Data processing method, electronic device, storage medium, and program product

Abstract

The application relates to the computer technical field and discloses a data processing method, an electronic device, a storage medium and a program product. The method comprises the following steps: obtaining a topic set and obtaining emotion labels of a plurality of candidate songs; the topic set comprises a plurality of original topics with a heat degree meeting preset conditions; emotional keywords of a plurality of selected topics are determined; the selected topic is an original topic selected from the topic set; the emotional keywords of each selected topic are aggregated to obtain an emotional keyword pool; a target song is selected from the plurality of candidate songs according to the matching degree between the emotion label of each candidate song and the emotional keyword pool; and a target video is generated according to the target song. Through the matching degree between the emotional keywords and the song emotion labels, the application realizes a non-random certainty song selection driven by a hot search, so that the created target video is more likely to cause user resonance, and the quality of the created video is improved.

Description

Technical Field

[0001] This application relates to the field of computer technology, specifically to data processing methods, electronic devices, storage media, and program products. Background Technology

[0002] In related technologies, to achieve automated or semi-automated production of short video content, it is typically necessary to aggregate trending topics from multiple data sources in real time and select background music and generate accompanying text based on manual configuration or simple tag matching. However, the heterogeneous topic data from different data platforms exhibits significant differences in field structure, update frequency, and trending calculation methods, leading to substantial computational pressure on the server during data cleaning, alignment, and standardization. Simultaneously, the candidate music library is enormous, making it difficult to achieve high-frequency computational matching across a vast number of songs using limited manual tagging methods, resulting in significant computational resource overhead. Furthermore, in the opening text generation stage, manual writing or generation using general text models requires multiple rounds of editing, verification, and feedback interactions between the terminal and server, generating additional signaling overhead and response latency. Therefore, how to achieve efficient and automated matching between trending topics and multimedia content elements while reducing server-side data processing pressure has become a pressing technical problem to be solved. Summary of the Invention

[0003] To address the aforementioned technical problems, this application provides a data processing method, an electronic device, a storage medium, and a program product.

[0004] In a first aspect, this application provides a data processing method, the method comprising: Obtain a set of topics and obtain the emotional tags of multiple candidate songs; the set of topics includes multiple original topics whose popularity meets preset conditions; Identify sentiment keywords for multiple selected topics; the selected topics are original topics chosen from the topic set. The sentiment keywords of each selected topic are aggregated to obtain a sentiment keyword pool; The target song is selected from the candidate songs based on the matching degree between the emotional tags of each candidate song and the emotional keyword pool. Generate a target video based on the target song.

[0005] In some alternative implementations, determining the sentiment keywords for multiple selected topics includes: For any original topic in the topic set, determine the sentiment keywords corresponding to the original topic; Based on the sentiment keywords corresponding to each of the original topics, the original topics are filtered to determine the topics to be determined; Determine the evaluation indicators for each of the proposed topics; the evaluation indicators are positively correlated with the popularity of the proposed topics. Topics whose evaluation indicators are greater than the indicator threshold are selected as undetermined topics.

[0006] In some optional implementations, determining the sentiment keywords corresponding to the original topic includes: The original topic is matched with a preset sentiment dictionary, and the sentiment words in the sentiment dictionary that match the original topic are used as the sentiment keywords corresponding to the original topic. If no sentiment word matching the original topic exists in the sentiment dictionary, then sentiment analysis is performed on the original topic based on the large language model to determine the corresponding sentiment keywords.

[0007] In some optional implementations, the step of filtering each original topic based on the sentiment keywords corresponding to each original topic to determine the candidate topics includes: For multiple original topics whose title similarity is greater than the first threshold, retain the original topic with the highest popularity. Determine the second similarity between the sentiment keywords of the retained original topic and the historical sentiment word set; the historical sentiment word set is used to record the sentiment keywords of the used topics; Original topics with a second similarity score less than the second threshold are designated as undetermined topics.

[0008] In some optional implementations, determining the evaluation metrics for each of the topics to be determined includes: The normalized popularity, category weight, and time factor determined based on the time difference between the current time and the collection time of the topic to be determined are determined; the category weight is used to represent the importance of the topic category to which the topic to be determined belongs, and the time factor and the time difference are negatively correlated. The evaluation index for the undetermined topic is determined based on the normalized popularity, the weight of its category, and the time factor; the evaluation index is positively correlated with the normalized popularity, the category weight, and the time factor.

[0009] In some optional implementations, selecting the target song from the candidate songs based on the matching degree between the emotion tags of each candidate song and the emotion keyword pool includes: Based on the attribute information of each candidate song, determine the configuration parameters of the candidate songs; The target song is selected from the candidate songs based on the configuration parameters of the candidate songs and the matching degree.

[0010] In some optional implementations, the method further includes: Identify target topics among the selected topics that match the emotional keywords with the emotional tags of the target song; The opening text of the target video is generated based on the emotional tags of the target song and the emotional keywords of the target topic.

[0011] In some optional implementations, generating the opening text of the target video based on the emotional tags of the target song and the emotional keywords of the target topic includes: Generate prompts containing the emotional tags of the target song, the emotional keywords of the target topic, and constraints on the creation method; the constraints on the creation method include the constraints on copywriting. The prompt words are input into the large language model to obtain the opening text generated by the large language model; the prompt words are used to instruct the large language model to generate the opening text according to the emotional tags of the target song and the emotional keywords of the target topic, in accordance with the constraints of the creation method.

[0012] In some optional implementations, the step of inputting the prompt word into a large language model and obtaining the opening text generated by the large language model includes: Input the prompt words into the large language model and obtain the initial text output by the large language model; The initial copy is validated; the validation includes: verifying whether the emotional tone of the copy is consistent with the emotional tags of the target song, and verifying whether the copy conforms to the constraints of the creative method. If the initial copy passes the copy verification, then the initial copy will be used as the opening copy; If the initial copy fails the copy verification, the prompt words are updated according to the verification result, and the copy is regenerated based on the updated prompt words until an opening copy that passes the copy verification is generated.

[0013] In some optional implementations, generating the target video based on the target song includes: Based on the source and category of the target topic, determine the corresponding video type; Based on the song information of the target song, the target song, and the opening text, generate a target video of the video type.

[0014] Secondly, this application provides an electronic device, including: a memory and a processor, which are communicatively connected to each other. The memory stores computer instructions, and the processor executes the computer instructions to perform the data processing method described in the first aspect or any corresponding embodiment.

[0015] Thirdly, this application provides a computer-readable storage medium storing computer instructions for causing a computer to perform the data processing method described in the first aspect or any corresponding embodiment.

[0016] Fourthly, this application provides a computer program product, including computer instructions for causing a computer to execute the data processing method described in the first aspect or any corresponding embodiment thereof.

[0017] The data processing method provided in this application selects highly trending topics from a data platform and extracts emotional keywords. By matching the emotional keywords with the emotional tags of songs, it can select target songs that have a high emotional match with the current public opinion environment. The selection process based on matching can establish a quantifiable, traceable, and reproducible association between trending topics and songs, ensuring that the song selection results are consistent with the direction of public opinion on that day, and achieving non-random deterministic song selection driven by trending topics, providing real and reliable data support for video creation. Furthermore, by matching the emotional keywords of the topic with the emotional tags of the song, suitable target songs can be selected with low computational load. The association between trending topics and content creation makes the created target videos more likely to resonate with users, which is conducive to improving the quality of the created videos. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the specific embodiments of this application or the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0019] Figure 1 This is a schematic diagram illustrating an application scenario according to an embodiment of this application; Figure 2 This is a schematic flowchart of a first type of data processing method according to an embodiment of this application; Figure 3 This is a schematic diagram of a second flow of a data processing method according to an embodiment of this application; Figure 4 This is a structural block diagram of a data processing apparatus according to an embodiment of this application; Figure 5This is a schematic diagram of the hardware structure of an electronic device according to an embodiment of this application. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0021] It is understood that before using the technical solutions disclosed in the various embodiments of this application, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in this application in an appropriate manner in accordance with relevant laws and regulations, and user authorization should be obtained.

[0022] For example, upon receiving a user's active request, a prompt message is sent to the user to explicitly inform them that the requested operation will require the acquisition and use of the user's personal information. This allows the user to independently choose whether to provide personal information to the software or hardware, such as the electronic device, application, server, or storage medium performing the operations of this disclosed technical solution, based on the prompt message.

[0023] As an optional but non-limiting implementation, in response to a user's active request, sending a prompt message to the user can be done via a pop-up window, where the prompt message can be presented in text format. Furthermore, the pop-up window can also include a selection control allowing the user to choose "agree" or "disagree" to provide personal information to the electronic device.

[0024] It is understood that the above notification and user authorization process are merely illustrative and do not constitute a limitation on the implementation of this disclosure. Other methods that comply with relevant laws and regulations may also be applied to the implementation of this disclosure.

[0025] It is understood that the data involved in this technical solution (including but not limited to the data itself, the acquisition or use of the data) shall comply with the requirements of relevant laws, regulations and related provisions.

[0026] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0027] Before providing a detailed description of the embodiments of this application, some of the terms and concepts involved in the embodiments of this application will be explained. These explanations are intended to make the embodiments of this application easier to understand and should not be considered as limiting the scope of protection claimed in this application.

[0028] (1) Hot Search: refers to the real-time hot topic list of each platform, which reflects the collection of content that users pay the most attention to at the current time. It has the characteristics of strong timeliness and emotional focus.

[0029] (2) Hook copy: The core text at the beginning of a short video (e.g., the first 3 seconds) used to attract users to stay. It should be short and impactful, with the goal of reducing the user's swipe rate.

[0030] (3) BPM (Beats Per Minute): A unit of musical tempo, which is the core parameter for measuring the speed of a song.

[0031] (4) Skip Rate: The percentage of users who swipe up to skip a video after watching it on a video platform. It is a core negative indicator for measuring the attractiveness of the beginning of a video.

[0032] As a mature content format, short lyric videos possess core competitiveness in two dimensions: firstly, the relevance of the background music to the user's current emotions; and secondly, the instant appeal of the opening text to the target audience. Current content production methods primarily rely on human experience and judgment, which presents several problems: (1) The selection of music is highly subjective and lacks data support. Content creators usually choose background music based on their personal familiarity with the music and their subjective feelings. They cannot systematically assess the emotional resonance potential of a song in the current public opinion environment, resulting in the selection of music being out of touch with the user's current emotional state.

[0033] (2) The disconnect between trending topics and content creation. Currently, in the content production process, the analysis of trending topics, music selection, and copywriting are independent manual processes, lacking a systematic connection mechanism. Creators need to manually browse trending content on multiple platforms and then transform it into content strategies based on subjective experience, resulting in low information utilization efficiency.

[0034] Music platforms can use collaborative filtering or content feature matching algorithms for music recommendations, but their recommendations are based on individual users' historical behavior and cannot be targeted to select songs based on trending topics of the day. They also lack the ability to integrate with content operation processes.

[0035] In addition, some trending search aggregation tools can access trending data from multiple data platforms, but these tools only provide data display functions, do not have analytical capabilities, and cannot form an automated connection with downstream content production processes.

[0036] As one optional application scenario in the embodiments of this application, such as Figure 1 As shown, application 101 is installed in terminal device 110, and user 130 can interact with application 101 through terminal device 110 and / or access device of terminal device 110.

[0037] For example, application 101 can be any application, such as a video platform application, that provides video creation functionality. Figure 1 In the application scenario shown, if application 101 is active, the terminal device 110 can display the interface 102 of application 101. The interface 102 may include various pages that application 101 can provide, such as interactive pages, settings pages, query pages, etc.

[0038] In some embodiments, terminal device 110 is communicatively connected to server 120 to provide services to application 101. Terminal device 110 may be a mobile terminal, fixed terminal, or portable terminal, including but not limited to mobile phones, desktop computers, laptop computers, multimedia tablets, e-book devices, gaming devices, or any combination thereof, including accessories and peripherals of these devices or any combination thereof. In some embodiments, terminal device 110 may also support any type of interface, and server 120 may be various types of computing systems or servers capable of providing computing power, including but not limited to mainframes, edge computing nodes, and computing devices in cloud environments.

[0039] It should be noted that, Figure 1 This is merely an example of an application scenario and does not limit the scope of protection of this application.

[0040] The embodiments of this application will be described below with reference to the accompanying drawings. It should be understood that the pages shown in the drawings are merely examples, and various page designs are possible in practice. The various graphic elements on the page may have different arrangements and different visual representations, one or more elements may be omitted or replaced, and one or more other elements may also be present; no limitations are imposed on the embodiments of this application. Furthermore, the embodiments are primarily described below with reference to terminal device 110. It should be understood that the actions described relative to terminal device 110 can be performed by application 101 on terminal device 110, or can be performed by application 101 in conjunction with its server (e.g., server 120).

[0041] This application provides a data processing method that selects trending topics from a data platform and extracts emotional keywords. By matching the emotional keywords with the emotional tags of songs, target songs with high emotional relevance to the current public opinion environment can be selected. The selection process based on matching can establish a quantifiable, traceable, and reproducible association between trending topics and songs, ensuring that the song selection results are consistent with the direction of public opinion on that day, and achieving non-random deterministic song selection driven by trending topics, providing real and reliable data support for video creation. Furthermore, the association between trending topics and content creation makes the created target videos more likely to resonate with users, which is conducive to improving the quality of created videos.

[0042] According to an embodiment of this application, a data processing method embodiment is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0043] This embodiment provides a data processing method that can be used in the aforementioned terminal devices, such as mobile terminals. Figure 2 This is a flowchart of a data processing method according to an embodiment of this application, such as... Figure 2 As shown, the process includes the following steps.

[0044] Step S201: Obtain a topic set and obtain the emotional tags of multiple candidate songs; the topic set includes multiple original topics whose popularity meets preset conditions.

[0045] In this embodiment, in a video creation scenario, relevant topics can be obtained based on popularity, thereby forming a corresponding topic set. For ease of description, the obtained topics are referred to as original topics, meaning that the topic set includes multiple original topics.

[0046] This involves retrieving high-trending original topics for the current time period (e.g., today) from relevant data platforms. Specifically, data platforms typically offer trending topics functionality, which includes multiple highly popular real-time topics, allowing the extraction of original topics. Furthermore, pre-setting relevant topic retrieval criteria allows for the acquisition of multiple original topics from the data platform based on these criteria.

[0047] For example, the preset conditions could be: retrieve all topics in the trending searches, retrieve the top K topics in the trending searches (e.g., K=10, 20, etc.), or prioritize topics with a popularity exceeding a preset popularity value as the original topics. The specific conditions can be determined based on business needs.

[0048] Furthermore, it is possible to initially identify songs that may be used to create videos, i.e., candidate songs; for example, some songs can be selected from the local music library as candidate songs.

[0049] Each candidate song is pre-labeled with a corresponding emotion tag, which is a marker to categorize the song according to its emotional dimension, such as "longing," "healing," "sweetness," and "anger." Emotion tags can be pre-labeled manually or automatically; this embodiment does not limit this approach. Generally, each song is tagged with at least one emotion tag.

[0050] Optionally, original topics can be obtained from multiple data platforms in parallel to form a multi-source fusion of hot topics.

[0051] For example, multiple data platforms such as Weibo, video platforms, and Q&A platforms can be used as data sources to collect trending topics from various data sources, thereby obtaining original topics from multiple aspects and creating a multi-source topic collection.

[0052] Step S202: Determine the sentiment keywords for multiple selected topics; the selected topics are the original topics chosen from the topic set.

[0053] In this embodiment, the number of original topics in the topic set is generally large. To reduce the amount of subsequent processing, a portion of the original topics can be selected as usable topics, i.e., selected topics. For example, N selected topics can be selected, where N can be 10, 15, 20, etc. It can be understood that if the number of original topics in the topic set is relatively small, all original topics can also be selected topics.

[0054] For any selected topic, performing sentiment analysis can identify the corresponding sentiment keywords. These sentiment keywords are core words representing the emotional tendency of the relevant trending topic, and can indicate the public sentiment of the day to a certain extent, which can be used for subsequent matching with the song's sentiment tags.

[0055] For example, relevant emotional keywords can be selected from the title of a chosen topic. These emotional keywords could be, for example, longing, loneliness, or healing, and are specifically related to the content of the chosen topic.

[0056] For a given topic, one or more sentiment keywords may be identified. This means the relationship between a topic and its sentiment keywords can be one-to-one or one-to-many, depending on the actual content of the topic. This embodiment does not limit the number of sentiment keywords corresponding to a topic.

[0057] Step S203: Aggregate the sentiment keywords of each selected topic to obtain a sentiment keyword pool.

[0058] For different selected topics, various emotional keywords can be aggregated according to certain rules to obtain a set containing the aggregated emotional keywords, i.e., an emotional keyword pool.

[0059] Specifically, the sentiment keywords identified from various selected topics may be duplicated. In this case, duplicate sentiment keywords can be deleted, thus forming a set containing sentiment keywords from each selected topic, i.e., a sentiment keyword pool. It can be understood that the sentiment keywords in this sentiment keyword pool are all different from each other.

[0060] For example, if topic A includes sentiment keywords a1 and a2, and topic B includes sentiment keywords b1 and a1, meaning that both contain the same sentiment keyword a1, then after deduplication, the resulting sentiment keyword pool includes sentiment keywords a1, a2, and b1.

[0061] Step S204: Select the target song from multiple candidate songs based on the matching degree between the emotional tags of each candidate song and the emotional keyword pool.

[0062] For any candidate song, the matching degree between the candidate song's emotion tag and the emotion keyword pool can be determined, thus enabling the selection of the optimal song, i.e., the target song, from multiple candidate songs. For example, the candidate song with the highest matching degree can be used as the target song.

[0063] This process involves calculating the intersection of the candidate song's emotion tags and the emotion keyword pool, and determining the matching degree based on the number of each emotion keyword in the intersection. Furthermore, it involves calculating the similarity between the candidate song's emotion tags and the matching emotion words in the emotion keyword pool, and determining the matching degree based on the combined similarity of each emotion tag.

[0064] For example, the similarity between the emotional tag of the candidate song (which may be one or more) and each emotional word in the emotional keyword pool can be calculated. The emotional word with the highest similarity and greater than a certain threshold is taken as the emotional word that matches the emotional tag (the number of such emotional words is the size of the intersection). Furthermore, the similarity of each emotional tag is weighted and summed to obtain the matching degree between the emotional tag of the candidate song and the emotional keyword pool.

[0065] Step S205: Generate the target video based on the target song.

[0066] In this embodiment, the target song identified above matches the emotional sentiment of a trending topic for the current time period. When generating a corresponding video based on this target song, the emotional matching degree between the video and the trending topic can be improved. Specifically, background music can be generated based on the target song, thereby generating a target song containing this background music. This makes the background music of the generated target video more consistent with the user's current mood and more likely to attract the user.

[0067] In this embodiment, the target song can be directly used as the background music for the target video, or the chorus or its corresponding accompaniment can be extracted from the target video and used as the background music. This embodiment does not limit the method of generating the background music. After obtaining the background music, it can be used as the background audio for the video to be published, i.e., the target video.

[0068] The data processing method provided in this embodiment selects trending topics from a data platform and extracts emotional keywords. By matching the emotional keywords with the emotional tags of songs, it can select target songs that have a high emotional match with the current public opinion environment. The selection process based on matching degree can establish a quantifiable, traceable, and reproducible association between trending topics and songs, ensuring that the song selection results are consistent with the direction of public opinion on that day, and realizing non-random deterministic song selection driven by trending topics, providing real and reliable data support for video creation. Furthermore, by matching the emotional keywords of the topic with the emotional tags of the song, a suitable target song can be selected with low computational load. The association between trending topics and content creation makes the created target videos more likely to resonate with users, which is conducive to improving the quality of the created videos.

[0069] This embodiment provides a data processing method that can be used in the aforementioned terminal devices, such as mobile terminals. Figure 3 This is a flowchart of a data processing method according to an embodiment of this application, such as... Figure 3 As shown, the process includes the following steps.

[0070] Step S301: Obtain a topic set and obtain the emotional tags of multiple candidate songs; the topic set includes multiple original topics whose popularity meets preset conditions.

[0071] Please see details Figure 2 Step S201 of the illustrated embodiment will not be described again here.

[0072] In this embodiment, original topics can be obtained from multiple data platforms such as Weibo, video platforms, and Q&A platforms.

[0073] For example, a video platform's topic categories could include: popular TV dramas, popular movies, popular variety shows, popular pairings (CP), popular captions, popular songs, etc. Different categories correspond to different emotional tendencies in the content.

[0074] For example, popular TV series / movies / variety shows reflect the emotions of the content and are adapted to popular video formats; popular couples reflect emotions such as love, longing, and sweetness; popular copywriting directly contains emotional expressions and can serve as a direct source of emotional keywords; popular songs indicate users' music consumption preferences for the day.

[0075] For the Weibo platform, its raw data includes a category field. This field can be used for pre-filtering, focusing on retaining topics in the categories of emotion, dramas, and variety shows, while filtering out categories such as finance and government affairs that are unrelated to the creation of emotional content, thereby improving the signal-to-noise ratio in subsequent processing.

[0076] For Q&A platforms, their trending topics, presented in the form of questions, possess unique value for sentiment analysis. Question titles can be pre-processed: identifying questions containing sentiment-driven expressions such as "why," "how to cope," and "have you ever," and using these as desired topics.

[0077] For each original topic collected from different platforms, a unified structured data can be used to represent it. For example, a corresponding topic object can be set for each original topic, which includes: source platform, topic title, category tag (corresponding to topic category), popularity value, collection timestamp, etc.

[0078] Furthermore, a unified emotional dictionary can be established, including terms such as longing, healing, sweetness, loneliness, anger, anxiety, nostalgia, inspiration, sadness, and joy. This allows data from different platforms to be uniformly mapped to the same emotional word system, enabling a cross-platform matching mechanism between trending topics and song emotions.

[0079] Optionally, in step S301, obtaining the emotional tags of multiple candidate songs may specifically include: for each song in the preset music library, filtering based on the song's metadata to select multiple candidate songs, and determining the emotional tag of each candidate song. The metadata includes at least one of the song's beat count, power index, and catchy markers.

[0080] Specifically, the metadata of each song is read from the structured index file of the music library (e.g., songs.json). The structured index file records metadata fields such as the title, artist, beat count (BPM), mood tag, duration, power index, catchy tag, and file path of each song.

[0081] Hard filtering can be performed based on song metadata to select candidate songs whose melodies meet the requirements.

[0082] For example, the filtering criteria could be: BPM range: 90-130 (the optimal tempo range for short video platforms); chorus burst power index: valid (judged by the energy field in the song's metadata or by manual annotation); catchy intro marker: valid (marked by the intro_hook field).

[0083] Songs that do not meet the above screening criteria can be directly excluded; only songs that meet the criteria will be considered as candidate songs for subsequent use.

[0084] Step S302: Determine the sentiment keywords for multiple selected topics; the selected topics are the original topics chosen from the topic set.

[0085] Please see details Figure 2 Step S202 of the illustrated embodiment will not be described again here.

[0086] In some alternative implementations, step S302, “determining sentiment keywords for multiple selected topics,” may include steps a1 to a4.

[0087] Step a1: For any original topic in the topic set, determine the sentiment keywords corresponding to the original topic.

[0088] In this embodiment, after determining the topic set, sentiment analysis can be performed on each original topic to extract sentiment keywords. For example, each original topic can be labeled with 1 to 3 sentiment keywords for subsequent matching calculations with song sentiment tags.

[0089] Optionally, step a1, "determining the sentiment keywords corresponding to the original topic," may include steps a11 to a12.

[0090] Step a11: Match the original topic with the preset sentiment dictionary, and use the sentiment words in the sentiment dictionary that match the original topic as the sentiment keywords corresponding to the original topic.

[0091] Step a12: If there is no sentiment word matching the original topic in the sentiment dictionary, then perform sentiment analysis on the original topic based on the large language model to determine the sentiment keywords corresponding to the original topic.

[0092] In this embodiment, the extraction of emotional keywords is implemented according to a two-layer extraction logic.

[0093] The first layer (rule layer) maintains the sentiment dictionary and matches it with various original topics through regular expression matching and other methods to identify sentiment words with high confidence. For example, the word "breakup" in the topic title matches the sentiment words "longing" / "sadness", "sweet" → "sweet" / "love", "healing" → "healing" / "peace", and then the sentiment words identified in the sentiment dictionary are used as the sentiment keywords of the corresponding original topics.

[0094] The second layer (semantic layer): For the original topics that the rule layer failed to match, the large language model interface is called again to judge the sentiment tendency, and the output is a structured list of sentiment keywords.

[0095] Since relying solely on the title text of a topic has limitations, this embodiment adopts a two-layer extraction method, with the rule dictionary layer and the semantic model layer complementing each other to jointly ensure the reliability of sentiment keyword extraction; the large model has a strong ability to handle colloquialisms and new words, and is suitable for real-time changing hot search scenarios.

[0096] Step a2: Based on the sentiment keywords corresponding to each original topic, filter the original topics to determine the topics to be determined.

[0097] In this embodiment, when collecting original topics from multiple trending search lists obtained from multiple data platforms, there may be situations where different original topics are quite similar. In this case, the original topics can be filtered to improve the quality of the topics used subsequently.

[0098] Optionally, step a2, "screening each original topic based on the sentiment keywords corresponding to each original topic and determining the topics to be determined," may include steps a21 to a23.

[0099] Step a21: For multiple original topics whose first similarity between topic titles is greater than the first threshold, retain the original topic with the highest popularity.

[0100] Step a22: Determine the second similarity between the sentiment keywords of the retained original topic and the historical sentiment word set; the historical sentiment word set is used to record the sentiment keywords of the topics that have been used.

[0101] Step a23: Select original topics with a second similarity less than the second threshold as undetermined topics.

[0102] In this embodiment, the original topics are filtered by deduplication, which may involve two deduplication dimensions.

[0103] First, internal deduplication is performed among the original topics, that is, duplicate original topics are removed. Specifically, topics with identical titles are directly filtered out to achieve precise deduplication; in this process, the similarity between each original topic can be calculated (e.g., cosine similarity, Jaccard similarity, etc.). If the first similarity of two original topics is greater than a preset first threshold (e.g., 0.8, 0.9, etc.), it means that the two are duplicates, and in this case, only the original topic with higher popularity can be retained.

[0104] The first similarity between topic titles can be determined directly based on the text of the two topic titles, or it can be determined based on the sentiment keywords extracted from the topic title text.

[0105] In this embodiment, a set of sentiment keywords for recording used topics is maintained, namely, the historical sentiment word set. For the original topics after internal deduplication, a second similarity between the sentiment keywords of each original topic and the historical sentiment word set can be determined. Topics with a second similarity greater than a second threshold (e.g., 0.8, 0.9, etc., which can be the same as or different from the first threshold) are removed, and only the original topics with a similarity less than the second threshold are retained. The topics that are ultimately retained are the undetermined topics required for subsequent processing.

[0106] It is understandable that the target topic determined later is the topic used in this instance. Therefore, after creating the corresponding video based on the target topic, the target topic becomes a topic that has already been used, and its corresponding sentiment keywords will be added to the historical sentiment keyword set to achieve rolling deduplication.

[0107] In addition, for each emotional keyword in the historical emotional keyword set, a certain survival time (such as 7 days, one month, etc.) can be set. After the survival time is reached, the emotional keyword is deleted, so that the topic corresponding to the emotional keyword can be used to generate videos again after a period of time.

[0108] Step a3: Determine the evaluation indicators for each pending topic; the evaluation indicators are positively correlated with the popularity of the pending topics.

[0109] Step a4: Select topics that are pending topics and whose evaluation indicators are greater than the indicator threshold.

[0110] For each deduplicated pending topic, an evaluation metric can be determined based on its popularity. This metric can, to some extent, indicate the degree of match between the topic and the current online public opinion environment or sentiment. Therefore, based on the evaluation metrics of each pending topic, further screening can be performed to select several pending topics with higher evaluation metrics as the final selected topics. For example, N selected topics can be chosen from multiple pending topics.

[0111] The threshold value for this indicator can be a preset fixed value or a value set based on actual circumstances. For example, if it is necessary to select the top N undetermined topics as selected topics, then the threshold value corresponds to the evaluation indicator of the Nth undetermined topic.

[0112] Optionally, step a3, “determining the evaluation criteria for each pending topic,” may include steps a31 to a32.

[0113] Step a31: Determine the normalized popularity, category weight, and time factor of the pending topic based on the time difference between the current time and the collection time of the pending topic; the category weight is used to represent the importance of the topic category to which the pending topic belongs, and the time factor and the time difference are negatively correlated.

[0114] Step a32: Determine the evaluation index of the topic to be determined based on the normalized popularity, the weight of the topic category, and the time factor; the evaluation index is positively correlated with the normalized popularity, the category weight, and the time factor.

[0115] In this embodiment, in addition to using the popularity of the topic, a composite evaluation index can also be obtained by weighting based on the topic category to which the topic belongs and the corresponding time factor.

[0116] Specifically, the popularity of an undetermined topic can be normalized to obtain a normalized popularity, for example, normalized to the [0,1] interval.

[0117] Each pending topic belongs to a certain topic category, and each topic category has a pre-set corresponding category weight; for example, the weight of content-related categories such as emotions / dramas / variety shows is higher than the weight of unrelated categories such as finance / politics.

[0118] Furthermore, based on the time difference between the current time and the collection time of the pending topic, a factor representing the degree of time decay can be determined, namely the time factor. For example, the time factor remains at 1.0 within 1 hour of collection, decays to 0.8 after 6 hours, halves after 24 hours, and is discarded after 24 hours (i.e., the time factor is 0).

[0119] For any undetermined topics collected from various data platforms, evaluation metrics can be determined using the unified method described above. For example, evaluation metric = normalized popularity × category weight × time factor. Finally, N suitable topics are selected from the topics collected from multiple platforms.

[0120] Step S303: Aggregate the sentiment keywords of each selected topic to obtain a sentiment keyword pool.

[0121] Please see details Figure 2 Step S203 of the illustrated embodiment will not be described again here.

[0122] Step S304: Select the target song from multiple candidate songs based on the matching degree between the emotional tags of each candidate song and the emotional keyword pool.

[0123] Please see details Figure 2 Step S204 of the illustrated embodiment will not be described again here.

[0124] Optionally, step S304 above, "selecting the target song from multiple candidate songs based on the matching degree between the emotion tags of each candidate song and the emotional keyword pool", may include steps b1 to b2.

[0125] Step b1: Determine the configuration parameters of the candidate songs based on the attribute information of each candidate song.

[0126] Step b2: Select the target song from multiple candidate songs based on the configuration parameters and matching degree of the candidate songs.

[0127] In this embodiment, in the process of screening songs, in addition to initially selecting candidate songs based on the song's metadata according to the above screening method, further screening is required based on the matching degree between the song and emotional keywords.

[0128] Specifically, each candidate song possesses attribute information that characterizes its importance. This attribute information may include, for example, the popularity of the candidate song. For any candidate song, parameters—configuration parameters—can be determined based on its attribute information to refine the match between the candidate song and the sentiment keyword pool. After determining the configuration parameters for each candidate song, the match between the candidate song's sentiment tag and the sentiment keyword pool can be refined based on these parameters. The target song is then selected from multiple candidate songs based on the refinement results.

[0129] For example, the corrected result is the corrected matching degree, and the candidate songs with the highest corrected matching degree can be used as the target songs.

[0130] Specifically, step S304 may include: removing candidate songs whose usage interval is less than a time threshold; determining the weight of each remaining candidate song based on the popularity of the selected topic or the category weight of the topic category; adjusting the matching degree between the emotional tags and the emotional keyword pool of the candidate songs based on their weights; and selecting the candidate song with the highest adjusted matching degree as the target song.

[0131] In this embodiment, candidate songs can be further filtered to remove those with usage intervals less than a time threshold; for example, the usage interval of the same song must exceed a set threshold (e.g., 7 days) to avoid repeatedly pushing the same song. Alternatively, candidate songs can be sorted based on their matching degree, retaining the M candidate songs with the highest matching degree. By performing a secondary filtering of candidate songs, the subsequent computational load can be further reduced.

[0132] For each candidate song remaining after the second round of screening, the corresponding selected topic can be determined. For example, the intersection of the emotional tags of the candidate song and the emotional keyword pool can be determined, and the candidate topic that best matches the intersection of the emotional tags can be taken as the selected topic corresponding to the candidate song; or, the similarity (e.g., the size of the intersection) between the emotional tags of the candidate song and the emotional keyword pools of each selected topic can be calculated, and the selected topic with the highest similarity can be taken as the selected topic corresponding to the candidate song.

[0133] For a selected topic corresponding to a candidate song, the popularity of the selected topic and / or the category weight of its category can be determined and used as attribute information for the candidate song, and the weight of the candidate song can be calculated; the higher the popularity and category weight of the selected topic, the higher the weight of the candidate song. The calculated weight of the candidate song can then be used as a configuration parameter.

[0134] After determining the weight of the candidate songs, for example, the weight of the candidate song can be multiplied by the matching degree to obtain the adjusted matching degree between the emotional tag of the candidate song and the emotional keyword pool. Then, it can be determined which candidate song corresponds to the highest matching degree after adjustment, and that candidate song can be used as the final target song.

[0135] In this embodiment, the process of selecting target songs relies entirely on the target song's metadata, including BMP and emotion tags, and there is no random selection, ensuring that the selection decision process is traceable and reproducible. Furthermore, adjusting the matching degree through song weights effectively avoids situations where songs with the same matching degree cannot be selected.

[0136] Step S305: Identify the target topic among multiple selected topics whose emotional keywords match the emotional tags of the target song.

[0137] Besides background music, the opening caption (hook) is also a crucial video element. Current caption generation solutions primarily rely on manual writing or generic text generation models, which cannot guarantee caption quality or relevance to the song's mood. This can easily result in inefficient captions that sound like advertisements or marketing jargon. If users make decisions based on trending topics, the decision-making process lacks transparency and struggles to ensure efficiency and consistency. Furthermore, some content generation solutions use content IDs or URLs for deduplication, lacking semantic deduplication capabilities based on topic sentiment. This can lead to the repeated use of topics with similar emotional types but different expressions, causing viewer fatigue.

[0138] In this embodiment, after the target song is determined, an opening script is generated based on the relevant information of the target song, which can ensure the relevance of the opening script to the mood of the song.

[0139] Specifically, for a given target song, a matching target topic can be determined from multiple selected topics. The emotional keywords of this target topic must match the emotional tags of the target song. For example, the match between the emotional keywords of the target topic and the emotional tags of the target song is the strongest, exceeding the match between the emotional keywords of other selected topics and the emotional tags of the target song.

[0140] Similar to the method described above for determining the selected topics corresponding to candidate songs, for example, the intersection of the emotional tags of the target song and the emotional keyword pool can be determined, and the undetermined topic that best matches the intersection of the emotional tags can be taken as the selected topic corresponding to the target song, i.e., the target topic; or, the similarity (e.g., the size of the intersection) between the emotional tags of the target song and the emotional keyword pools of each selected topic can be calculated, and the selected topic with the highest similarity can be taken as the selected topic corresponding to the target song, i.e., the target topic.

[0141] Step S306: Generate the opening text for the target video based on the emotional tags of the target song and the emotional keywords of the target topic.

[0142] In this embodiment, the emotional tags of the target song (e.g., longing, loneliness, etc.) are used as the main keywords, and the emotional keywords of the target topic are used to provide contextual references to generate an opening text that matches the emotional tags and emotional keywords. This opening text can then be used as the text for the preset time period (e.g., the first 3 seconds) at the beginning of the target video.

[0143] For example, you can set up copywriting templates and fill in the corresponding copywriting templates based on the emotional tags of the target song and the emotional keywords of the target topic to get the opening copywriting.

[0144] In some optional implementations, step S306, "generating the opening text of the target video based on the emotional tags of the target song and the emotional keywords of the target topic," may include steps c1 to c2.

[0145] Step c1 generates emotional tags for the target song, emotional keywords for the target topic, and prompts for creative method constraints; the creative method constraints include the constraints on copywriting.

[0146] Step c2: Input the prompt words into the large language model to obtain the opening text generated by the large language model; the prompt words are used to instruct the large language model to generate the opening text according to the emotional tags of the target song and the emotional keywords of the target topic, in accordance with the constraints of the creation method.

[0147] In this embodiment, copywriting is based on a large language model, and constraints on the writing method are also introduced.

[0148] Specifically, after determining the emotional tags of the target song and the emotional keywords of the target topic, we can further obtain the creative method constraints related to the opening copy. These creative method constraints are used to represent the constraints when generating the opening copy, and can be configured manually or summarized based on a large language model.

[0149] For example, creative method constraints may include the scene immersion method and the behavior explanation method, and one of the two can be chosen or used in combination.

[0150] For example, the scenario-based approach: using the sentence template "[Time / Scene] People who saw this song" to place the target audience in a specific context. The behavior-based approach: using the sentence template "Do you also [behavior description]?" to target the hidden behaviors or psychology of the target audience.

[0151] Optionally, in addition to including the emotional tags of the target song, the emotional keywords of the target topic, and creative method constraints, the prompt may also include copy length constraints, a list of prohibited words, etc.

[0152] For example, length constraint: output no more than 15 Chinese characters (including punctuation). Prohibited word list: advertising-style words (such as "absolutely", "hot-selling", "must-see"), marketing-style words (such as "click", "follow", "share"), overly sentimental words, etc.

[0153] By inputting the prompt word into the large language model, the powerful reasoning ability of the large language model can be used to generate a suitable opening message.

[0154] Optionally, since the initial output text of the large language model may be inappropriate due to factors such as illusions, this embodiment further introduces a verification mechanism. Specifically, step c2, "inputting the prompt words into the large language model and obtaining the opening text generated by the large language model," may include steps c21 to c24.

[0155] Step c21: Input the prompt words into the large language model and obtain the initial text output by the large language model.

[0156] Step c22: Verify the initial copy; the copy verification includes: verifying whether the emotional tone of the copy is consistent with the emotional tags of the target song, and verifying whether the copy conforms to the constraints of the creative method.

[0157] Step c23: If the initial copy passes the copy verification, then the initial copy will be used as the opening copy.

[0158] Step c24: If the initial copy fails the copy verification, update the prompt words according to the verification result, and regenerate the copy based on the updated prompt words until an opening copy that passes the copy verification is generated.

[0159] In this embodiment, one or more initial texts can be directly generated using a large language model. For example, if the model is limited to generating 3 texts each time, then the number of initial texts is 3, which can be used for subsequent verification.

[0160] For any initial copy, multiple validations can be performed. Specifically, it is based on whether the emotional tendency of the initial copy (e.g., emotional keywords) is consistent with the emotional tags of the song (judged by an emotional dictionary or model), which can verify emotional consistency. In addition, it is also necessary to verify whether the initial copy conforms to the constraints of the creative method, such as whether the initial copy can be classified as the scene substitution method or the behavior confession method.

[0161] Furthermore, it can also perform length checks (character count ≤ 15) and prohibited word checks (not including any words in the prohibited word list), etc.

[0162] For the text that passes the verification, it can be used as the final opening text; if none of the generated texts pass the verification, the prompt words are updated according to the verification results, such as adding the verification results to the prompt words, to regenerate the text based on the large language model, until the generated text passes the verification.

[0163] If multiple candidate texts (such as the initial text generated for the first time) pass the verification, the score of each candidate text can be determined, and the candidate text with the highest score will be used as the opening text.

[0164] For example, a score can be calculated based on the candidate copy's emotional consistency, technical suitability, length appropriateness, and language naturalness. For instance, the candidate copy's score could be: score = 0.4 × emotional consistency + 0.3 × technical suitability + 0.2 × length appropriateness + 0.1 × language naturalness.

[0165] Among them, emotional consistency refers to the similarity (0-1) between the emotional keywords extracted from the text and the emotional tags of the song; technique fit refers to whether it is clearly classified as scene imitation or behavior revelation (0 or 1), a reasonable mixture of the two = 0.8, and no classification = 0; length fit can be specifically: ≤12 characters = 1.0, 13-15 characters = 0.8, >15 characters are directly rejected (i.e., failed the validation); language naturalness refers to the evaluation index output by the large language model (such as the inverse of perplexity).

[0166] The number of retries for copy generation can be limited, for example, a maximum of two retries. If a satisfactory copy cannot be generated after multiple retries, a reminder message can be sent, or the target song can be changed to regenerate a suitable opening copy.

[0167] Step S307: Generate the target video based on the target song and opening text.

[0168] Please see details Figure 2 Step S205 of the illustrated embodiment will not be described again here.

[0169] In some optional implementations, the process of generating the target video in step S307 may specifically include step d1.

[0170] Step d1: Generate the target video based on the song information and opening text of the target song; the song information should include at least the emotional tag of the target song.

[0171] In this embodiment, after determining the background music and opening text corresponding to the target song, the song information of the target song (specifically including the target song's mood tag, and may also include BPM, artist, genre, lyrics, etc.) can be determined based on the song's metadata, thereby generating the target video. It can be understood that the background audio track of the target video includes the background music, the opening text is the opening text, and the song information of the target song can be used to assist in video generation, such as matching certain frames of the video with the lyrics of the song.

[0172] Optionally, step d1, "Generate the target video based on the song information and opening text of the target song," may specifically include steps d11 to d12.

[0173] Step d11: Determine the appropriate video type based on the source and category of the target topic; Step d12: Generate a target video of video type based on the song information of the target song, the target song, and the opening text.

[0174] In this embodiment, the appropriate video type can be determined based on the source of the target topic (which data platform it comes from) and its category (e.g., film, song, etc.). For example, for TV series / movies / variety show categories on video platforms, or drama series categories on Weibo platforms, the video type can be decided as popular film and television; for other cases (e.g., copywriting, song categories), the dynamic background category can be decided.

[0175] Input the target song information, the target song (which may be the background music corresponding to the target song), the opening text, and the corresponding video type into the downstream video generation module to generate the corresponding target video.

[0176] For example, for dynamic backgrounds, the downstream video generation module can call the dynamic background generation pipeline to produce a dynamic background with emotion matching, line-by-line lyric fading in, and an emotion template combination using a large model for text / image generation. For popular film and television content, the downstream video generation module can call the film and television clip editing pipeline to capture and match film and television clips or posters as backgrounds, with lyrics subtitles overlaid. Furthermore, both types of videos have capabilities such as lyric alignment, subtitle animation, and cover images to ensure the quality of the generated target video.

[0177] For popular film and television content, the focus is on copyright-compliant material libraries. Creation is completed through methods such as searching film and television material libraries, video editing, and overlaying lyrics and subtitles, which does not require complete reliance on large language models.

[0178] The data processing method provided in this embodiment, by matching the emotional keywords with the song's emotional tags, can select target songs with a high emotional match to the current public opinion environment. It establishes a quantifiable, traceable, and reproducible association between trending topics and songs, ensuring that the song selection results are consistent with the direction of public opinion on that day, and achieving non-random deterministic song selection driven by trending topics. Further generating opening text based on relevant information of the target song ensures the correlation between the opening text and the song's emotion. Introducing creative method constraints and completing the closed loop through post-verification ensures both creative diversity and compliance with the basic requirements of video platform dissemination rules. By determining the video type, it can generate target videos of the corresponding video type, thereby transforming content format selection from manual experience-based decision-making into a rule-based, automatically executable system behavior, achieving autonomous system decision-making.

[0179] This embodiment also provides a data processing apparatus for implementing the above embodiments and preferred embodiments; details already described will not be repeated. As used below, the term "module" can refer to a combination of software and / or hardware that performs a predetermined function. Although the apparatus described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0180] This embodiment provides a data processing device, such as... Figure 4 As shown, the device includes: The acquisition module 401 is used to acquire a topic set and acquire the emotion tags of multiple candidate songs; the topic set includes multiple original topics whose popularity meets preset conditions; The sentiment analysis module 402 is used to determine sentiment keywords for multiple selected topics; the selected topics are original topics selected from the topic set. The aggregation module 403 is used to aggregate the sentiment keywords of each selected topic to obtain a sentiment keyword pool; Processing module 404 is used to select a target song from multiple candidate songs based on the matching degree between the emotion tags of each candidate song and the emotion keyword pool; The generation module 405 is used to generate a target video based on the target song.

[0181] In some alternative implementations, determining the sentiment keywords for multiple selected topics includes: For any original topic in the topic set, determine the sentiment keywords corresponding to the original topic; Based on the sentiment keywords corresponding to each of the original topics, the original topics are filtered to determine the topics to be determined; Determine the evaluation indicators for each of the proposed topics; the evaluation indicators are positively correlated with the popularity of the proposed topics. Topics whose evaluation indicators are greater than the indicator threshold are selected as undetermined topics.

[0182] In some optional implementations, determining the sentiment keywords corresponding to the original topic includes: The original topic is matched with a preset sentiment dictionary, and the sentiment words in the sentiment dictionary that match the original topic are used as the sentiment keywords corresponding to the original topic. If no sentiment word matching the original topic exists in the sentiment dictionary, then sentiment analysis is performed on the original topic based on the large language model to determine the corresponding sentiment keywords.

[0183] In some optional implementations, the step of filtering each original topic based on the sentiment keywords corresponding to each original topic to determine the candidate topics includes: For multiple original topics whose title similarity is greater than the first threshold, retain the original topic with the highest popularity. Determine the second similarity between the sentiment keywords of the retained original topic and the historical sentiment word set; the historical sentiment word set is used to record the sentiment keywords of the used topics; Original topics with a second similarity score less than the second threshold are designated as undetermined topics.

[0184] In some optional implementations, determining the evaluation metrics for each of the topics to be determined includes: The normalized popularity, category weight, and time factor determined based on the time difference between the current time and the collection time of the topic to be determined are determined; the category weight is used to represent the importance of the topic category to which the topic to be determined belongs, and the time factor and the time difference are negatively correlated. The evaluation index for the undetermined topic is determined based on the normalized popularity, the weight of its category, and the time factor; the evaluation index is positively correlated with the normalized popularity, the category weight, and the time factor.

[0185] In some optional implementations, selecting the target song from the candidate songs based on the matching degree between the emotion tags of each candidate song and the emotion keyword pool includes: Based on the attribute information of each candidate song, determine the configuration parameters of the candidate songs; The target song is selected from the candidate songs based on the configuration parameters of the candidate songs and the matching degree.

[0186] In some optional implementations, the generation module 405 is further configured to: Identify target topics among the selected topics that match the emotional keywords with the emotional tags of the target song; The opening text of the target video is generated based on the emotional tags of the target song and the emotional keywords of the target topic.

[0187] In some optional implementations, generating the opening text of the target video based on the emotional tags of the target song and the emotional keywords of the target topic includes: Generate prompts containing the emotional tags of the target song, the emotional keywords of the target topic, and constraints on the creation method; the constraints on the creation method include the constraints on copywriting. The prompt words are input into the large language model to obtain the opening text generated by the large language model; the prompt words are used to instruct the large language model to generate the opening text according to the emotional tags of the target song and the emotional keywords of the target topic, in accordance with the constraints of the creation method.

[0188] In some optional implementations, the step of inputting the prompt word into a large language model and obtaining the opening text generated by the large language model includes: Input the prompt words into the large language model and obtain the initial text output by the large language model; The initial copy is validated; the validation includes: verifying whether the emotional tone of the copy is consistent with the emotional tags of the target song, and verifying whether the copy conforms to the constraints of the creative method. If the initial copy passes the copy verification, then the initial copy will be used as the opening copy; If the initial copy fails the copy verification, the prompt words are updated according to the verification result, and the copy is regenerated based on the updated prompt words until an opening copy that passes the copy verification is generated.

[0189] In some alternative implementations, the generation module 405 is used to: Based on the source and category of the target topic, determine the corresponding video type; Based on the song information of the target song, the target song, and the opening text, a target video of the video type is generated.

[0190] The data processing apparatus provided in this disclosure can execute the data processing method provided in any embodiment of this disclosure, and has the corresponding functional modules and beneficial effects for executing the method. Further functional descriptions of the various modules and units described above are the same as in the corresponding embodiments described above, and will not be repeated here.

[0191] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0192] The following is a detailed reference. Figure 5 The diagram illustrates a structural schematic suitable for implementing the electronic device described in the embodiments of this application. The electronic device may include a processor (e.g., a central processing unit, graphics processor, etc.) 501, which can perform various appropriate actions and processes according to a program stored in read-only memory (ROM) 502 or a program loaded from memory 508 into random access memory (RAM) 503. The RAM 503 also stores various programs and data required for the operation of the electronic device. The processor 501, ROM 502, and RAM 503 are interconnected via a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0193] Typically, the following devices can be connected to I / O interface 505: input devices 506 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 507 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; memory devices 508 including, for example, magnetic tapes, hard disks, etc.; and communication devices 509. Communication device 509 allows electronic devices to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 5 Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown, and more or fewer devices may be implemented or have instead.

[0194] Specifically, according to embodiments of this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this application include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 509, or installed from a memory 508, or installed from a ROM 502. When the computer program is executed by the processor 501, it performs the functions defined in the data processing method of the embodiments of this application.

[0195] Figure 5 The electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0196] This application also provides a computer-readable storage medium. The methods described in this application can be implemented in hardware or firmware, or implemented as recordable on a storage medium, or implemented as computer code downloaded over a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the data processing methods shown in the above embodiments are implemented.

[0197] A portion of this application can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide the methods and / or technical solutions according to this application through the operation of the computer. Those skilled in the art will understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executing the instructions, or the computer compiling the instructions and then executing the corresponding compiled program, or the computer reading and executing the instructions, or the computer reading and installing the instructions and then executing the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to a computer.

[0198] Although embodiments of this application have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of this application, and all such modifications and variations fall within the scope defined by the appended claims.

Claims

1. A data processing method, characterized in that, The method includes: Obtain a set of topics and obtain the emotional tags of multiple candidate songs; the set of topics includes multiple original topics whose popularity meets preset conditions; Identify sentiment keywords for multiple selected topics; the selected topics are original topics chosen from the topic set. The sentiment keywords of each selected topic are aggregated to obtain a sentiment keyword pool; The target song is selected from the candidate songs based on the matching degree between the emotional tags of each candidate song and the emotional keyword pool. Generate a target video based on the target song.

2. The method according to claim 1, characterized in that, The determination of sentiment keywords for multiple selected topics includes: For any original topic in the topic set, determine the sentiment keywords corresponding to the original topic; Based on the sentiment keywords corresponding to each of the original topics, the original topics are filtered to determine the topics to be determined; Determine the evaluation indicators for each of the proposed topics; the evaluation indicators are positively correlated with the popularity of the proposed topics. Topics whose evaluation indicators are greater than the indicator threshold are selected as undetermined topics.

3. The method according to claim 2, characterized in that, The process of determining the sentiment keywords corresponding to the original topic includes: The original topic is matched with a preset sentiment dictionary, and the sentiment words in the sentiment dictionary that match the original topic are used as the sentiment keywords corresponding to the original topic. If no sentiment word matching the original topic exists in the sentiment dictionary, then sentiment analysis is performed on the original topic based on the large language model to determine the corresponding sentiment keywords.

4. The method according to claim 2, characterized in that, The step of filtering the original topics based on the sentiment keywords corresponding to each original topic to determine the topics to be determined includes: For multiple original topics whose first similarity between topic titles is greater than the first threshold, retain the original topic with the highest popularity. Determine the second similarity between the sentiment keywords of the retained original topic and the historical sentiment word set; the historical sentiment word set is used to record the sentiment keywords of the used topics; Original topics with a second similarity score less than the second threshold are designated as undetermined topics.

5. The method according to claim 2, characterized in that, The determination of evaluation indicators for each of the undetermined topics includes: The normalized popularity, category weight, and time factor determined based on the time difference between the current time and the collection time of the topic to be determined are determined; the category weight is used to represent the importance of the topic category to which the topic to be determined belongs, and the time factor and the time difference are negatively correlated. The evaluation index for the undetermined topic is determined based on the normalized popularity, the weight of its category, and the time factor; the evaluation index is positively correlated with the normalized popularity, the category weight, and the time factor.

6. The method according to claim 1, characterized in that, The step of selecting a target song from multiple candidate songs based on the matching degree between the emotion tags of each candidate song and the emotion keyword pool includes: Based on the attribute information of each candidate song, determine the configuration parameters of the candidate songs; The target song is selected from the candidate songs based on the configuration parameters of the candidate songs and the matching degree.

7. The method according to claim 1, characterized in that, The method further includes: Identify target topics among the selected topics that match the emotional keywords with the emotional tags of the target song; The opening text of the target video is generated based on the emotional tags of the target song and the emotional keywords of the target topic.

8. An electronic device, characterized in that, include: A memory and a processor are communicatively connected, the memory stores computer instructions, and the processor executes the computer instructions to perform the data processing method of any one of claims 1 to 7.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to perform the data processing method according to any one of claims 1 to 7.

10. A computer program product, characterized in that, Includes computer instructions for causing a computer to perform the data processing method according to any one of claims 1 to 7.

Images