of executing the method.

[0152] Figure 5 This is a schematic diagram of the structure of an electronic device provided in this embodiment. The electronic device may include a video highlights generation device, such as... Figure 5 As shown, the electronic device 500 includes a processor 510, a memory 520, an input device 530, and an output device 540; the number of processors 510 in the electronic device can be one or more. Figure 5 Taking a processor 510 as an example; the processor 510, memory 520, input device 530, and output device 540 in the electronic device can be connected via a bus or other means. Figure 5 Taking the example of a connection between China and Israel via a bus.

[0153] The memory 520, as a computer-readable storage medium, can be used to store software programs, computer-executable programs, and modules, such as the program instructions / modules corresponding to the video highlight generation method in this embodiment of the invention. The processor 510 executes various functional applications and data processing of the electronic device by running the software programs, instructions, and modules stored in the memory 520, thereby implementing the video highlight generation method provided in this embodiment of the invention.

[0154] The memory 520 may primarily include a program storage area and a data storage area. The program storage area may store the operating system and at least one application program required for a given function; the data storage area may store data created based on terminal usage. Furthermore, the memory 520 may include high-speed random access memory and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other non-volatile solid-state storage device. In some instances, the memory 520 may further include memory remotely located relative to the processor 510, which can be connected to the electronic device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0155] Input device 530 can be used to receive input digital or character information, and to generate key signal inputs related to user settings and function control of the electronic device, and may include a keyboard, mouse, etc. Output device 540 may include a display device such as a screen.

[0156] This embodiment also provides a storage medium containing computer-executable instructions, which, when executed by a computer processor, are used to implement the video highlight generation method provided in this embodiment of the invention.

[0157] Of course, the computer-executable instructions provided in the embodiments of the present invention are not limited to the above-described method operations, but can also perform related operations in the video highlight generation method provided in any embodiment of the present invention.

[0158] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0159] It is worth noting that in the embodiments of the video highlight generation device described above, the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy differentiation and are not used to limit the scope of protection of the present invention.

[0160] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0161] The above are merely specific embodiments of this application, enabling those skilled in the art to understand or implement this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to these embodiments, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for generating video highlights, characterized in that, The method includes: Determine the topic information of the bullet comments, which includes bullet comment keywords and the time period corresponding to the bullet comment keywords. The bullet comment keywords are phrases determined based on multiple bullet comment texts with related relationships. The time period corresponding to the bullet comment keywords is the time period during which the multiple bullet comment texts appear in the initial video. The initial video is the video watched by the user when the user inputs the bullet comment text. The plot summary whose time period overlaps with the time period corresponding to the bullet screen theme keyword is used as the target plot summary, and the time period corresponding to the plot summary is the time period of the video segment corresponding to the plot summary in the initial video. Based on the character names in the bullet screen keywords, standard character names in the bullet screen keywords are determined. The character name refers to the name that identifies a specific character in the bullet screen keywords. The character name refers to the name that identifies the character shown in the initial video. The standard character name refers to the character name that conforms to the preset naming rules. Based on the standard character name, the character information of the standard character name is obtained from the character information database, wherein the character information database includes multiple character information, and the character information refers to descriptive information related to the character in the initial video; The video clip's highlights are generated based on the bullet screen keywords, the target plot summary, and the character information of the standard character names.

2. The method according to claim 1, characterized in that, Plot summaries that overlap with the time periods corresponding to the plot summary and the bullet screen keywords ， As a target plot summary, it includes: Obtain the start and end times of the time period corresponding to the plot summary; Obtain the start and end times of the time period corresponding to the topic of the bullet comments; The minimum value between the end time of the plot summary and the end time of the bullet screen theme word is taken as the first time point; The maximum value between the start time of the plot summary and the start time of the bullet screen keyword is taken as the second time point; If the difference between the first time point and the second time point is greater than a set time threshold, then the plot summary is taken as the target plot summary.

3. The method according to claim 1 or 2, characterized in that, Also includes: Dialogue content whose time period overlaps with the time period of the bullet screen keyword is taken as target dialogue content, and the time period corresponding to the dialogue content is the time period of the video segment corresponding to the dialogue content in the initial video. The process of generating highlight information for the video clip based on the bullet screen keywords, the target plot summary, and the character information of the standard character names includes: Based on the bullet screen keywords, the target plot summary, the character information of the standard character name, and the target dialogue content, the highlights information of the video clip is generated.

4. The method according to claim 3, characterized in that, Dialogue content whose time period overlaps with the time period corresponding to the bullet screen keywords is considered as target dialogue content, including: Obtain the start and end times of the time period corresponding to the dialogue content; Obtain the start and end times of the time period corresponding to the topic of the bullet comments; The minimum value between the end time of the dialogue content and the end time of the bullet screen keyword is taken as the third time point; The maximum value between the start time of the dialogue content and the start time of the bullet screen keyword is taken as the fourth time point; If the difference between the third time point and the fourth time point is greater than a set time threshold, then the dialogue content will be used as the target dialogue content.

5. The method according to claim 1, characterized in that, Based on the character names in the bullet screen keywords, the standard character names in the bullet screen keywords are determined, including: Extract the character names from the keywords in the bullet comments; Determine whether the character name is a standard name, where a standard name refers to a name that conforms to a preset naming rule; If the character name is a standard name, then determine whether the character name is a standard role name; If the character name is a standard character name, then the character name will be used as the standard character name in the bullet screen theme.

6. The method according to claim 5, characterized in that, Also includes: If the character name is not a standard character name, then it is determined whether the character name is a standard actor name. The standard actor name refers to the actor name that conforms to the preset naming rules. The actor name refers to the name of the performer who plays the role. If the character name is a standard actor name, query the standard character name corresponding to the standard actor name in the actor role relationship; The standard character names found in the query will be used as the standard character names in the bullet screen keywords.

7. The method according to claim 5, characterized in that, Also includes: If the character name is not a standard name, extract the standard character names included in the target plot summary; Retrieve character information including standard character names from the target plot summary from the character information database; Semantic analysis is performed on the character name and the role information to determine the standard role name corresponding to the character name.

8. The method according to claim 5 or 6, characterized in that, Also includes: If the character name is a standard actor name or the character name is not a standard name, replace the character name in the bullet screen theme with the standard role name corresponding to the character name to obtain a new bullet screen theme. Based on the bullet screen keywords, the target plot summary, and the character information of the standard character names, the highlight information of the video clip is generated, including: The video clip's highlights are generated based on the new bullet screen keywords, the target plot summary, and the character information of the standard character names.

9. The method according to claim 1, characterized in that, Based on the bullet screen keywords, the target plot summary, and the character information of the standard character names, the highlight information of the video clip is generated, including: Obtain pre-set prompt words, which are used to indicate the type of plot highlights output by the text-generating large model; The input information is generated from the prompt words, the bullet screen theme words, the target plot summary, and the character information of the standard character name, and then input into the generative large model; The system receives highlight information from video clips output from the generative large model, wherein the type of highlight information is consistent with the type indicated by the prompt word.

10. A video highlight generation device, characterized in that, The device includes: The topic information determination module is used to determine the topic information of the bullet screen. The topic information of the bullet screen includes the topic words of the bullet screen and the time period corresponding to the topic words. The topic words of the bullet screen are phrases determined based on multiple bullet screen texts with related relationships. The time period corresponding to the topic words of the bullet screen is the time period during which the multiple bullet screen texts appear in the initial video. The initial video is the video watched by the user when the user inputs the bullet screen text. The target plot summary acquisition module is used to take plot summaries whose time periods overlap with the time periods corresponding to the bullet screen keywords as target plot summaries, wherein the time periods corresponding to the plot summaries are the time periods of the video segments corresponding to the plot summaries in the initial video; The target character name extraction module is used to determine the standard character name in the bullet screen theme words based on the character names in the bullet screen theme words. The character name refers to the name that identifies a specific character in the bullet screen theme words. The character name refers to the name that identifies the character shown in the initial video. The standard character name refers to the character name that conforms to the preset naming rules. The character information acquisition module is used to acquire character information of the standard character name from the character information database based on the standard character name, wherein the character information database includes multiple character information, and the character information refers to descriptive information related to the character in the initial video; The highlight information generation module is used to generate highlight information for the video clip based on the bullet screen keywords, the target plot summary, and the character information of the standard character name.

11. An electronic device, characterized in that, The device includes: One or more processors; Storage device for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the video highlight generation method as described in any one of claims 1-9.

12. A storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the video highlight generation method as described in any one of claims 1-9.

Images