Explosive content prediction method and device, electronic equipment and storage medium

Abstract

The invention relates to an implosion content prediction method and device, electronic equipment and a storage medium. The method comprises the steps that firstly, hot topic data provided by different social platforms are obtained, then semantic analysis and feature mapping are conducted on the hot topic data based on a pre-constructed hot topic feature database, hot topic associated information is generated, and the hot topic associated information at least comprises candidate hot topics; according to different types of hot topic associated information, heat scores corresponding to the candidate hot topics are calculated, the multiple heat scores are subjected to weighted fusion to obtain a target hot topic index, and finally, the target hot topics are screened from all the candidate hot topics according to the target hot topic index. And predicting brand popular content associated with the target hot topic content. According to the scheme provided by the invention, the target hot topics can be accurately and efficiently identified according to the target hot topics, and then the brand popular content associated with the target hot topics can be accurately predicted.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to methods, apparatus, electronic devices and storage media for predicting trending content. Background Technology

[0002] With advancements in artificial intelligence and big data technologies, businesses can identify potential trending topics, gain insights into user behavior trends, and formulate precise content operation and marketing strategies accordingly. Trending topics refer to topics that circulate frequently within a certain timeframe, reflecting the public's interests on social media platforms.

[0003] Currently, relevant trending topic identification methods typically rely on explicit metrics such as likes, shares, and comments to rank topics by popularity and thus determine their level of popularity. For example, by statistically analyzing the number of likes, shares, comments, and saves for a particular topic on social media platforms, these metrics are summed to calculate the total interaction volume for each topic within a specific time period, and the trending topics are directly identified based on the total interaction volume.

[0004] In practical applications, while the above methods can quickly filter out high-profile topics to a certain extent and provide reference data for enterprises, they mainly rely on explicit interaction data and easily overlook the potential spread of topics. At the same time, they cannot guarantee that topics with high interaction volume are consistent with the enterprise's products, which makes it difficult for enterprises to identify hot topics that are consistent with their own brands in a timely manner, make it difficult to effectively predict viral content, and miss the best content marketing opportunities. Summary of the Invention

[0005] To address or partially address the problems existing in related technologies, this application provides a method, apparatus, electronic device, and storage medium for predicting trending content, which can improve the accuracy of predicting trending content.

[0006] The first aspect of this application provides a method for predicting viral content, including: Obtain trending topic data from different social media platforms; Based on a pre-built hot topic feature database, semantic analysis and feature mapping are performed on the hot topic data to generate hot topic association information; the hot topic association information includes at least candidate hot topics. For different types of trending topic association information, the popularity score corresponding to the candidate trending topic is calculated, and multiple popularity scores are weighted and fused to obtain the target trending index; Based on the target hot topic index, target hot topics are selected from various candidate hot topics, and brand hit content associated with the target hot topic content is predicted.

[0007] In one instance, the step of performing semantic analysis and feature mapping on the hot topic data based on a pre-built hot topic feature database to generate hot topic association information includes: Extract industry attributes and historical brand characteristics from a pre-built database of hotspot features; Semantic analysis is performed on the hot topic data to obtain the candidate hot topics and candidate hot keywords of the candidate hot topics; If the candidate hot keywords belong to the industry attribute, then the type of the candidate hot topic is determined to be industry-related. If the candidate hot keywords belong to the historical brand characteristics, then the type of the candidate hot topic is determined to be brand association type; The candidate hot topics, their types, and their keywords are structured to generate associated information about the hot topics.

[0008] In one instance, performing semantic analysis on the trending topic data to obtain the candidate trending topics and candidate trending keywords for the candidate trending topics includes: Semantic analysis was performed on the aforementioned trending topic data to obtain multiple topics; Calculate the popularity growth rate of each topic in the first set time window, and calculate the popularity trend score of each topic in the second set time window; Based on the popularity growth rate and the popularity trend score, each topic is scored to obtain a basic popularity score. Topics with a basic popularity score greater than or equal to a preset popularity threshold are identified as candidate hot topics, and candidate hot keywords corresponding to the candidate hot topics are extracted from the hot topic data.

[0009] In one instance, the popularity score includes an industry matching score, and the calculation of the popularity score corresponding to the candidate hot topic for different types of hot topic association information includes: When the type of information associated with the hot topic is the industry-related type, extract the industry keywords and the first sentiment keyword from the candidate hot keywords; Calculate the first text similarity between the industry keywords and the industry attributes; According to the preset score mapping relationship, the first emotion score corresponding to the first emotion keyword is obtained; The industry matching score is obtained by weighted fusion of the first text similarity and the first sentiment score.

[0010] In one instance, the popularity score includes a brand matching score, and the calculation of the popularity score corresponding to the candidate hot topic for different types of hot topic association information includes: When the type of information associated with the hot topic is the brand association type, extract the brand keyword and the second emotion keyword from the candidate hot topic keywords; Calculate the second text similarity between the brand keywords and the historical brand features; According to the preset score mapping relationship, the second emotion score corresponding to the second emotion keyword is obtained; The brand matching score is obtained by weighted fusion of the second text similarity and the second emotion score.

[0011] In one instance, the weighted fusion of multiple popularity scores to obtain the target popularity index includes: For each popularity score, assign a popularity weight; The multiple popularity scores are weighted and fused according to the popularity weight to generate the target hot topic index, which is data representing the overall popularity of the candidate hot topics.

[0012] In one instance, the step of filtering target hot topics from various candidate hot topics based on the target hot topic index and predicting brand-related viral content associated with the target hot topic content includes: The candidate hot topics are sorted according to the value of the target hot topic index to generate a list of hot topics; The top N candidate hot topics in the hot topic list are determined as the target hot topic, where N is a positive integer; The target trending topic is the input data of a pre-trained trending content generation model, which enables the trending content generation model to automatically generate brand-related viral content associated with the target trending topic.

[0013] A second aspect of this application provides a device for predicting viral content, comprising: The multi-source data acquisition module is used to acquire trending topic data from different social media platforms; The association information generation module is used to perform semantic analysis and feature mapping on the hot topic data based on a pre-built hot topic feature database to generate hot topic association information; the hot topic association information includes at least candidate hot topics; The hot topic index generation module is used to calculate the heat score corresponding to the candidate hot topic for different types of hot topic association information, and to perform weighted fusion of multiple heat scores to obtain the target hot topic index. The content prediction module is used to filter target hot topics from various candidate hot topics based on the target hot topic index, and predict brand hit content associated with the target hot topic content.

[0014] A third aspect of this application provides an electronic device, comprising: Processor; and A memory that stores executable code, which, when executed by the processor, causes the processor to perform the method described above.

[0015] A fourth aspect of this application provides a computer-readable storage medium having executable code stored thereon, which, when executed by a processor of an electronic device, causes the processor to perform the method described above.

[0016] The fifth aspect of this application provides a computer program product comprising computer instructions that, when executed by a processor, implement the method described above.

[0017] The technical solution provided in this application may include the following beneficial results: In this application, firstly, trending topic data from different social media platforms is acquired. Then, based on a pre-built trending feature database, semantic analysis and feature mapping are performed on the trending topic data to generate trending topic association information. The trending topic association information includes at least candidate trending topics. For different types of trending topic association information, the popularity score corresponding to the candidate trending topics is calculated, and multiple popularity scores are weighted and fused to obtain a target trending index. Finally, target trending topics are selected from each candidate trending topic based on the target trending index, and brand hit content associated with the target trending topic content is predicted.

[0018] Compared with related technologies, the technical solution of this application has the following advantages: First, by acquiring hot topic data from different social media platforms, cross-platform statistical analysis of hot topic data is achieved to comprehensively understand the overall trend of hot topics. Furthermore, through in-depth semantic analysis and feature mapping of hot topic data, topic-related information that is highly consistent with the company's products can be generated. Second, this application specifically calculates the popularity scores corresponding to hot topic-related information under different types, and then weights and merges the popularity scores of different dimensions to obtain the target hot topic index. The target hot topic index obtained in this way not only takes into account the popularity level of the topic in different dimensions, but also comprehensively reflects the dissemination potential of the topic. Thus, based on the target hot topic index, the target hot topic can be accurately and efficiently identified, and the brand's hit content associated with the target hot topic can be predicted more accurately.

[0019] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0020] The above and other objects, features and advantages of this application will become more apparent from the more detailed description of exemplary embodiments thereof in conjunction with the accompanying drawings, wherein the same reference numerals generally represent the same components in the exemplary embodiments thereof.

[0021] Figure 1 This is a flowchart illustrating a method for predicting viral content, as shown in an embodiment of this application. Figure 2 This is another flowchart illustrating a method for predicting viral content, as shown in an embodiment of this application; Figure 3 This is a schematic diagram of the structure of a popular content prediction device shown in an embodiment of this application; Figure 4 This is a schematic diagram of the structure of an electronic device shown in an embodiment of this application. Detailed Implementation

[0022] Embodiments of this application will now be described in more detail with reference to the accompanying drawings. While embodiments of this application are shown in the drawings, it should be understood that this application may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to make this application more thorough and complete, and to fully convey the scope of this application to those skilled in the art.

[0023] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used in this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0024] It should be understood that although the terms "first," "second," "third," etc., may be used in this application to describe various information, this information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Thus, 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.

[0025] With advancements in artificial intelligence and big data technologies, businesses can identify potential trending topics, gain insights into user behavior trends, and formulate precise content operation and marketing strategies accordingly. Trending topics refer to topics that circulate frequently within a certain timeframe, reflecting the public's interests on social media platforms.

[0026] Currently, relevant trending topic identification methods typically rely on explicit metrics such as likes, shares, and comments to rank topics by popularity and thus determine their level of popularity. For example, by counting the number of likes, shares, comments, and saves for a particular topic on social media platforms, these metrics are summed to calculate the total interaction volume for each topic within a specific time period, and the trending topic is determined directly based on the total interaction volume.

[0027] In practical applications, while the above methods can quickly filter out high-profile topics to a certain extent and provide reference data for enterprises, they mainly rely on explicit interaction data and easily overlook the potential spread of topics. At the same time, they cannot guarantee that topics with high interaction volume are consistent with the enterprise's products, which makes it difficult for enterprises to identify hot topics that are consistent with their own brands in a timely manner, make it difficult to effectively predict viral content, and miss the best content marketing opportunities.

[0028] Among the related technologies, there are problems such as enterprises being unable to identify trending topics that align with their own brands in a timely manner, having difficulty effectively predicting viral content, and missing the best content marketing opportunities.

[0029] To address the aforementioned issues, this application provides a method for predicting trending content, which can improve the accuracy of trending content prediction.

[0030] The technical solutions of the embodiments of this application are described in detail below with reference to the accompanying drawings.

[0031] Figure 1 This is a flowchart illustrating a method for predicting viral content, as shown in an embodiment of this application.

[0032] See Figure 1 The method includes at least the following steps: Step 101: Obtain trending topic data from different social media platforms.

[0033] In this application embodiment, trending topic data provided by different social platforms can be acquired in real time or at regular intervals, thereby covering the user group of the entire platform. It can also capture the differences in the spread of the same topic on different platforms, which is conducive to fully exploring the spread potential of the topic.

[0034] Optionally, "social platform" refers to online social media platforms commonly used by users. Because the display styles and data structures of trending topics differ across social platforms, the collected trending topic data includes, but is not limited to, lists of trending topics and popular discussion posts.

[0035] As an example, this application can perform data preprocessing on trending topic data provided by different social media platforms, such as deduplication, filtering invalid data, and unifying data formats, to obtain valid data with consistent formats.

[0036] Step 102: Based on the pre-built hot topic feature database, perform semantic analysis and feature mapping on the hot topic data to generate hot topic association information, which includes at least candidate hot topics.

[0037] In this embodiment of the application, the hot topic data collected in step 101 can be semantically analyzed and feature mapped based on a pre-built hot topic feature database to generate hot topic association information, and the hot topic association information includes at least candidate hot topics.

[0038] Optionally, a hot topic feature database refers to a pre-built database containing industry attributes, audience profiles, historical brand characteristics, historical hot topic information, sentiment keywords, and other hot topic-related content. In practical applications, a single database can be used for centralized storage, with all industry attributes, brand characteristics, historical hot topic information, etc., stored in the same database table for easy unified management and retrieval. Alternatively, multiple databases can be used, such as storing industry-related information in an industry database and brand-related information in a brand database, with each database managed independently.

[0039] Semantic analysis refers to the use of Natural Language Processing (NLP) technology to understand and process text data related to trending topics, including keyword extraction, sentiment analysis, and classification, in order to identify candidate trending topics and their corresponding keywords.

[0040] Feature mapping refers to matching candidate hot topics and their candidate hot keywords obtained through semantic analysis with industry attributes, historical brand characteristics, or sentiment keywords in a hot feature database to generate related information.

[0041] Hot topic association information refers to the association information generated after the above semantic analysis and feature mapping, which includes at least candidate hot topics, candidate hot keywords, and topic types (such as industry association types and brand association types).

[0042] As an example, suppose the trending topics data collected from social media platforms A and B include the topic "New skincare product attracts user attention; the product's ingredients are gentle." First, semantic analysis is used to extract candidate trending keywords from the topic content, such as "skincare," "new product," and "user attention." Next, these candidate keywords are mapped to a pre-built trending feature database. The candidate keyword "skincare" is matched with the industry attribute "skin," and the candidate keyword "gentle" is matched with the brand. Finally, the matching results are used as one type of data in the trending topic association information.

[0043] Step 103: For different types of trending topic association information, calculate the popularity score corresponding to the candidate trending topic, and perform weighted fusion of multiple popularity scores to obtain the target trending index.

[0044] In this embodiment of the application, the popularity score corresponding to the candidate hot topic can be calculated for different types of hot topic association information, and then the multiple popularity scores can be weighted and fused to obtain the target hot topic index.

[0045] Optionally, the popularity score represents the popularity level of a candidate hot topic in a specific dimension. In this application, the popularity level of candidate hot topics is quantified in the form of a score, so as to conduct a comprehensive analysis of candidate hot topics and select suitable target hot topics.

[0046] The weighted fusion process includes linear weighting, nonlinear weighting, and machine learning model weighting. Weighted fusion can more accurately reflect the comprehensive value of candidate hot topics.

[0047] The target hot topic index refers to the comprehensive score of candidate hot topics, reflecting multiple indicators such as the attention received by candidate hot topics across the entire platform, industry relevance, brand fit, and sentiment.

[0048] As an example, assuming the target hot topic index = popularity score multiplied by weight parameters, if the popularity scores corresponding to the candidate hot topics are calculated as y1, y2, and y3, then this application can perform weighted fusion of the popularity scores of different dimensions based on the preset weight parameters w1, w2, and w3, that is, target hot topic index = y1w1 + y2w2 + y3w3.

[0049] By flexibly adjusting the proportion of popularity scores under different dimensions in the overall assessment, the rationality of the target popularity index can be improved, and the assessment results can be avoided due to the overemphasis of a single dimension.

[0050] Step 104: Select target hot topics from various candidate hot topics based on the target hot topic index, and predict brand hit content related to the target hot topic content.

[0051] In this embodiment of the application, based on the target hot topic index obtained in step 103, target hot topics with dissemination potential are selected from the candidate hot topic set.

[0052] Optionally, the target hot topic refers to a topic selected from multiple candidate hot topics that is highly relevant to specific industry attributes and brand characteristics and has strong dissemination potential. For example, candidate hot topics include: {Topic ①: "New skincare product ingredients are safe, gentle and non-irritating"; Topic ②: "A celebrity's lipstick becomes a hot topic among fans"; Topic ③: "Sports event championship sparks nationwide discussion"}. After processing in step 103, it is found that Topic ① is highly relevant to the industry attribute "beauty and skincare", and its candidate keywords "new product" and "gentle" highly match the keywords in the hot topic data feature library. At the same time, it has maintained an upward trend in popularity in the past 7 days, and its target hot topic index meets the preset conditions. Therefore, Topic ① can be identified as the target hot topic. Although Topic ② has a certain level of popularity, it does not fully match the brand's goals, and its target hot topic index does not meet the preset conditions. Although Topic ③ has a high volume of discussion, it is not directly related to industry attributes and the brand, and therefore does not meet the preset conditions.

[0053] In practical applications, when a target trending topic is identified, brand-related viral content related to the company's products or services can also be generated based on that topic.

[0054] Brand-driven viral content refers to content created by combining target trending topics with brand characteristics. For example, if the target trending topic is "gentle ingredients in new skincare products," it can be combined with features such as "gentle formula" and "suitable for sensitive skin" from a trending topic feature database to automatically generate brand marketing copy suitable for dissemination on social media platforms. This helps businesses seize opportunities to spread trending topics in a timely manner and achieve precise marketing.

[0055] In this embodiment, firstly, trending topic data from different social media platforms is acquired. Then, based on a pre-built trending feature database, semantic analysis and feature mapping are performed on the trending topic data to generate trending topic association information. The trending topic association information includes at least candidate trending topics. For different types of trending topic association information, the popularity score corresponding to the candidate trending topics is calculated, and multiple popularity scores are weighted and fused to obtain the target trending index. Finally, the target trending topic is selected from each candidate trending topic based on the target trending index, and the brand's hit content associated with the target trending topic is predicted.

[0056] Compared with related technologies, the technical solution of this application has the following advantages: First, by acquiring hot topic data from different social media platforms, cross-platform statistical analysis of hot topic data is achieved to comprehensively understand the overall trend of hot topics. Furthermore, through in-depth semantic analysis and feature mapping of hot topic data, topic-related information that is highly consistent with the company's products can be generated. Second, this application specifically calculates the popularity scores corresponding to hot topic-related information under different types, and then weights and merges the popularity scores of different dimensions to obtain the target hot topic index. The target hot topic index obtained in this way not only takes into account the popularity level of the topic in different dimensions, but also comprehensively reflects the dissemination potential of the topic. Thus, based on the target hot topic index, the target hot topic can be accurately and efficiently identified, and the brand's hit content associated with the target hot topic can be predicted more accurately.

[0057] Figure 2 This is another flowchart illustrating a method for predicting trending content, as shown in an embodiment of this application. Figure 2 relatively Figure 1 The technical solutions of the embodiments of this application are described in more detail.

[0058] Step 201: Obtain trending topic data from different social media platforms.

[0059] In this embodiment, the data extraction, analysis, and understanding processes utilize the AIUC (Artificial Intelligence Content Understanding) engine. This AIUC engine integrates relevant knowledge graphs to perform deep understanding and real-time analysis of multimodal data such as text, images, and videos. It can quickly identify which topics are growing rapidly and which are likely to become trends in the short term, ensuring that the selected target hot topics are forward-looking. For example, by using the AIUC engine to deeply analyze user behavior and content they follow, more targeted and real-time identification of hot topic data can be achieved, thereby improving the accuracy of target hot topic identification. Simultaneously, it helps brands accurately grasp genuine user feedback, improving marketing efficiency and effectiveness.

[0060] In practical applications, by pre-configuring APIs (Application Programming Interfaces) adapted to the AIUC engine and using crawler technology that complies with the Robots Exclusion Protocol, relevant data such as trending topics, search keywords, and basic popularity data are collected from different social platforms. After data preprocessing, trending topic data is generated.

[0061] The basic popularity data covers both explicit and implicit data regarding the level of attention and dissemination behavior of a topic. For example, explicit data includes the number of likes, shares, comments, favorites, views, impressions, clicks, posting timestamps, and platform trending rankings, while implicit data includes sentiment tags, region, and content type.

[0062] Step 202: Based on the pre-built hot topic feature database, perform semantic analysis and feature mapping on the hot topic data to generate hot topic association information, which includes at least candidate hot topics.

[0063] In this embodiment, the main process involves extracting industry attributes and historical brand features from a pre-built hot topic feature database, performing semantic analysis on the hot topic data to obtain candidate hot topics and candidate hot keywords. If the candidate hot keywords belong to industry attributes, the type of the candidate hot topic is determined to be industry-related. If the candidate hot keywords belong to historical brand features, the type of the candidate hot topic is determined to be brand-related. The candidate hot topics, their types, and the candidate hot keywords are then structured to generate hot topic association information.

[0064] The construction process of the hot topic feature database can be as follows: Collect industry-related information and historical brand data, mainly including industry attributes, historical brand characteristics, audience profiles, and historical topic information. Preprocess and clean the collected data to obtain text data related to industry attributes and historical brand characteristics. Extract keywords from the text data and label the category of each keyword. Further, label and classify historical hot topics, associating them with industry attributes and historical brand characteristics to generate a brand knowledge graph for a specific brand. Each data entry in the brand knowledge graph records the corresponding keywords, association type, and related historical topic information. Store the integrated information in MySQL (My Structured Query Language, a relational database) or MongoDB (MongoDatabase, a non-relational database) to obtain the hot topic feature database.

[0065] It is worth noting that the hot topic feature database can be updated regularly, so that newly added industry attributes, brand characteristics and hot topics can be stored in the hot topic feature database in a timely manner.

[0066] In addition, structured processing refers to storing candidate trending topics, their types, and keywords according to a prescribed data format. For example, storing relevant data in the order of {candidate trending topic name; candidate trending topic type; candidate trending topic keywords} generates trending topic association information.

[0067] Regarding the selection method for candidate trending topics, this application mainly utilizes NLP technology to perform semantic analysis on trending topic data, deeply understand the topic content and user sentiment of trending topic data across the entire network, and thus select candidate trending topics that conform to industry attributes and brand characteristics from massive amounts of trending topic data.

[0068] The process of using NLP to perform semantic analysis on hot topic data to obtain candidate hot topics and candidate hot keywords for candidate hot topics includes at least the following: performing semantic analysis on hot topic data to obtain multiple topics, calculating the popularity growth rate of each topic in a first set time window, and calculating the popularity trend score of each topic in a second set time window, scoring each topic based on the popularity growth rate and popularity trend score to obtain a basic popularity score, identifying topics with a basic popularity score greater than or equal to a preset popularity threshold as candidate hot topics, and extracting candidate hot keywords corresponding to the candidate hot topics from the hot topic data.

[0069] In this application, the first set time window is shorter than the second set time window, such as the first set time window being 24 hours and the second set time window being 7 days.

[0070] The heat growth rate calculated in the first set time window refers to: The interaction volume is the sum of the number of likes, comments, shares, and favorites within the first set time window.

[0071] The popularity trend score calculated in the second set time window refers to the change in the topic's popularity within the second set time window, calculated using methods such as regression fitting, averaging, and normalization. For example, it represents the average growth rate of daily interactions for a certain topic over 7 days.

[0072] After identifying multiple strongly related topics, the popularity growth rate of each topic within a first defined time window is calculated to measure the short-term explosive power of the topic's interaction volume. A popularity trend score within a second defined time window is used to obtain the topic's popularity trend over a longer period, such as continuously rising popularity, maintaining high popularity, or continuously declining popularity, thus measuring the topic's long-term trend. By combining short-term explosive power and long-term popularity trends, a preliminary basic popularity score for the topic can be obtained.

[0073] As an example, this application divides the popularity growth rate into different score ranges. For example, a popularity growth rate of less than or equal to 50% corresponds to score a, a popularity growth rate of greater than 50% and less than 200% corresponds to score b, and a popularity growth rate of greater than 200% corresponds to score c. Correspondingly, different ranges are divided for the popularity trend score, and a corresponding score is assigned to each range. Finally, the basic popularity score for each topic is obtained by combining the score corresponding to the popularity growth rate with the score corresponding to the popularity trend score.

[0074] As another example, this application can normalize the popularity growth rate and popularity trend score, directly convert the normalized value into the corresponding score, and finally obtain the basic popularity score of each topic by combining the score corresponding to the popularity growth rate and the score corresponding to the popularity trend score.

[0075] After the above calculations, the basic popularity score of each topic can be obtained. Topics with a basic popularity score greater than or equal to the preset popularity threshold are identified as candidate hot topics. This can avoid misidentifying low-popularity topics or noisy topics caused by short-term fluctuations, thereby ensuring that the selected candidate hot topics have sustained attention and dissemination potential for a certain period of time.

[0076] Step 203: When the type of information associated with a hot topic is industry-related, calculate the industry matching score corresponding to the candidate hot topic.

[0077] In this embodiment, the popularity score includes not only the basic popularity score but also an industry matching score. The industry matching score describes the degree of relevance or matching between the candidate hot topic and industry attributes. A higher industry matching score indicates a higher degree of matching between the candidate hot topic and industry attributes, while a lower industry matching score indicates a lower degree of matching between the candidate hot topic and industry attributes.

[0078] When the type of information associated with a hot topic is industry-related, industry keywords and first sentiment keywords are extracted from candidate hot keywords. The first text similarity between industry keywords and industry attributes is calculated. According to the preset score mapping relationship, the first sentiment score corresponding to the first sentiment keyword is obtained. The first text similarity and the first sentiment score are weighted and fused to obtain the industry matching score.

[0079] In this application, the calculation process for the first text similarity in the industry matching score is mainly as follows: semantically similar industry keywords are extracted from candidate hot keywords using TF-IDF (Term Frequency–Inverse Document Frequency) and Word2Vec (Word to Vector model), and then the cosine similarity between the industry keywords and industry attributes is calculated. The result of the cosine similarity is used as the first text similarity to measure the semantic closeness of the two keywords.

[0080] Industry attributes refer to a set of tags or keywords that describe the industry to which an object (such as a topic, product, content, or user behavior) belongs. For example, the industry attributes of the beauty industry include: beauty, skincare, makeup, face mask, refreshing, night cream, etc.

[0081] As an example, candidate hot keywords can be used as input corpus. The TF-IDF value of each keyword in the text can be calculated. The keywords can be sorted according to the size of the TF-IDF value. The top-ranked high-weight words can be selected as industry keywords for the topic. These industry keywords can be compared with the industry attributes in the hot feature database to determine whether they are similar.

[0082] As another example, the Word2Vec model is used to vectorize the words in the candidate hot keywords to obtain the word vectors of industry keywords. Similarly, these industry keywords are compared with the industry attributes in the hot feature database to determine whether they are similar.

[0083] In addition, this application extracts users' emotional keywords, which refer to words that can reflect users' emotions or attitudes from candidate hot keywords. For example, positive emotional words such as "like", "satisfied", and "easy to use", and negative emotional words such as "disappointed" and "dissatisfied".

[0084] To fully consider users' emotional inclinations toward candidate trending topics, this application introduces a score mapping mechanism corresponding to emotional keywords—a preset score mapping relationship.

[0085] The preset score mapping relationship refers to associating each emotional keyword with a preset emotional score to generate a one-to-one correspondence.

[0086] For example, positive emotion words can be assigned positive scores, and different scores can be allocated to different positive emotion words according to actual needs to express the strength of positive emotions, such as assigning 1 point to "good" and 2 points to "very good," etc. This application does not limit this. Similarly, negative emotion words can be assigned negative scores, and different scores can be allocated to different negative emotion words according to actual needs to express the strength of negative emotions, such as assigning 1 point to "very average" and 2 points to "disappointed," etc. As for neutral emotion words, since their positive or negative impact on the spread trend is not obvious, they can be directly assigned 0 points, which preserves their meaning in semantic analysis and avoids their bias on the final calculation results.

[0087] In practical applications, a candidate trending topic often contains multiple emotional keywords. This application can sum the scores of all identified emotional keywords to obtain the emotional score of the candidate trending topic. In the popularity assessment, in addition to considering traditional explicit data such as the number of likes, reposts, comments, and collections, it can also gain a deeper understanding of users' emotional tendencies, thereby improving the accuracy of predicting trending content.

[0088] Step 204: When the type of information associated with a hot topic is brand association, calculate the brand matching score corresponding to the candidate hot topic.

[0089] In this embodiment of the application, the popularity score also includes a brand matching score. The brand matching score can describe the degree of relevance or matching between the candidate hot topic and the historical brand. The higher the brand matching score, the higher the degree of matching between the candidate hot topic and the historical brand characteristics of the current brand. The lower the brand matching score, the lower the degree of matching between the candidate hot topic and the historical brand characteristics of the current brand.

[0090] When the type of information associated with a trending topic is brand-related, brand keywords and second emotion keywords are extracted from candidate trending keywords. The second text similarity between the brand keywords and historical brand features is calculated. According to the preset score mapping relationship, the second emotion score corresponding to the second emotion keyword is obtained. The second text similarity and the second emotion score are weighted and fused to obtain the brand matching score.

[0091] In this application, the calculation process for the second text similarity in the brand matching score is mainly as follows: semantically similar brand keywords are extracted from candidate hot keywords using TF-IDF and Word2Vec, and then the cosine similarity between the brand keywords and historical brand features is calculated. The result of the cosine similarity is used as the second text similarity to measure the semantic closeness of the two keywords. The principle of obtaining brand keywords using TF-IDF and Word2Vec is similar to the principle of obtaining industry keywords using TF-IDF and Word2Vec, and will not be elaborated further here.

[0092] Historical brand characteristics refer to the semantic information, keywords, typical topics, brand image tags, and other related features associated with a brand. For example, for a cosmetics brand, its historical brand characteristics may include keywords such as "skincare," "whitening," "gentle," "suitable for sensitive skin," and "celebrity endorsement," as well as keywords related to the brand's past marketing activities.

[0093] By comprehensively considering the above three aspects—basic popularity score, industry matching score, and brand matching score—this application can quickly filter out candidate hot topics that truly match the industry positioning and brand needs from massive amounts of hot topic data, and quantify the positive and negative impact of users' emotional tendencies on candidate hot topics.

[0094] Step 205: Weight and merge multiple popularity scores to obtain the target popularity index.

[0095] In this embodiment of the application, a heat weight is assigned to each heat score, and multiple heat scores are weighted and merged according to the heat weight to generate a target heat index.

[0096] As described in steps 202, 203, and 204 above, the popularity score of this application includes a basic popularity score, an industry matching score, and a brand matching score. The basic popularity score reflects the basic dissemination and user interaction of the topic on social media platforms; the industry matching score reflects the degree of relevance of the topic to the current industry; and the brand matching score reflects the degree of relevance of the topic to the current brand. Since the different dimensions of the score have varying degrees of impact on the topic's popularity, corresponding popularity weights need to be assigned to the basic popularity score, industry matching score, and brand matching score respectively. The corresponding popularity weights for the basic popularity score, industry matching score, and brand matching score can be the same or different; a higher weight coefficient indicates a more important popularity score.

[0097] Assuming that the base popularity score is assigned a first popularity weight, the industry matching score is assigned a second popularity weight, and the brand matching score is assigned a third popularity weight, then the target hot topic index = (base popularity score × first popularity weight) + (industry matching score × second popularity weight) + (brand matching score × third popularity weight) can be obtained, which can comprehensively reflect the overall performance of candidate hot topics under different dimensions.

[0098] In scenarios where industry relevance is crucial, the AIUC engine can appropriately increase the weight of the second popularity metric in the industry matching score to enhance industry relevance and more accurately capture potential trending topics within the industry. For example, in some industries, the relevance of trending topics is often closely related to industry trends. The AIUC engine can combine industry knowledge graphs and user behavior data for analysis to deeply mine industry-related keywords, points of interest, and popular content, helping brands identify trending topics closely related to their own business in a complex market environment.

[0099] In scenarios emphasizing brand characteristics, the AIUC engine can increase the weight of the third trending factor in brand relevance scores, enabling intelligent filtering based on brand characteristics to ensure that the selected trending topics align with the brand. For example, if a brand's communication focuses on differentiated marketing, the AIUC engine can utilize the brand's knowledge graph for analysis and calculation to assess its relevance, ensuring that the selected trending topics better match the brand's communication direction and market positioning.

[0100] The target hot topic index obtained through the above methods not only takes into account the popularity of the topic itself, but also integrates industry relevance and brand fit.

[0101] Step 206: Select target hot topics from various candidate hot topics based on the target hot topic index, and predict brand hit content related to the target hot topic content.

[0102] In this embodiment of the application, each candidate hot topic is sorted from high to low according to the value of the target hot topic index to generate a hot topic list. The top N candidate hot topics in the hot topic list are determined as the target hot topics, where N is a positive integer.

[0103] The trending topics list refers to a list of candidate trending topics sorted from highest to lowest according to the target trending index value. Higher priority candidate trending topics are listed first, and lower priority candidate trending topics are listed last.

[0104] Understandably, in practical applications, the method of identifying target hot topics in this application can not only be by filtering the Top-N candidate hot topics, but also by directly determining the candidate hot topics whose target hot topic index is greater than or equal to the preset hot topic index threshold as target hot topics, or by combining multiple constraints, requiring the candidate hot topics to simultaneously meet the following constraints: the target hot topic index is greater than the preset hot topic index threshold, and the first emotion score and the second emotion score are both greater than the set emotion score, thereby ensuring that the selected target hot topics are not only highly popular, but also in line with the user's emotional tendency.

[0105] As an example, this application can use the target trending topic as one of the input data for a pre-trained trending content generation model, so that the trending content generation model can automatically generate brand-related viral content associated with the target trending topic.

[0106] In this application, the trending content generation model is a third-order quantitative prediction model. It calculates the basic popularity score of the target trending topic and its relevance to the business (industry matching score, brand matching score) to obtain a comprehensive "target trending index." Based on this index, the model identifies the target trending topic. Upon receiving the target trending topic, it uses it as input data and combines it with other trending topic data to automatically generate brand-related viral content closely related to the target trending topic. For example, when the target trending topic is "new skincare products attracting user attention," the trending content generation model can generate marketing copy, promotional headlines, and other brand-related viral content related to the skincare product, thereby helping companies quickly respond to trending trends and output highly relevant content.

[0107] As another example, this application can convert popular brand content into content suggestion reports and output them to user terminals for easy viewing: When a target trending topic with a high target trending index is received, a topic extension direction generation request is automatically triggered. Using a preset content template and the corresponding trending topic data, the trending topic data is automatically populated into a suitable content template, generating several topic extension direction suggestions. These extension direction suggestions include recommended content formats (such as text and images, short videos, and Q&A), suitable platform channels, suggested tone, and target audience. Furthermore, after generating topic extension direction suggestions, a seed content generation request is triggered, inputting the target trending topic with a high target trending index and the topic extension direction suggestions into a large language model (such as a trending content generation model) to generate a sample copy or content outline, providing inspirational content for creators.

[0108] To better understand the method for predicting trending content in this application, a complete example will be used to illustrate it below.

[0109] Step 1: Collect trending topics and trending topic data from across the internet.

[0110] Obtain trending topics, search keywords, and basic popularity data through social media platform APIs and web crawlers, including at least metrics such as readership and discussion volume.

[0111] Step two involves storing the collected data in a database (such as MySQL or MongoDB) to form the original topic database for subsequent analysis.

[0112] Step 3: Determine whether the collected topics are trending topics.

[0113] 1) Calculate the basic heat score Short-term explosive power: Calculate the growth rate of each topic's popularity in the last 24 hours to assess the topic's immediate attention.

[0114] Sustained vitality: Calculate the popularity trend of each topic over the past 7 days to determine whether the topic can continue to attract attention.

[0115] Overall score: The basic popularity score for each topic is obtained by combining short-term and long-term trends.

[0116] 2) Calculate the industry matching score Industry Keyword Library Definition: A collection of keywords that represent the functional, social, and emotional attributes of the industry. For example, keywords in the beauty industry include: efficacy, ingredients, skincare, appearance, confidence, etc.

[0117] Similarity calculation: NLP techniques (such as TF-IDF and Word2Vec cosine similarity) are used to compare the relevance of candidate hot topics with the industry keyword database and generate industry matching scores.

[0118] 3) Calculate the brand matching score Brand Keyword Library Definition: A collection of keywords that represent the brand's functions, social and emotional attributes. For example, the current brand focuses on "professionalism", "high-end" and "luxury".

[0119] Similarity calculation: NLP technology is also used to compare the relevance between candidate trending topics and the brand keyword database to obtain a brand matching score.

[0120] Step 4: Calculate the hit index (target hot topic index).

[0121] The basic popularity score, industry matching score, and brand matching score are weighted and combined to obtain the final ranking prediction score: Popularity Index = Basic Popularity Score × w1 + Industry Matching Score × w2 + Brand Matching Score × w3.

[0122] The weights can be initially set to 1:1:1, and can be adjusted and optimized based on data feedback.

[0123] Step 5: Generate a list of trending topics.

[0124] Sort candidate trending topics from highest to lowest based on their popularity index, and output a TOP-N list.

[0125] Step six: Generate suggestions for extending the topic.

[0126] Fill the preset content template with topic keywords, industry keywords, brand keywords, and emotion tags.

[0127] Suggested directions for extending the topic include: suggested entry points (based on recent "hot topics", you can start from "industry keywords" and "brand keywords" to explore content creation), recommended content formats (text / images / short videos / Q&A), suitable channels, content tone, target audience, priority, and recommended release time window, etc.

[0128] Step 7: Generate AIGC (Artificial Intelligence Generated Content) seed content.

[0129] High-index topics and their extensions, brand context information, and channel constraints are encapsulated as standardized inputs. A large language model is then used to generate viral brand content, such as sample copy or content outlines, including marketing copy, short video scripts, graphic outlines, and suggestions for titles and tags.

[0130] In this embodiment, firstly, trending topic data from different social media platforms is acquired. Then, based on a pre-built trending feature database, semantic analysis and feature mapping are performed on the trending topic data to generate trending topic association information. The trending topic association information includes at least candidate trending topics. When the type of trending topic association information is industry association, the industry matching score corresponding to the candidate trending topic is calculated. When the type of trending topic association information is brand association, the brand matching score corresponding to the candidate trending topic is calculated. Multiple popularity scores are weighted and fused to obtain the target trending index. Finally, the target trending topic is selected from each candidate trending topic based on the target trending index.

[0131] Compared with related technologies, the technical solution of this application has the following advantages: First, by acquiring hot topic data from different social media platforms, cross-platform statistical analysis of hot topic data is achieved to comprehensively understand the overall trend of hot topics. Furthermore, through in-depth semantic analysis and feature mapping of hot topic data, topic-related information that is highly consistent with the company's products can be generated. Second, this application specifically calculates the popularity scores corresponding to hot topic-related information under different types, and then weights and merges the popularity scores of different dimensions to obtain the target hot topic index. The target hot topic index obtained in this way not only takes into account the popularity level of the topic in different dimensions, but also comprehensively reflects the dissemination potential of the topic. Thus, based on the target hot topic index, the target hot topic can be accurately and efficiently identified, and the brand's hit content associated with the target hot topic can be predicted more accurately.

[0132] Corresponding to the aforementioned application function implementation method embodiments, this application also provides a popular content prediction device, electronic device, and corresponding embodiments.

[0133] Figure 3This is a schematic diagram of the structure of a popular content prediction device shown in an embodiment of this application.

[0134] See Figure 3 The device includes at least the following modules: The multi-source data acquisition module 301 is used to acquire trending topic data provided by different social media platforms; The association information generation module 302 is used to perform semantic analysis and feature mapping on hot topic data based on a pre-built hot topic feature database to generate hot topic association information; the hot topic association information includes at least candidate hot topics; The hot topic index generation module 303 is used to calculate the heat score corresponding to the candidate hot topic for different types of hot topic association information, and to perform weighted fusion of multiple heat scores to obtain the target hot topic index. The content prediction module 304 is used to filter target hot topics from various candidate hot topics based on the target hot topic index, and predict the brand's hit content related to the target hot topic content.

[0135] As an optional example of an embodiment of this application, the associated information generation module 302 includes: The data extraction submodule is used to extract industry attributes and historical brand characteristics from a pre-built hotspot feature database; The candidate hot topic identification submodule is used to perform semantic analysis on hot topic data to obtain candidate hot topics and candidate hot keywords for the candidate hot topics. The first type determination submodule is used to determine the type of candidate hot topics as industry-related if the candidate hot keywords belong to industry attributes; The second type determination submodule is used to determine the type of candidate hot topic as brand association type if the candidate hot keywords belong to historical brand characteristics; The related information generation submodule is used to perform structured processing on candidate hot topics, the types of candidate hot topics, and candidate hot keywords to generate related information for hot topics.

[0136] As an optional example of an embodiment of this application, the candidate hot topic identification submodule is used for: Semantic analysis of trending topic data yields multiple topics; Calculate the popularity growth rate of each topic in the first set time window, and calculate the popularity trend score of each topic in the second set time window; Based on the popularity growth rate and popularity trend score, each topic is scored to obtain a basic popularity score. Topics with a base popularity score greater than or equal to a preset popularity threshold are identified as candidate hot topics, and candidate hot keywords corresponding to the candidate hot topics are extracted from the hot topic data.

[0137] As an optional example of an embodiment of this application, the popularity score includes an industry matching score, and the hot topic index generation module 303 is used for: When the type of information associated with a trending topic is industry-related, extract industry keywords and the primary sentiment keyword from the candidate trending keywords. Calculate the first text similarity between industry keywords and industry attributes; According to the preset score mapping relationship, the first emotion score corresponding to the first emotion keyword is obtained; The industry matching score is obtained by weighting and fusing the first text similarity score and the first sentiment score.

[0138] As an optional example of an embodiment of this application, the popularity score includes a brand matching score, and the popularity index generation module 303 is used for: When the type of information associated with a trending topic is brand-related, extract brand keywords and second sentiment keywords from candidate trending keywords. Calculate the second text similarity between brand keywords and historical brand features; According to the preset score mapping relationship, the second emotion score corresponding to the second emotion keyword is obtained; The brand matching score is obtained by weighting and fusing the second text similarity score and the second emotion score.

[0139] As an optional example of an embodiment of this application, the hotspot index generation module 303 is used for: For each popularity score, assign a popularity weight; Multiple popularity scores are weighted and merged according to their popularity weights to generate a target hot topic index, which is a data representation of the overall popularity of candidate hot topics.

[0140] As an optional example of an embodiment of this application, the content prediction module 304 is used for: The candidate hot topics are sorted according to the target hot topic index value to generate a list of hot topics; The top N candidate hot topics in the hot topic list are identified as target hot topics, where N is a positive integer. The target trending topics are the input data for a pre-trained trending content generation model, which enables the model to automatically generate brand-related viral content associated with the target trending topics.

[0141] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated further here.

[0142] Figure 4 This is a schematic diagram of the structure of an electronic device shown in an embodiment of this application.

[0143] See Figure 4 The electronic device 400 includes a memory 410 and a processor 420.

[0144] The processor 420 can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor. Memory 410 may include various types of storage units, such as system memory, read-only memory (ROM), and permanent storage devices. ROM may store static data or instructions required by processor 420 or other modules of the computer. Permanent storage devices may be read-write storage devices. Permanent storage devices may be non-volatile storage devices that retain stored instructions and data even when the computer is powered off. In some embodiments, permanent storage devices use mass storage devices (e.g., magnetic or optical disks, flash memory) as permanent storage devices. In other embodiments, permanent storage devices may be removable storage devices (e.g., floppy disks, optical drives). System memory may be a read-write storage device or a volatile read-write storage device, such as dynamic random access memory. System memory may store some or all of the instructions and data required by the processor during operation. Furthermore, memory 410 may include any combination of computer-readable storage media, including various types of semiconductor memory chips (e.g., DRAM, SRAM, SDRAM, flash memory, programmable read-only memory), and disks and / or optical disks may also be used. In some embodiments, memory 410 may include a removable storage device that is readable and / or writable, such as a laser disc (CD), a read-only digital multifunction optical disc (e.g., DVD-ROM, dual-layer DVD-ROM), a read-only Blu-ray disc, an ultra-high density optical disc, a flash memory card (e.g., SD card, mini SD card, Micro-SD card, etc.), a magnetic floppy disk, etc. Computer-readable storage media do not contain carrier waves or transient electronic signals transmitted wirelessly or via wired connections.

[0145] The memory 410 stores executable code, which, when processed by the processor 420, can cause the processor 420 to execute part or all of the methods described above.

[0146] Furthermore, the method according to this application can also be implemented as a computer program or computer program product, which includes computer program code instructions for performing some or all of the steps in the method described above.

[0147] Alternatively, this application may be implemented as a computer-readable storage medium (or a non-transitory machine-readable storage medium or a machine-readable storage medium) storing executable code (or computer program or computer instruction code) that, when executed by a processor of an electronic device (or server, etc.), causes the processor to perform part or all of the steps of the methods described above according to this application.

[0148] This application also provides a computer program product, which includes computer instructions that, when executed by a processor, implement the method described above.

[0149] The various embodiments of this application have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical application, or improvement of the technology in the market, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. A method for predicting viral content, characterized in that, include: Obtain trending topic data from different social media platforms; Based on a pre-built hot topic feature database, semantic analysis and feature mapping are performed on the hot topic data to generate hot topic association information; the hot topic association information includes at least candidate hot topics. For different types of trending topic association information, the popularity score corresponding to the candidate trending topic is calculated, and multiple popularity scores are weighted and fused to obtain the target trending index; Based on the target hot topic index, target hot topics are selected from various candidate hot topics, and brand hit content associated with the target hot topic content is predicted.

2. The method according to claim 1, characterized in that, The process involves performing semantic analysis and feature mapping on the hot topic data based on a pre-built hot topic feature database to generate hot topic association information, including: Extract industry attributes and historical brand characteristics from a pre-built database of hotspot features; Semantic analysis is performed on the hot topic data to obtain the candidate hot topics and candidate hot keywords of the candidate hot topics; If the candidate hot keywords belong to the industry attribute, then the type of the candidate hot topic is determined to be industry-related. If the candidate hot keywords belong to the historical brand characteristics, then the type of the candidate hot topic is determined to be brand association type; The candidate hot topics, their types, and their keywords are structured to generate associated information about the hot topics.

3. The method according to claim 2, characterized in that, The step of performing semantic analysis on the hot topic data to obtain the candidate hot topics and candidate hot keywords for the candidate hot topics includes: Semantic analysis was performed on the aforementioned trending topic data to obtain multiple topics; Calculate the popularity growth rate of each topic in the first set time window, and calculate the popularity trend score of each topic in the second set time window; Based on the popularity growth rate and the popularity trend score, each topic is scored to obtain a basic popularity score. Topics with a basic popularity score greater than or equal to a preset popularity threshold are identified as candidate hot topics, and candidate hot keywords corresponding to the candidate hot topics are extracted from the hot topic data.

4. The method according to claim 2, characterized in that, The popularity score includes an industry matching score. The calculation of the popularity score corresponding to the candidate hot topics for different types of hot topic association information includes: When the type of information associated with the hot topic is the industry-related type, extract the industry keywords and the first sentiment keyword from the candidate hot keywords; Calculate the first text similarity between the industry keywords and the industry attributes; According to the preset score mapping relationship, the first emotion score corresponding to the first emotion keyword is obtained; The industry matching score is obtained by weighted fusion of the first text similarity and the first sentiment score.

5. The method according to claim 2, characterized in that, The popularity score includes a brand matching score. The calculation of the popularity score corresponding to the candidate trending topics for different types of trending topic association information includes: When the type of information associated with the hot topic is the brand association type, extract the brand keyword and the second emotion keyword from the candidate hot topic keywords; Calculate the second text similarity between the brand keywords and the historical brand features; According to the preset score mapping relationship, the second emotion score corresponding to the second emotion keyword is obtained; The brand matching score is obtained by weighted fusion of the second text similarity and the second emotion score.

6. The method according to any one of claims 1-5, characterized in that, The weighted fusion of multiple popularity scores to obtain the target hot topic index includes: For each popularity score, assign a popularity weight; The multiple popularity scores are weighted and fused according to the popularity weight to generate the target hot topic index, which is data representing the overall popularity of the candidate hot topics.

7. The method according to claim 1, characterized in that, The step of filtering target hot topics from various candidate hot topics based on the target hot topic index and predicting brand-related viral content associated with the target hot topic includes: The candidate hot topics are sorted according to the value of the target hot topic index to generate a list of hot topics; The top N candidate hot topics in the hot topic list are determined as the target hot topic, where N is a positive integer; The target trending topic is the input data of a pre-trained trending content generation model, which enables the trending content generation model to automatically generate brand-related viral content associated with the target trending topic.

8. A device for predicting viral content, characterized in that, include: The multi-source data acquisition module is used to acquire trending topic data from different social media platforms; The association information generation module is used to perform semantic analysis and feature mapping on the hot topic data based on a pre-built hot topic feature database to generate hot topic association information; the hot topic association information includes at least candidate hot topics; The hot topic index generation module is used to calculate the heat score corresponding to the candidate hot topic for different types of hot topic association information, and to perform weighted fusion of multiple heat scores to obtain the target hot topic index. The content prediction module is used to filter target hot topics from various candidate hot topics based on the target hot topic index, and predict brand hit content associated with the target hot topic content.

9. An electronic device, characterized in that, include: processor; as well as A memory having executable code stored thereon, which, when executed by the processor, causes the processor to perform the method as described in any one of claims 1-7.

10. A computer-readable storage medium having executable code stored thereon, which, when executed by a processor of an electronic device, causes the processor to perform the method as described in any one of claims 1-7.

Images