Video classification method and device

Abstract

The embodiment of the invention provides a video classification method. The method comprises the following steps: acquiring a plurality of target video files; obtaining interaction information of each target video file; wherein the interaction information comprises multiple pieces of comment information and / or bullet screen information; obtaining an interaction keyword in each piece of interaction information; and aggregating the plurality of target video files with the same or similar interaction keywords into a file set to perform video classification. According to the video classification method and device, the videos are classified in combination with the interactive behaviors (such as bullet screens and comments) of the users under the different videos, so that the interactive contents in the different videos can be more comprehensively obtained, the classification of the videos is more in line with the actual opinions of the users, and the accuracy and comprehensiveness of the video classification are improved. Besides, by aggregating the file set, centralized management and analysis of the video files under the same or similar interactive contents are realized, so that the video classification efficiency is improved.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of data processing technology, and in particular, to a video classification method, apparatus, computer equipment, computer-readable storage medium, and computer program product. Background Art

[0002] With the continuous development of network technology and terminal devices, video has become a major communication medium. By classifying multiple videos, video platforms can more effectively provide users with relevant content and enhance user experience. At the same time, the platform can also better manage video-related interactive content based on video classification results.

[0003] However, existing video classification methods rely on video titles or video content, and the accuracy of video classification is low.

[0004] It should be noted that the above content is not necessarily prior art, nor is it intended to limit the scope of patent protection of this application. Summary of the invention

[0005] The embodiments of the present application provide a video classification method, apparatus, computer device, computer-readable storage medium, and computer program product to solve or alleviate one or more of the technical problems raised above.

[0006] One aspect of an embodiment of the present application provides a video classification method, the method comprising: Get multiple target video files; Obtaining interactive information of each of the target video files; wherein the interactive information includes a plurality of comment information and / or barrage information; Obtaining interactive keywords in each of the interactive information; A plurality of the target video files having the same or similar interactive keywords are aggregated into a file set to correspond to one interactive content.

[0007] Optionally, obtaining the interaction keyword in each of the interaction information includes: Based on the word segmentation library, a plurality of word segments are obtained from the plurality of video interaction information; Obtaining current word frequency distribution of a plurality of the word segments; According to the current word frequency distribution and the historical word frequency distribution of the plurality of word segments, obtaining the prominence of each of the word segments; The segmented words whose significance is greater than a first preset threshold are determined as the interactive keywords.

[0008] Optionally, the interactive information includes one or more words, and adjacent words may form a word combination; the word segmentation library updates new words according to the plurality of video interactive information, and the updating operation includes: According to the frequency of occurrence of each of the words and the word combination in the interactive information, obtaining the cohesion degree of the word combination; When the cohesion degree is greater than a second preset threshold, determining the word combination as a target word; Performing left-right entropy analysis on the target word to obtain a left-right entropy value of the target word; the left-right entropy value is used to indicate the degree of independence of the target word; When the left-right entropy value is greater than a third preset threshold, the target word is determined to be the new word.

[0009] Optionally, the method further comprises: According to the prominence, sorting the plurality of interactive keywords in the same target video file in descending order; According to the sorting order of each of the interactive keywords, a sorting index number is assigned to each of the interactive keywords.

[0010] Optionally, the plurality of target video files include a first target video file and a second target video file, and aggregating the plurality of target video files having the same or similar interactive keywords into a file set includes: Determining the overlapping interaction keywords in the first target video file and the second target video file as target interaction keywords; The interaction content similarity between the first target video file and the second target video file is determined based on the target interaction keyword, the sorting index number of the target interaction keyword in the first target video file, and the sorting index number of the target interaction keyword in the second target video file; the interaction content similarity is used to determine whether the first target video file and the second target video file are aggregated.

[0011] Optionally, aggregating a plurality of the target video files having the same or similar interactive keywords into a file set further comprises: Constructing a connection graph structure; wherein the connection graph structure includes a first node and a second node, the first node and the second node are connected by an edge, and the edge corresponds to a weight value; Adding the first target video file to the first node of the connection graph structure; Adding the second target video file to a second node of the connection graph structure; Setting the interactive content similarity between the first target video file and the second target video file as the weight value of the edge connecting the first node and the second node; Based on the connection graph structure, multiple nodes are aggregated into a node cluster to obtain the corresponding file set.

[0012] Optionally, multiple target video files are obtained, including: Get multiple video files; Obtaining the amount of interaction information of each of the plurality of video files; A plurality of target video files are obtained from the plurality of video files, wherein the amount of interactive information of the target video files is greater than a fourth preset threshold.

[0013] Optionally, the method further comprises: Obtaining the video file title of the target video file in the file set; Based on the video file title, a core title is extracted from a plurality of the video file titles as the title of the interactive content.

[0014] Another aspect of an embodiment of the present application provides a video classification device, the device comprising: An acquisition module, used to acquire multiple target video files; used to acquire interactive information of each of the target video files; wherein the interactive information includes multiple comment information; and used to acquire interactive keywords in each of the interactive information; The aggregation module is used to aggregate a number of the target video files having the same or similar interactive keywords into a file set for video classification.

[0015] Another aspect of an embodiment of the present application provides a computer device, including: at least one processor; and a memory communicatively coupled to the at least one processor; Wherein: the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method as described above.

[0016] Another aspect of an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer instructions, and when the computer instructions are executed by a processor, the method described above is implemented.

[0017] Another aspect of an embodiment of the present application provides a computer program product, including a computer program, which implements the method described above when executed by a processor.

[0018] The above technical solution adopted in the embodiment of the present application may have the following advantages: By obtaining interactive information such as comment information or bullet screen information of the target video file, obtaining interactive keywords based on the interactive information, and then aggregating several target video files with the same or similar interactive keywords into a file set, video classification under the same or similar interactive information can be achieved. This embodiment classifies videos by combining the interactive behaviors of users under different videos (such as bullet screen, comments, etc.), and can more comprehensively obtain the interactive content in different videos, so that the classification of videos is more in line with the actual opinions of users, thereby improving the accuracy and comprehensiveness of video classification. In addition, by aggregating file sets, centralized management and analysis of video files under the same or similar interactive information can be achieved, thereby improving the efficiency of video classification. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The accompanying drawings exemplarily illustrate the embodiments and constitute a part of the specification, and together with the text description of the specification, are used to explain the exemplary implementation of the embodiments. The embodiments shown are for illustrative purposes only and do not limit the scope of the claims. In all drawings, the same reference numerals refer to similar but not necessarily identical elements.

[0020] Figure 1 A flowchart of a video classification method according to Embodiment 1 of the present application is schematically shown; Figure 2 Schematically shows Figure 1 Flow chart of sub-steps of step S104; Figure 3 Schematically shows Figure 2 Flow chart of sub-steps of step S200; Figure 4 The following schematically shows a newly added flow chart of the video classification method according to the first embodiment of the present application; Figure 5 Schematically shows Figure 1 Flow chart of sub-steps of step S106; Figure 6 Schematically shows Figure 1 Another sub-step flow chart of step S106; Figure 7 Schematically shows Figure 1 Flow chart of sub-steps of step S100; Figure 8 The following schematically shows a newly added flow chart of the video classification method according to the first embodiment of the present application; Fig. 9 A block diagram schematically shows a video classification device according to the second embodiment of the present application; and Fig.10 The hardware architecture diagram of the computer device according to the third embodiment of the present application is schematically shown. DETAILED DESCRIPTION

[0021] In order to make the purpose, technical solutions and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present application.

[0022] It should be noted that the descriptions involving "first", "second", etc. in the embodiments of the present application are only for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, the technical solutions between the various embodiments can be combined with each other, but they must be based on the ability of ordinary technicians in the field to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such combination of technical solutions does not exist and is not within the scope of protection required by this application.

[0023] In the description of the present application, it should be understood that the numerical labels before the steps do not indicate the order in which the steps are executed, but are only used to facilitate the description of the present application and to distinguish each step, and therefore should not be understood as a limitation on the present application.

[0024] First, the following terms are explained: Point-to-point mutual information: is a measure of the strength of the association between two events or words that appear together.

[0025] Left-right entropy: It is an indicator to measure the diversity of the context around a word, used in new word discovery and text analysis. The richer the distribution of left-right neighbor words and the greater the randomness, the higher the entropy value, which is used to determine whether a phrase is an independent new word.

[0026] Bayesian average: refers to an average calculation method that combines sample data and prior information. It can effectively reduce the extreme impact caused by small sample size or data bias, thereby obtaining a more robust estimate.

[0027] Community discovery algorithm: An algorithm used to identify closely connected groups or communities between nodes in a network, which can find subsets of closely related nodes in the network.

[0028] Secondly, in order to facilitate those skilled in the art to understand the technical solutions provided in the embodiments of the present application, the relevant technologies are described below: With the continuous development of network technology and terminal devices, video has become a major communication medium. Video creators publish video works through video platforms, and users can discuss video works through methods such as bullet screens and comments, thereby generating corresponding interactive content. By classifying multiple videos, video platforms can more effectively provide users with relevant content and enhance user experience. At the same time, the platform can also better manage video-related interactive content based on video classification results.

[0029] Existing video classification methods rely on video titles or video content, and the accuracy of video classification is low.

[0030] To this end, an embodiment of the present application provides a technical solution for video classification. In this technical solution, by obtaining interactive information such as comment information or barrage information of the target video file, interactive keywords are obtained based on the interactive information, and then several target video files with the same or similar interactive keywords are aggregated into a file set to achieve video classification under the same or similar interactive information. This embodiment classifies videos by combining the interactive behaviors of users under different videos (such as barrage, comments, etc.), so as to more comprehensively obtain the interactive content in different videos, so that the classification of videos is more in line with the actual opinions of users, thereby improving the accuracy and comprehensiveness of video classification. In addition, by aggregating file sets, centralized management and analysis of video files under the same or similar interactive information can be achieved, thereby improving the efficiency of video classification. See below for details.

[0031] Embodiment 1 Figure 1 The flowchart of the video classification method according to the first embodiment of the present application is schematically shown.

[0032] like Figure 1 As shown, the video classification method may include steps S100 to S106, wherein: Step S100, obtaining multiple target video files.

[0033] Step S102, obtaining interactive information of each of the target video files; wherein the interactive information includes multiple comment information and / or barrage information.

[0034] Step S104, obtaining the interaction keywords in each of the interaction information.

[0035] Step S106: Aggregate a number of the target video files having the same or similar interactive keywords into a file set for video classification.

[0036] The video classification method provided in this embodiment obtains interactive information such as comment information or bullet screen information of the target video file, obtains interactive keywords based on the interactive information, and then aggregates several target video files with the same or similar interactive keywords into a file set to achieve video classification under the same or similar interactive information. This embodiment classifies videos by combining the interactive behaviors of users under different videos (such as bullet screen, comments, etc.), and can more comprehensively obtain the interactive content in different videos, so that the classification of videos is more in line with the actual opinions of users, thereby improving the accuracy and comprehensiveness of video classification. In addition, by aggregating file sets, centralized management and analysis of video files under the same or similar interactive information is achieved, thereby improving the efficiency of video classification.

[0037] The following combination Figure 1 , each step in steps S100~S106 and other optional steps are explained in detail.

[0038] Step S100 , get multiple target video files.

[0039] In video classification, multiple target video files need to be collected to provide data support for subsequent video classification.

[0040] In one example, the video files released in a video platform in one day can be obtained, and the files with low discussion levels can be deleted. Due to the low discussion level, the video files have less interactive content and low reference value. By filtering out the files with low discussion levels, the running time of the mining algorithm can be reduced. For example, video files with a comment level of more than 200 on a specific day can be selected as target video files.

[0041] In one example, the API provided by the video platform can be used to automatically obtain multiple video files that meet the conditions based on keywords, time range and other screening conditions. By setting certain rules, for example, you can select videos related to a hot topic or hot event to ensure that the target video obtained meets the needs of the current video classification.

[0042] For example, if you need to mine interactive content related to the "World Cup", you can call the video platform API interface, use "World Cup" as the search keyword, set the time range from the start to the end of the game, and automatically capture the game videos, news videos or commentary videos about the World Cup released during this time period. The acquired videos can include video files from different media sources, and interactive content information related to the World Cup can be extracted from these video files.

[0043] Step S102 , obtain interactive information of each of the target video files; wherein the interactive information includes multiple comment information and / or barrage information.

[0044] The interactive content mined based on video titles or video content depends on the subjectivity of the video producer, resulting in low accuracy of video classification. Therefore, this embodiment can more comprehensively obtain the interactive content in different videos by combining the interactive behaviors of users under different videos (such as bullet screens, comments, etc.), making the video classification more consistent with the actual interaction of users, thereby improving the accuracy and comprehensiveness of video classification.

[0045] In one example, all interactive information related to the target video file can be obtained through the API interface or data statistics technology of the video platform. Interactive information may include users' evaluations, discussions and opinions on the video content or comments sent by viewers in real time while watching the video. Before obtaining the user's interactive information, the user may be informed through a pop-up window that his or her interactive information will be collected. The content of the notification may include the purpose of collecting the interactive information (for example, improving the user experience) to obtain the user's informed consent. In addition, data encryption technology or anonymization technology may be used to collect interactive information, and data storage and processing may be carried out in accordance with relevant data protection laws and regulations (such as the Cybersecurity Law) to protect user privacy and prevent data leakage.

[0046] In one example, after obtaining the interactive information of each target video file, the advertising recognition model can also be used to automatically screen the interactive information and filter out the identified advertisements, irrelevant content or spam. The advertising recognition model can be based on text classification algorithms, rule engines and other technologies, combined with preset advertising keywords, sentence patterns, graphic information, etc., to identify and exclude advertising content, thereby ensuring the data quality of the interactive information. Exemplarily, the interactive information can also be screened according to preset rules. For example, if certain keywords such as "discount", "free", "click link", etc. appear in the interactive information content, it is marked as an advertisement and the interactive information is deleted.

[0047] Step S104 , obtain the interactive keywords in each of the interactive information.

[0048] In one example, an interaction keyword may refer to a frequently appearing word, which may reflect the core topics that users are concerned about and help analyze the sentiment tendency and discussion focus of the interaction content.

[0049] In one example, interactive keywords can be extracted from interactive information through text mining technology. Each comment or bullet message can be segmented to separate the interactive information text into separate words. Then, natural language processing (NLP) technology is used in combination with methods such as word frequency statistics to identify keywords that appear frequently in the text, are representative, and are useful for interactive content analysis as interactive keywords.

[0050] In one example, since word segmentation depends on a word segmentation corpus, and the interactive information may involve emerging industry technologies, brand names, or user-created words that may not exist in the original word segmentation corpus. Therefore, a new word discovery algorithm can be used to process the interactive information, mine new words, and update the new words to the word segmentation corpus. For example, a new word discovery algorithm, such as the method based on pointwise mutual information and information entropy, can be used to automatically extract potential new words from the interactive information.

[0051] In one example, before mining keywords from the interactive information, the text data of the interactive information can be cleaned first, for example, removing stop words, punctuation marks, irrelevant noise information, etc., to improve the efficiency and accuracy of word segmentation and keyword extraction. Among them, stop words can refer to words that frequently appear in the text but are irrelevant to the analysis, such as "de", "le", "shi", etc. Irrelevant noise information can be, for example, advertising content, meaningless characters, duplicate content, etc. Information, and useless interactive information text can be removed through rule matching or model filtering.

[0052] Step S106 , aggregating several of the target video files with the same or similar interaction keywords into a file set for video classification.

[0053] By aggregating multiple target video files with the same or similar interaction keywords and grouping them into the same file set, the classification of interactive content can be achieved. Furthermore, it can help video platform managers quickly understand the situation of a specific interactive content, enabling video platform managers to formulate more accurate interactive content monitoring and response strategies.

[0054] In one example, the relationship between target video files can be calculated based on the keyword significance degree between the target video files. For example, for each pair of video manuscripts, the significance degree of the common keywords can be calculated through the Bayesian average algorithm, and the relationship between the videos can be weighted according to the significance degree of these keywords. After obtaining the weighted relationship between each pair of video manuscripts, these video manuscripts can be regarded as nodes in a graph, the keyword relationship between the videos as edges, and different weights can be assigned to each edge. Community discovery algorithms, such as the Louvain algorithm, the Infomap algorithm, etc., can be used to cluster the target video files on all nodes into different communities, thereby quickly aggregating the target video files describing the same or similar interactive content into the same file set.

[0055] In an alternative embodiment, as Figure 2 shown, step S104 may include: Step S200, obtaining multiple word segments from the multiple video interaction information according to the word segmentation library.

[0056] Step S202, obtaining the current word frequency distribution of the plurality of word segments.

[0057] Step S204: acquiring the prominence of each of the segmented words according to the current frequency distribution and the historical frequency distribution of the multiple segmented words.

[0058] Step S206: Determine the segmented words whose significance is greater than a first preset threshold as the interactive keywords.

[0059] In one example, in interactive content analysis, video interactive information (such as comments, bullet screens, etc.) is constantly changing, and new words and topics may be generated over time. By using a new word discovery algorithm, these new words that have not yet been included in the word segmentation library can be identified. For example, when mining interactive content about "smart home", some new words appear in the user's comments, such as "smart assistant leakage" or "privacy shield" and other words that do not exist in the original word segmentation library. Since the above new words do not exist in the original word segmentation library, effective interactive keywords cannot be mined during word segmentation, resulting in reduced video classification accuracy. Through the new word discovery algorithm, new words will be identified and updated to the word segmentation library in a timely manner, thereby ensuring that the subsequent word segmentation is more accurate and comprehensive.

[0060] In an example, the historical word frequency distribution may refer to the word frequency distribution of the word segment in the past word frequency statistics, for example, the word frequency statistics of last month, or the word frequency statistics of any time point in the past.

[0061] In one example, in the actual segmentation data, there may be a large number of stop words and high-frequency irrelevant words. These words are irrelevant to the interactive content of the video and may interfere with the accuracy of video classification. Therefore, after obtaining the current word frequency distribution and historical word frequency distribution data of multiple segmentations, the significance of each segmentation can be obtained through the Yebaes average algorithm. The higher the significance, the closer and more representative the segmentation is to the interactive content. The interactive keywords can be determined based on the significance of the segmentation. Exemplarily, the significance can be calculated by the following formula:

[0062] Among them, W represents the prominence of a certain word, R represents the basic score of a certain word, v represents the word frequency of a certain word in the video file, m represents the average word frequency of a certain word among all word frequencies in the video file, that is, m=all word frequencies in the video file / v, and C represents the average basic score of all words, that is, C=ΣR / the sum of word frequencies of all word segments. The basic score R of a certain word can be calculated by the following formula:

[0063] in, Indicates the historical frequency of a certain participle, Indicates the current word frequency of a certain word.

[0064] The YeBas average algorithm corrects the current data through historical data, so that high-frequency but meaningless word segmentations will not have too much impact on the analysis of interactive content.

[0065] After obtaining the prominence of each segmented word, it is possible to determine which segmented words are most relevant to the core topic in the current interactive content based on the prominence. By setting a first preset threshold, representative interactive keywords that are more valuable for analyzing the interactive content can be screened out.

[0066] In this embodiment, new words are obtained through a new word discovery algorithm and updated to the word segmentation library, and multiple word segments are obtained based on the updated word segmentation library, so that new words in video interactive information can be dynamically discovered, thereby improving word segmentation reliability. The significance of each word segmentation is obtained based on the obtained word segmentation statistical word frequency distribution and historical word frequency distribution, and then the interactive keywords are screened out according to the significance of each word segmentation, which can reduce the interference of common words, stop words, etc. on video classification, and screen out interactive keywords that are truly representative of interactive content, thereby improving the quality of video classification.

[0067] The word segmentation library can be updated regularly or on an implementation basis. For example, a new word discovery algorithm can be used to update new words. The New Word Discovery Algorithm is an algorithm in the field of natural language processing (NLP), which is mainly used to automatically identify and extract newly emerged words that have not yet been defined in dictionaries or vocabularies from a large amount of text data. New words can refer to some new, field-specific words, or words that suddenly become popular at a certain moment for some reason. The goal of the new word discovery algorithm is to be able to discover these words and to be able to effectively identify, annotate, count and classify them. The new word discovery algorithm can construct a word co-occurrence network or semantic graph in the text, discover possible new words through structural analysis of the graph, or use deep learning models (such as BERT, GPT, etc.) to automatically identify and predict new words through contextual information.

[0068] In an optional embodiment, if Figure 3 As shown, the interactive information includes one or more words, and adjacent words can form a word combination; the word segmentation library updates new words according to the plurality of video interactive information, and the updating operation is: Step S300: acquiring the cohesion degree of the word combination according to the frequency of occurrence of each of the words and the word combination in the interactive information.

[0069] Step S302: when the cohesion degree is greater than a second preset threshold, determining the word combination as a target word.

[0070] Step S304, performing left-right entropy analysis on the target word to obtain a left-right entropy value of the target word; the left-right entropy value is used to indicate the degree of independence of the target word.

[0071] Step S306: When the left-right entropy value is greater than a third preset threshold, determining that the target word is the new word.

[0072] In one example, the interactive information includes one or more words, and the multiple words can be divided by means of a sliding window. Specifically, the sliding window can be used to divide the interactive information text into continuous word combinations of different lengths. Exemplarily, when the window is set to 2, a combination of two continuous words can be generated. For example, for the interactive information "This movie is really wonderful!", if the sliding window is set to 2, the following two-word combinations can be generated from left to right: "this", "this movie", "movie", "true film", "really", "too", "too exquisite", "wonderful", "colorful". The way of dividing words by sliding windows can reduce the possibility of missing possible word combinations, thereby improving the accuracy of obtaining interactive keywords.

[0073] By analyzing the frequency of each word and word combination in the interactive information, we can identify two words that appear frequently and are closely related in the interactive information, and then obtain the cohesion degree of the two words. The cohesion degree of the word combination can measure the possibility of forming a new word between the two words.

[0074] In one example, the degree of cohesion between words can be calculated through point-to-point mutual information, thereby measuring whether two words tend to co-appear in interactive information such as comments and bullet comments, and constitute independent vocabulary or concepts. For example, if two words always appear together, that is, the frequency of the word combination in interactive information is high, and the frequency of the words appearing independently is relatively low, then the PMI value will be high, indicating that they are more likely to be a new word with independent meaning. PMI can be calculated by the following formula:

[0075] in, represents the frequency of word a, represents the frequency of word b, Indicates the frequency of word combinations of word a and word b.

[0076] In one example, after obtaining the cohesion degree of each word combination, a threshold of the cohesion degree can be set to screen out word combinations with strong correlation or semantic independence, thereby reducing the situation where irrelevant combinations are mistaken for new words.

[0077] For example, if word a is "this" and word b is "movie", is 2000, is 3000, is 1000, and according to the point-to-point mutual information calculation formula, the PMI value of "this movie" can be obtained as 5. When the second preset threshold is 2, since the PMI value of "this movie" is greater than the second preset threshold, "this movie" can be determined as the target word.

[0078] After obtaining the target word, the left-right entropy analysis can be used to further evaluate whether the target word can be used as an independent new word. By combining the cohesion degree of word combination and the left-right entropy analysis, new words with practical significance can be more accurately identified, thereby improving the accuracy of video classification. The higher the left-right entropy value of the target word after the left-right entropy analysis, the richer the collocation of the target word with other words, and the higher the possibility of the target word being an independent word.

[0079] For example, left-right entropy analysis can calculate the richness of the phrases constructed by adding characters to the left and right of the keyword, indicating whether the target word can be formed independently. Specifically, left-right entropy analysis calculates the probability distribution of the left and right adjacent characters of the keyword to determine the independence of the word in different contexts. The calculation formula is as follows:

[0080] in, Represents a character in the target word.

[0081] For example, the target word is "Light Speed ​​Movie", and the left and right entropy values ​​of the two words "Light Speed" and "Movie" are calculated by the formula. If the left and right entropy value of "Light Speed ​​Movie" is calculated to be 3.5. Among them, the left and right entropy value of "Light Speed" is relatively high, because its context combination is very rich, for example, it can be used in combination with words such as "fast", "electricity", and "car", which shows that it has strong independence. The left and right entropy value of "Movie" is relatively low, because its context is mostly fixed, such as "action", "love", "science fiction", etc., lacking diversity. After calculating the left and right entropy value of the target word, the third preset threshold can be set to 2. Since the left and right entropy value is greater than the third preset threshold, it can be determined that "Light Speed ​​Movie" is a new word and can be used independently.

[0082] In this embodiment, the target word is obtained by obtaining the cohesion degree of the word combination, and then the left-right entropy analysis is performed on the target word to further verify the independence of the target word in the context and the possibility of independent use, so as to screen out high-quality new words that can be independently formed, and reduce the situation where low-frequency or meaningless words that are too dependent on the context are determined as new words. When determining interactive keywords, it is possible to analyze effective new words, thereby improving the efficiency and accuracy of video classification.

[0083] In an optional embodiment, if Figure 4 As shown, the video classification method also includes: Step S400: sorting the plurality of interactive keywords in the same target video file in descending order according to the prominence.

[0084] Step S402: assigning a sorting index number to each of the interactive keywords according to the sorting order of each of the interactive keywords.

[0085] In one example, the prominence can reflect the importance of the interactive keywords in the video, and words with high prominence are usually closely related to the theme of the video content and public discussion points. For example, the interactive information of the target video file discusses multiple topics, among which some interactive keywords appear repeatedly and are highly relevant, while other interactive keywords appear less frequently or have weaker relevance, and the ranking can be performed based on the prominence value of the interactive keywords.

[0086] In one example, a corresponding index number may be assigned to each interactive keyword according to its sorting order from large to small. In this case, an interactive keyword with a high degree of significance will receive a lower index number, and an interactive keyword with a low degree of significance will receive a higher index number, thereby reflecting the priority and analysis focus of the interactive keyword.

[0087] For example, assume that at this time, the ranking results of four interactive keywords from the highest to the lowest in terms of prominence have been obtained: "match", "win", "error", and "referee". A ranking index number is assigned to each interactive keyword based on the ranking result. "match" is the most significant, so it will be assigned a ranking index number of 1; "win" is assigned a ranking index number of 2; "error" is assigned a ranking index number of 3; and "referee" is assigned a ranking index number of 4. Obtaining the interactive keywords and their corresponding ranking index numbers can ensure that the most important content is given priority.

[0088] In this embodiment, the interactive keywords in the video content are prioritized by their significance, thereby more accurately reflecting the relevance of the interactive keywords to the interactive content of the target video file. This ensures that in the subsequent interactive content analysis, sentiment tendency assessment or user behavior research, the interactive keywords that best represent the video content or best reflect the public's attention can be processed first, further improving the efficiency and accuracy of video classification.

[0089] In an optional embodiment, if Figure 5 As shown, the plurality of target video files include a first target video file and a second target video file, and step S106 includes: Step S500: determining the overlapping interaction keywords in the first target video file and the second target video file as target interaction keywords.

[0090] Step S502, determining the interaction content similarity between the first target video file and the second target video file based on the target interaction keyword, the sorting index number of the target interaction keyword in the first target video file, and the sorting index number of the target interaction keyword in the second target video file; the interaction content similarity is used to determine whether the first target video file and the second target video file are aggregated.

[0091] In one example, the interaction keywords in the two video files can be compared to find the intersection, and then the overlapping keywords can be screened out. Specifically, the interaction keywords of the first target video and the second target video can be regarded as two sets, and their intersection is calculated to find the overlapping interaction keywords in the two video files as target interaction keywords.

[0092] In one example, after the target interaction keyword is obtained, the formula used to determine the interaction content similarity between the first target video file and the second target video file according to the target interaction keyword, the sorting index number of the target interaction keyword in the first target video file, and the sorting index number of the target interaction keyword in the second target video file may be:

[0093] Wherein, W represents the sum of the weights of each target keyword. The higher W is, the higher the correlation between the first target video file and the second target video file is. I represents the number of target interaction keywords, and M represents the sorting index number of the target interaction keywords. Since the target interaction keywords correspond to two different sorting index numbers in the first target video file and the second target video file, M selects the larger sorting index number.

[0094] For example, assume that there are two video manuscripts, video A and video B: Interactive keywords in video A: "match", "victory", "referee".

[0095] Interactive keywords in video B: "match", "victory", "misjudgment", and "audience".

[0096] The overlapping keywords "match" and "win" are sorted according to the prominence of the overlapping keywords. If the ranking index of "match" in video A is 4 and in video B is 2, and the ranking index of "win" in video A is 2 and in video B is 1. According to the above weight calculation formula, the weights of the overlapping keywords "match" and "win" are calculated separately. Since we take the maximum ranking index value as the calculation basis, when calculating the weight, we use the maximum ranking value corresponding to these overlapping keywords in the two videos, that is: The weights for "match" are:

[0097] The weights for "win" are:

[0098] Add up the weights of all overlapping keywords to get the total weight W between the two videos ( + ), and then get the interactive content similarity between the two videos. The higher the interactive content similarity value, the more similar the discussion topics between the two video files are, and they can be aggregated into a video file set. The lower the interactive content similarity, it means that there are large differences in the discussion topics between the two videos, and it is not suitable to be aggregated into a video file set.

[0099] In this embodiment, by calculating the similarity of the interactive content as a basis for whether to aggregate multiple target video files with the same or similar interactive keywords, videos describing the same or similar interactive content are automatically aggregated together, reducing the workload of manual analysis and screening and improving the analysis speed and efficiency of the interactive content.

[0100] In an optional embodiment, if Figure 6 As shown, step S106 also includes: Step S600, constructing a connection graph structure; wherein the connection graph structure includes a first node and a second node, the first node and the second node are connected by an edge, and the edge corresponds to a weight value.

[0101] Step S602: adding the first target video file to the first node of the connection graph structure.

[0102] Step S604: adding the second target video file to the second node of the connection graph structure.

[0103] Step S606: setting the interactive content similarity between the first target video file and the second target video file as the weight value of the edge connecting the first node and the second node.

[0104] Step S608: Based on the connection graph structure, aggregating multiple nodes into a node cluster to obtain the corresponding file set.

[0105] In one example, the connection graph structure can be constructed by calculating the similarity of the interaction content between nodes. If the similarity of the interaction content between two nodes is high, there will be an edge between the two nodes, and the weight of the edge represents the size of the similarity of the interaction content.

[0106] In one example, each video file may be considered as a node, and each node may include interaction keywords with the video file and their corresponding prominence and ranking information.

[0107] In one example, each target video file is taken as a node in the graph and added to different node positions in the graph structure. By calculating the similarity of the interactive content between the two video files, the edge between the two nodes is determined, and the corresponding interactive content similarity is used as the weight value of the edge. The relationship between nodes can be represented by a graph structure such as an adjacency list or an adjacency matrix, and the similarity between nodes can be reflected by setting appropriate weight values.

[0108] In one example, based on the nodes and edges in the graph structure, the nodes in the graph structure can be aggregated using a graph clustering algorithm, a community discovery algorithm, etc., to identify a node cluster of video files with similar interactive content. When aggregating, the weight value of the edge can be referenced. The higher the weight of the edge, the greater the similarity of the interactive content between the two video files, and the higher the possibility of being classified into the same cluster.

[0109] Exemplarily, after obtaining the connection graph structure, one of the community discovery algorithms, such as the Louvain algorithm, the Girvan-Newman algorithm or the spectral clustering algorithm, etc., can be selected to identify the community in the connection graph. For example, the Louvain algorithm can be used, which is based on the idea of ​​modularity optimization and searches for the local optimal community structure by gradually merging nodes. For example, the connection graph structure has five nodes, corresponding to five target videos. In the initial state, video A, video B, video C, video D, and video E are each an independent community. The similarity between video A, video B, and video C is high, and the weight of the edge is large, so the Louvain algorithm will merge them into the same community. The similarity between video D and video E is also high, and the Louvain algorithm merges them into the same community. After obtaining two different communities, the video files with similar interactive content are aggregated into the same community to form two file sets.

[0110] In this embodiment, by constructing a connection graph structure, with the target video file as the node and the similarity of the interactive content between the videos as the weight value of the edge, a set of video files with similar interactive content can be efficiently identified. The connection graph structure can make full use of the connectivity, edge weights and structured relationships between nodes, combined with the graph clustering algorithm and weight calculation, to automatically identify and aggregate nodes with similar features without manual intervention. This makes data analysis more efficient and accurate, thereby improving the efficiency of video classification.

[0111] In an optional embodiment, if Figure 7 As shown, step S100 includes: Step S700, obtaining multiple video files.

[0112] Step S702, obtaining the amount of interaction information of each of the plurality of video files.

[0113] Step S704: acquiring a plurality of target video files from the plurality of video files, wherein the amount of interactive information of the target video files is greater than a fourth preset threshold.

[0114] In one example, video files may come from different video platforms. For example, the API interface or crawler technology of the video platform may be used to batch download relevant information of the video files from public resources, such as the title of the video, the number of interactive messages, the duration, the upload date, etc.

[0115] In one example, the amount of interactive information may include the number of comments, likes, shares, and views of the video, which may reflect the intensity of interaction between the audience and the video content.

[0116] By way of example, a fourth preset threshold can be set, such as when the number of comments is greater than 2000, to select target video files with a high number of interactive messages as target video files. The target video files have relatively rich interactive messages, which means that the videos have a high degree of audience participation, and further video classification or other research can be performed.

[0117] In this embodiment, by screening videos with high interaction, computing and analysis resources are allocated more efficiently. Since low-interaction videos contribute less to the overall analysis, focusing video classification on high-interaction videos can improve analysis efficiency and reduce unnecessary computation. In addition, video classification based on high-interaction videos can better reflect the audience's real opinions and hot topics, ensuring more accurate video classification.

[0118] In an optional embodiment, if Figure 8 As shown, the video classification method also includes: Step S800: obtaining the video file title of the target video file in the file set.

[0119] Step S802: based on the video file titles, extract a core title from a plurality of the video file titles as the title of the interactive content.

[0120] In an example, a file set corresponding to an interactive content may correspond to multiple video file titles, and it is difficult to extract the core information. Summary extraction models such as bertsum, matchsum, and textrank4zh can be selected to extract effective information as the title of the interactive content.

[0121] In one example, after obtaining the interactive content title, the interactive content can also be sorted to help users quickly obtain interactive content with high discussion. Specifically, relevant comment data can be obtained from each video file. The more comments there are, the higher the interactive content heat value corresponding to the video file. Based on the heat value of each interactive content, each interactive content is sorted.

[0122] In this embodiment, the title of the interactive content is extracted based on the titles of multiple video files in the file set, so that the user group can quickly understand the core content of the interactive content.

[0123] In order to make the present application easier to understand, an exemplary application is provided below.

[0124] S11: Get the video manuscripts published on the platform in one day.

[0125] S12: Delete the video manuscripts with less than 200 comments to obtain multiple videos to be processed.

[0126] S13: Obtain all comment data texts of the video to be processed.

[0127] S14: Divide all comment data texts into multiple words through a sliding window, traverse each text content from left to right, and set the window threshold to 2, then divide the comment data text into multiple two-character words.

[0128] S15: Calculate the occurrence frequency of each two-character word in all comment data texts, and the occurrence frequency of a combination of two adjacent two-character words.

[0129] S16: Using the formula for calculating mutual information between points , substitute the frequency value in S15, and calculate the cohesion degree PMI of each word combination.

[0130] S17: When the PMI is greater than 20, a left-right entropy analysis is performed on the word combination to obtain a left-right entropy value of the word combination, and the left-right entropy value is used as a basis for judging whether the word combination can be a new word.

[0131] S18: Load the new words obtained by the new word algorithm into the jieba word segmenter.

[0132] S19: Put the comment data texts of multiple videos to be processed into the Jieba word segmenter for processing to obtain multiple keywords and corresponding word frequencies.

[0133] S20: Based on the acquired keyword frequency and the historical keyword frequency at the same time point in the previous month, the significance of each keyword is calculated through the Bayesian average algorithm to filter out stop words and useless words.

[0134] S21: Sort the keywords in descending order of significance, and set a corresponding sorting index number for each keyword.

[0135] S22: Analyze the similarity of the interactive content of each two videos to be processed, first obtain the overlapping keywords and corresponding sorting index numbers between the two videos to be processed, and then use the formula , calculate the interactive content similarity of the two videos to be processed. Where M represents the value of the larger sorting index number corresponding to the overlapping keywords.

[0136] S23: The video to be processed is taken as a node, the value of W is taken as the weight value of the edge between two nodes, and multiple manuscripts to be processed that describe the same or similar interactive content are aggregated through a community discovery algorithm to obtain a file set.

[0137] In this exemplary application, by combining the interactive behaviors of users under different videos (such as bullet screens, comments, etc.) to classify videos, the interactive content in different videos can be more comprehensively obtained, so that the classification of videos is more in line with the actual interactive situation of users, thereby improving the accuracy and comprehensiveness of video classification. In addition, by aggregating file sets, centralized management and analysis of video files with the same or similar interactive content can be achieved, thereby improving the efficiency of video classification.

[0138] After obtaining different types of file sets, you can: (1) Displaying the file set and the interactive content represented by the file set on a display interface for user reference and execution of content management strategy; (2) Sending the file set and the interactive content represented by the file set to an opinion processing platform, so that the opinion processing platform executes a corresponding content management strategy according to the file set and the interactive content.

[0139] Embodiment 2 Fig. 9 The block diagram of the video classification device according to the second embodiment of the present application is schematically shown. The device can be divided into one or more program modules, one or more program modules are stored in a storage medium and executed by one or more processors to complete the embodiment of the present application. The program module referred to in the embodiment of the present application refers to a series of computer program instruction segments that can complete specific functions. The following description will specifically introduce the functions of each program module in this embodiment. Fig. 9 As shown, the device 1600 may include: an acquisition module 1610 and an aggregation module 1620, wherein: The acquisition module 1610 is used to acquire multiple target video files; to acquire interactive information of each of the target video files; wherein the interactive information includes multiple comment information; and to acquire interactive keywords in each of the interactive information; The aggregation module 1620 is used to aggregate a plurality of the target video files having the same or similar interactive keywords into a file set for video classification.

[0140] As an optional embodiment, the device 1600 is further used for: Based on the word segmentation library, a plurality of word segments are obtained from the plurality of video interaction information; Obtaining current word frequency distribution of a plurality of the word segments; According to the current word frequency distribution and the historical word frequency distribution of the plurality of word segments, obtaining the prominence of each of the word segments; The segmented words whose significance is greater than a first preset threshold are determined as the interactive keywords.

[0141] As an optional embodiment, the interactive information includes one or more words, and adjacent words can form a word combination; the word segmentation library updates new words according to the plurality of video interactive information, and the updating operation is as follows: According to the frequency of occurrence of each of the words and the word combination in the interactive information, obtaining the cohesion degree of the word combination; When the cohesion degree is greater than a second preset threshold, determining the word combination as a target word; Performing left-right entropy analysis on the target word to obtain a left-right entropy value of the target word; the left-right entropy value is used to indicate the degree of independence of the target word; When the left-right entropy value is greater than a third preset threshold, the target word is determined to be the new word.

[0142] As an optional embodiment, the device 1600 is further used for: According to the prominence, sorting the plurality of interactive keywords in the same target video file in descending order; According to the sorting order of each of the interactive keywords, a sorting index number is assigned to each of the interactive keywords.

[0143] As an optional embodiment, the plurality of target video files include a first target video file and a second target video file, and the apparatus 1600 is further configured to: Determining the overlapping interaction keywords in the first target video file and the second target video file as target interaction keywords; The interaction content similarity between the first target video file and the second target video file is determined based on the target interaction keyword, the sorting index number of the target interaction keyword in the first target video file, and the sorting index number of the target interaction keyword in the second target video file; the interaction content similarity is used to determine whether the first target video file and the second target video file are aggregated.

[0144] As an optional embodiment, the device 1600 is further used for: Constructing a connection graph structure; wherein the connection graph structure includes a first node and a second node, the first node and the second node are connected by an edge, and the edge corresponds to a weight value; Adding the first target video file to the first node of the connection graph structure; Adding the second target video file to a second node of the connection graph structure; Setting the interactive content similarity between the first target video file and the second target video file as the weight value of the edge connecting the first node and the second node; Based on the connection graph structure, multiple nodes are aggregated into a node cluster to obtain the corresponding file set.

[0145] As an optional embodiment, the device 1600 is further used for: Get multiple video files; Obtaining the amount of interaction information of each of the plurality of video files; A plurality of target video files are obtained from the plurality of video files, wherein the amount of interactive information of the target video files is greater than a fourth preset threshold.

[0146] As an optional embodiment, the device 1600 is further used for: Obtaining the video file title of the target video file in the file set; Based on the video file title, a core title is extracted from a plurality of the video file titles as the title of the interactive content.

[0147] Embodiment 3 Fig.10 The schematic diagram of the hardware architecture of a computer device 10000 suitable for implementing the video classification method according to the third embodiment of the present application is shown schematically. In some embodiments, the computer device 10000 may be a terminal device such as a smart phone, a wearable device, a tablet computer, a personal computer, a vehicle terminal, a game console, a virtual device, a workbench, a digital assistant, a set-top box, a robot, etc. In other embodiments, the computer device 10000 may be a rack server, a blade server, a tower server, or a cabinet server (including an independent server, or a server cluster composed of multiple servers), etc. Fig.10 As shown, the computer device 10000 includes but is not limited to: a memory 10010, a processor 10020, and a network interface 10030 that can communicate with each other through a system bus. Among them: The memory 10010 includes at least one type of computer-readable storage medium, and the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (such as SD or DX memory), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 10010 can be an internal storage module of the computer device 10000, such as a hard disk or memory of the computer device 10000. In other embodiments, the memory 10010 can also be an external storage device of the computer device 10000, such as a plug-in hard disk equipped on the computer device 10000, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. Of course, the memory 10010 can also include both the internal storage module of the computer device 10000 and its external storage device. In this embodiment, the memory 10010 is generally used to store an operating system and various application software installed in the computer device 10000, such as program codes of a video classification method, etc. In addition, the memory 10010 can also be used to temporarily store various data that have been output or are to be output.

[0148] In some embodiments, the processor 10020 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other chips. The processor 10020 is generally used to control the overall operation of the computer device 10000, such as performing control and processing related to data interaction or communication with the computer device 10000. In this embodiment, the processor 10020 is used to run the program code stored in the memory 10010 or process data.

[0149] The network interface 10030 may include a wireless network interface or a wired network interface, and the network interface 10030 is generally used to establish a communication link between the computer device 10000 and other computer devices. For example, the network interface 10030 is used to connect the computer device 10000 to an external terminal through a network, and to establish a data transmission channel and a communication link between the computer device 10000 and the external terminal. The network may be a wireless or wired network such as an intranet, the Internet, the Global System of Mobile communication (GSM), Wideband Code Division Multiple Access (WCDMA), 4G network, 5G network, Bluetooth, Wi-Fi, etc.

[0150] It should be pointed out that Fig.10 Only a computer device having components 10010 - 10030 is shown, but it should be understood that implementation of all of the components shown is not a requirement, and more or fewer components may alternatively be implemented.

[0151] In this embodiment, the video classification method stored in the memory 10010 can also be divided into one or more program modules and executed by one or more processors (such as processor 10020) to complete the embodiment of the present application.

[0152] Embodiment 4 The embodiment of the present application further provides a computer-readable storage medium, on which a computer program is stored, wherein when the computer program is executed by a processor, the steps of the video classification method in the embodiment are implemented.

[0153] In this embodiment, the computer-readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, disk, optical disk, etc. In some embodiments, the computer-readable storage medium can be an internal storage unit of a computer device, such as a hard disk or memory of the computer device. In other embodiments, the computer-readable storage medium can also be an external storage device of a computer device, such as a plug-in hard disk equipped on the computer device, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. Of course, the computer-readable storage medium can also include both the internal storage unit of the computer device and its external storage device. In this embodiment, the computer-readable storage medium is generally used to store an operating system and various application software installed on the computer device, such as the program code of the video classification method in the embodiment. In addition, the computer-readable storage medium can also be used to temporarily store various types of data that have been output or are to be output.

[0154] Embodiment 5 An embodiment of the present application also provides a computer program product, including a computer program, which implements the method in the above embodiment when executed by a processor.

[0155] Obviously, those skilled in the art should understand that the modules or steps of the above-mentioned embodiments of the present application can be implemented by general-purpose computer devices, they can be concentrated on a single computer device, or distributed on a network composed of multiple computer devices, optionally, they can be implemented by executable program codes of computer devices, so that they can be stored in a storage device and executed by the computer device, and in some cases, the steps shown or described can be executed in a different order from that herein, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. In this way, the embodiments of the present application are not limited to any specific combination of hardware and software.

[0156] It should be noted that the above are only preferred embodiments of the present application, and the patent protection scope of the present application is not limited thereto. Any equivalent structure or equivalent process transformation made using the contents of the specification and drawings of the present application, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A video classification method, characterized in that: The method comprises: Get multiple target video files; Obtaining interactive information of each of the target video files; wherein the interactive information includes a plurality of comment information and / or barrage information; Obtaining interactive keywords in each of the interactive information; A plurality of the target video files having the same or similar interactive keywords are aggregated into a file set for video classification.

2. The method according to claim 1, characterized in that Obtaining the interactive keywords in each interactive information includes: Based on the word segmentation library, a plurality of word segments are obtained from the plurality of video interaction information; Obtaining current word frequency distribution of a plurality of the word segments; According to the current word frequency distribution and the historical word frequency distribution of the plurality of word segments, obtaining the prominence of each of the word segments; The segmented words whose significance is greater than a first preset threshold are determined as the interactive keywords.

3. The method according to claim 2, characterized in that The interactive information includes one or more words, and adjacent words may form a word combination; The word segmentation library updates new words according to the plurality of video interaction information, and the updating operation includes: According to the frequency of occurrence of each of the words and the word combination in the interactive information, obtaining the cohesion degree of the word combination; When the cohesion degree is greater than a second preset threshold, determining the word combination as a target word; Performing left-right entropy analysis on the target word to obtain a left-right entropy value of the target word; the left-right entropy value is used to indicate the degree of independence of the target word; When the left-right entropy value is greater than a third preset threshold, the target word is determined to be the new word.

4. The method according to claim 2, characterized in that: The method further comprises: According to the prominence, sorting the plurality of interactive keywords in the same target video file in descending order; According to the sorting order of each of the interactive keywords, a sorting index number is assigned to each of the interactive keywords.

5. The method according to claim 4, characterized in that The plurality of target video files include a first target video file and a second target video file, and aggregating the plurality of target video files having the same or similar interactive keywords into a file set includes: Determining the overlapping interaction keywords in the first target video file and the second target video file as target interaction keywords; The interaction content similarity between the first target video file and the second target video file is determined based on the target interaction keyword, the sorting index number of the target interaction keyword in the first target video file, and the sorting index number of the target interaction keyword in the second target video file; the interaction content similarity is used to determine whether the first target video file and the second target video file are aggregated.

6. The method according to claim 5, characterized in that Aggregating a plurality of the target video files having the same or similar interactive keywords into a file set further includes: Constructing a connection graph structure; wherein the connection graph structure includes a first node and a second node, the first node and the second node are connected by an edge, and the edge corresponds to a weight value; Adding the first target video file to the first node of the connection graph structure; Adding the second target video file to a second node of the connection graph structure; Setting the interactive content similarity between the first target video file and the second target video file as the weight value of the edge connecting the first node and the second node; Based on the connection graph structure, multiple nodes are aggregated into a node cluster to obtain the corresponding file set.

7. The method according to any one of claims 1 to 6, characterized in that: Get multiple target video files, including: Get multiple video files; Obtaining the amount of interaction information of each of the plurality of video files; A plurality of target video files are obtained from the plurality of video files, wherein the amount of interactive information of the target video files is greater than a fourth preset threshold.

8. The method according to any one of claims 1 to 6, characterized in that: The method further comprises: Obtaining the video file title of the target video file in the file set; Based on the video file title, a core title is extracted from a plurality of the video file titles as the title of the interactive content.

9. A video classification device, characterized in that: The device comprises: An acquisition module, used to acquire multiple target video files; used to acquire interactive information of each of the target video files; wherein the interactive information includes multiple comment information; and used to acquire interactive keywords in each of the interactive information; The aggregation module is used to aggregate a number of the target video files having the same or similar interactive keywords into a file set for video classification.

10. A computer device, characterized in that: include: at least one processor; and a memory communicatively connected to the at least one processor; wherein: The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 8.

11. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and when the computer instructions are executed by a processor, the method according to any one of claims 1 to 8 is implemented.

12. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to claims 1 to 8 are implemented.

Images