Data pushing method, apparatus, device and medium

Abstract

Embodiments of the present application provide a data pushing method and device, equipment and medium, the method relates to the field of artificial intelligence, and the method comprises the following steps: receiving business object data uploaded by a first object through a first client; obtaining effective content corresponding to effective object behavior from historical content; taking content attribute values corresponding to the effective content as object attribute values of the first object, determining a weighted business object attribute value of the first object in a business object cycle based on the object attribute values, determining a first attribute tendency of the first object as an attribute tendency corresponding to the weighted business object attribute value; if the weighted business object attribute value meets a reverse attribute recommendation condition corresponding to the business object cycle, obtaining target recommendation content matched with a second attribute tendency, and issuing the target recommendation content to the first client. By adopting the present application, information stream content (i.e. target recommendation content) with an opposite attribute tendency can be pushed to the first object, and the flexibility of information stream content recommendation is improved.

Description

Technical Field

[0001] This application relates to the field of Internet technology, and in particular to a data push method, apparatus, device, and medium. Background Technology

[0002] Currently, news feed products primarily utilize object profiles (i.e., user profiles) of objects (e.g., object Y) to recommend news feed content (e.g., text, images, videos, etc.). For instance, news feed products can record behavioral data information of object Y (e.g., reading content, reading behavior, etc.) in various ways, and then perform object profile analysis (i.e., user profile analysis) on this behavioral data information to obtain an object profile used to represent object Y.

[0003] Understandably, the object profile of object Y can be used to infer object Y's interests and hobbies, and thus, information feed content that object Y might be interested in can be recommended. Therefore, when it is found that object Y tends to read a certain type of information feed content within a certain period of time, this type of information feed content is assumed to match object Y's interests and hobbies, and thus this type of information feed content will be continuously recommended to object Y, thereby reducing the flexibility of information feed content recommendation to some extent. Summary of the Invention

[0004] This application provides a data push method, apparatus, device, and medium that can push information stream content with opposite attribute tendencies (i.e., target recommended content) to a first object, thereby improving the flexibility of information stream content recommendation.

[0005] One embodiment of this application provides a data push method, which is executed by a server and includes:

[0006] Receive business object data uploaded by the first object through the first client; the business object data includes historical content provided by the first object during the business object's lifecycle.

[0007] The effective content corresponding to the effective object behavior is obtained from the historical content; the effective object behavior is determined by the object behavior recorded by the first object in relation to the historical content within the business object's lifecycle;

[0008] The content attribute value corresponding to the valid content is used as the object attribute value of the first object. Based on the object attribute value, the weighted business object attribute value of the first object in the business object cycle is determined, and the attribute tendency corresponding to the weighted business object attribute value is determined as the first attribute tendency of the first object.

[0009] If the weighted business object attribute value meets the reverse attribute recommendation condition corresponding to the business object period, then the target recommendation content that matches the second attribute tendency is obtained, and the target recommendation content is sent to the first client so that the first client can display the target recommendation content to the first object; the second attribute tendency is the opposite attribute tendency of the first attribute tendency.

[0010] One embodiment of this application provides a data push device, which runs on a server and includes:

[0011] The data receiving module is used to receive business object data uploaded by the first object through the first client; the business object data includes historical content provided by the first object during the business object's lifecycle;

[0012] The content acquisition module is used to retrieve valid content corresponding to valid object behaviors from historical content; valid object behaviors are determined by the object behaviors recorded by the first object in relation to historical content within the business object's lifecycle;

[0013] The attribute determination module is used to take the content attribute value corresponding to the valid content as the object attribute value of the first object, determine the weighted business object attribute value of the first object within the business object cycle based on the object attribute value, and determine the attribute tendency corresponding to the weighted business object attribute value as the first attribute tendency of the first object.

[0014] The content sending module is used to obtain target recommended content that matches the second attribute tendency if the weighted business object attribute value meets the reverse attribute recommendation condition corresponding to the business object period, and send the target recommended content to the first client so that the first client can display the target recommended content to the first object; the second attribute tendency is the opposite attribute tendency of the first attribute tendency.

[0015] If the historical content includes text and images, then the business object data also includes the object behavior corresponding to the text and images; the object behavior corresponding to the text and images includes the content duration corresponding to the first object.

[0016] The content acquisition module includes:

[0017] The speed determination unit is used to obtain the number of words in the content of the first object within the content duration and determine the content speed of the first object in relation to the content of the image and text.

[0018] The coefficient determination unit is used to obtain the content type of the graphic content and determine the dynamic adjustment coefficient corresponding to the graphic content based on the content type.

[0019] The speed comparison unit is used to obtain the base speed associated with the text and image content, adjust the base speed based on the dynamic adjustment coefficient, use the adjusted base speed as the dynamic speed, compare the content speed with the dynamic speed, and obtain the speed comparison result.

[0020] The content filtering unit is used to determine that the object behavior corresponding to the text and image content is a valid object behavior if the speed comparison result indicates that the content speed is less than the dynamic speed, and to select the text and image content that corresponds to the valid object behavior from the text and image content as valid content.

[0021] The attribute determination module includes:

[0022] The first processing unit is used to take the content attribute value corresponding to the valid content as the object attribute value of the first object, perform a first accumulation process on the object attribute value, and obtain the accumulated content attribute value.

[0023] The second processing unit is used to obtain the number of valid content items and determine the weighted business object attribute value of the first object within the business object cycle based on the accumulated content attribute value and the number of content items.

[0024] The tendency determination unit is used to determine the attribute tendency corresponding to the weighted business object attribute value as the first attribute tendency of the first object.

[0025] The device also includes:

[0026] The condition acquisition module is used to acquire the reverse attribute recommendation conditions corresponding to the business object cycle; the reverse attribute recommendation conditions include the first attribute threshold associated with the first attribute tendency and the second attribute threshold associated with the second attribute tendency.

[0027] The threshold comparison module is used to compare the weighted business object attribute value with the first attribute threshold and the second attribute threshold when the first attribute threshold is less than the second attribute threshold, and obtain the threshold comparison result.

[0028] The first determining module is used to determine that the weighted business object attribute value meets the reverse attribute recommendation condition if the threshold comparison result indicates that the weighted business object attribute value is less than the first attribute threshold or the weighted business object attribute value is greater than the second attribute threshold.

[0029] The device also includes:

[0030] The difference determination module is used to obtain the attribute configuration threshold associated with the first object and the attribute fluctuation value associated with the attribute configuration threshold, and to determine the first absolute difference between the weighted business object attribute value and the attribute configuration threshold;

[0031] The second determining module is used to determine that if the first absolute difference is greater than the attribute fluctuation value, the weighted business object attribute value satisfies the reverse attribute recommendation condition corresponding to the business object period.

[0032] The content sending module includes:

[0033] The reverse determination unit is used to obtain the attribute balance value in the reverse attribute recommendation conditions if the weighted business object attribute value meets the reverse attribute recommendation conditions corresponding to the business object period, and determine the reverse attribute value corresponding to the weighted business object attribute value based on the attribute balance value; the second absolute difference between the weighted business object attribute value and the attribute balance value is equal to the second absolute difference between the reverse attribute value and the attribute balance value.

[0034] The content matching unit is used to select the content to be recommended with the reverse attribute value as the target recommended content that matches the second attribute tendency, and then send the target recommended content to the first client.

[0035] The historical content consists of content to be recommended from the content database corresponding to the first client; the content database contains sample content uploaded by the second object through the second client.

[0036] The device also includes:

[0037] The attribute analysis module is used to obtain the content details information of the sample content, perform attribute analysis on the content details information to obtain the first attribute value and the second attribute value of the content details information, and determine the content attribute value of the sample content based on the first attribute value and the second attribute value;

[0038] The content storage module is used to store content details, content attribute values, first attribute values, and second attribute values ​​as content to be recommended in the content database.

[0039] The attribute analysis module includes:

[0040] The strategy acquisition unit is used to acquire the content text in the content details information of the sample content and the attribute analysis strategies associated with the sample content; the attribute analysis strategies include attribute dictionary analysis strategies;

[0041] The word segmentation determination unit is used to perform word segmentation on the content text based on the attribute dictionary analysis strategy, to obtain the first text segment of the content text, to perform stop word processing on the first text segment, and to use the first text segment after stop word processing as the second text segment of the content text.

[0042] The result determination unit is used to obtain a first word dictionary and a second word dictionary associated with the content text based on the attribute dictionary analysis strategy, and to perform string matching between the second text segmentation and the first word dictionary and the second word dictionary to obtain the string matching result.

[0043] The summation processing unit is used to determine the first attribute value and the second attribute value of the content details information based on the attribute matching values ​​in the string matching results, and to sum the first attribute value and the second attribute value to obtain the content attribute value of the sample content.

[0044] Among them, the attribute dictionary analysis strategy is used to indicate the acquisition of a first word dictionary associated with the first attribute tendency and a second word dictionary associated with the second attribute tendency;

[0045] The result determination unit includes:

[0046] The first matching subunit is used to search for a string that matches the second text segment in the first word dictionary when the first word dictionary associated with the content text is obtained based on the attribute dictionary analysis strategy. If a string that matches the second text segment is found in the first word dictionary, the found string is used as the first matching segment corresponding to the second text segment.

[0047] The first accumulation subunit is used to obtain the first weight information corresponding to the first matching word segmentation, and to perform a second accumulation process on the first weight information to obtain the first matching value associated with the first attribute tendency.

[0048] The second matching subunit is used to search for a string that matches the second text segment in the second word dictionary when a second word dictionary associated with the content text is obtained based on the attribute dictionary analysis strategy. If a string that matches the second text segment is found in the second word dictionary, the found string is used as the second matching segment corresponding to the second text segment.

[0049] The second accumulation subunit is used to obtain the second weight information corresponding to the second matching word segmentation, perform the second accumulation process on the second weight information, and obtain the second matching value associated with the second attribute tendency.

[0050] The result determination subunit is used to obtain the string matching result based on the first matching value and the second matching value.

[0051] Among them, attribute analysis strategies include model learning analysis strategies;

[0052] The attribute analysis module also includes:

[0053] The model acquisition unit is used to acquire a target network model associated with the content text based on the model learning analysis strategy; the target network model is obtained by training the initial network model based on the training text content and sample labels;

[0054] The feature extraction unit is used to input the content text into the target network model, and extract the features of the content text through the target network model to obtain the text attribute features corresponding to the content text.

[0055] The numerical prediction unit is used to determine the matching degree between the text attribute features and the sample attribute features in the classifier of the target network model, and to determine the first predicted value associated with the first attribute tendency and the second predicted value associated with the second attribute tendency based on the matching degree.

[0056] The result update unit is used to update the string matching result based on the first predicted value and the second predicted value.

[0057] This application provides a data push method, executed by a first client, comprising:

[0058] In response to a first trigger operation performed by a first object on historical content in a first client, the business object data of the first object on the historical content is recorded based on the first trigger operation, and the recorded business object data is sent to the server; the historical content includes valid content corresponding to valid object behavior; the valid object behavior is determined by the object behavior recorded by the first object on the historical content within the business object cycle;

[0059] The receiving server sends target recommended content associated with valid content based on business object data; the target recommended content is recommended content that matches the second attribute tendency; the second attribute tendency is the opposite attribute tendency of the first attribute tendency; the first attribute tendency is the attribute tendency corresponding to the weighted business object attribute value of the first object within the business object period; the weighted business object attribute value is determined based on the object attribute value of the first object; the object attribute value is determined by the content attribute value corresponding to the valid content;

[0060] The target recommended content is displayed to the first object in the first client.

[0061] One embodiment of this application provides a data push device, which runs in a first client, including:

[0062] The data sending module is used to respond to a first trigger operation performed by the first object on historical content in the first client, record business object data of the first object on the historical content based on the first trigger operation, and send the recorded business object data to the server; the historical content includes valid content corresponding to valid object behavior; the valid object behavior is determined by the object behavior recorded by the first object on the historical content within the business object period;

[0063] The content receiving module is used to receive target recommended content associated with valid content sent by the server based on business object data; the target recommended content is recommended content that matches the second attribute tendency; the second attribute tendency is the opposite attribute tendency of the first attribute tendency; the first attribute tendency is the attribute tendency corresponding to the weighted business object attribute value of the first object within the business object period; the weighted business object attribute value is determined based on the object attribute value of the first object; the object attribute value is determined by the content attribute value corresponding to the valid content;

[0064] The content display module is used to display the target recommended content to the first object in the first client.

[0065] One embodiment of this application provides a computer device, including: a processor and a memory;

[0066] The processor is connected to a memory, which stores a computer program. When the computer program is executed by the processor, it causes the computer device to perform the method provided in the embodiments of this application.

[0067] One aspect of this application provides a computer-readable storage medium storing a computer program adapted to be loaded and executed by a processor, so that a computer device having the processor performs the method provided in this application.

[0068] One embodiment of this application provides a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the method provided in this application embodiment.

[0069] In this embodiment, the server can receive business object data uploaded by a first object through a first client, and then obtain valid content corresponding to valid object behavior from historical content. The business object data may include historical content provided by the first object within the business object period, and valid object behavior is determined by the object behavior recorded by the first object for the historical content within the business object period. Further, the server can use the content attribute value corresponding to the valid content as the object attribute value of the first object, and based on the object attribute value, determine the weighted business object attribute value of the first object within the business object period, and determine the attribute tendency corresponding to the weighted business object attribute value as the first attribute tendency of the first object. Further, if the weighted business object attribute value satisfies the reverse attribute recommendation condition corresponding to the business object period, the server can obtain target recommendation content matching the second attribute tendency (i.e., the opposite attribute tendency of the first attribute tendency), and send the target recommendation content to the first client so that the first client displays the target recommendation content to the first object. Therefore, this application embodiment can obtain effective content corresponding to effective object behavior from the information flow content viewed by the first object within the business object cycle by recording the browsing of information flow content (i.e., historical content) of an object (e.g., a first object) in the information flow product. Then, based on the weighted business object attribute value indicated by the effective content, the attribute tendency of the first object after reading the effective content can be identified. In this application embodiment, the identified attribute tendency can be used as the first attribute tendency of the first object. It should be understood that after the first attribute tendency of the first object is identified, when the weighted business object attribute value meets the reverse attribute recommendation condition, target recommendation content matching the opposite attribute tendency (i.e., the second attribute tendency) can be recommended to the first object. This means that this application embodiment can actively push differentiated information flow content (e.g., target recommendation content with a positive tendency) when the first object has been browsing information flow content with a single attribute tendency (e.g., effective content with a negative tendency) for a long time, thereby providing the first object with a sense of trust in the comprehensive and sound information flow product, and thus improving the flexibility of information flow content recommendation. Attached Figure Description

[0070] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0071] Figure 1 This is a schematic diagram of a network architecture provided in an embodiment of this application;

[0072] Figure 2This is a schematic diagram of a data interaction scenario provided in an embodiment of this application;

[0073] Figure 3 This is a flowchart illustrating a data push method provided in an embodiment of this application;

[0074] Figure 4 This is a schematic diagram of a process for uploading sample content provided in an embodiment of this application;

[0075] Figure 5 This is a schematic diagram illustrating a scenario of uploading sample content provided in an embodiment of this application;

[0076] Figure 6 This is a schematic diagram illustrating a scenario for viewing sample content provided in an embodiment of this application;

[0077] Figure 7a This is a flowchart illustrating an embodiment of the present application that does not meet the automatic recommendation mode;

[0078] Figure 7b This is a flowchart illustrating an automatic recommendation mode provided in an embodiment of this application;

[0079] Figure 8a This is a flowchart illustrating a scenario where a custom recommendation mode is not satisfied, as provided in an embodiment of this application.

[0080] Figure 8b This is a flowchart illustrating a custom recommendation mode provided in an embodiment of this application;

[0081] Figure 9 This is a flowchart illustrating a data push method provided in an embodiment of this application;

[0082] Figure 10 This is a schematic diagram illustrating a scenario of authorized statistical analysis provided in an embodiment of this application;

[0083] Figure 11 This is a schematic diagram of a scenario for adjusting the emotion configuration threshold provided in an embodiment of this application;

[0084] Figure 12 This is a schematic diagram of a data interaction process provided in an embodiment of this application;

[0085] Figure 13a This is a schematic diagram illustrating a scenario of directly recommending target content, as provided in an embodiment of this application.

[0086] Figure 13b This is a schematic diagram illustrating a scenario of indirectly recommending target content provided in an embodiment of this application;

[0087] Figure 14This is a schematic diagram of the structure of a data push device provided in an embodiment of this application;

[0088] Figure 15 This is a schematic diagram of the structure of a data push device provided in an embodiment of this application;

[0089] Figure 16 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application;

[0090] Figure 17 This is a schematic diagram of the structure of a data push system provided in an embodiment of this application. Detailed Implementation

[0091] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0092] It should be understood that Artificial Intelligence (AI) is the theory, methods, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to achieve optimal results. In other words, AI is a comprehensive technology within computer science that attempts to understand the essence of intelligence and produce a new kind of intelligent machine that can react in a way similar to human intelligence. AI studies the design principles and implementation methods of various intelligent machines, enabling them to possess the functions of perception, reasoning, and decision-making.

[0093] Artificial intelligence (AI) is a comprehensive discipline encompassing a wide range of fields, including both hardware and software technologies. Fundamental AI technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies primarily include computer vision, speech processing, natural language processing, as well as machine learning / deep learning, autonomous driving, and intelligent transportation.

[0094] The solutions provided in this application mainly involve artificial intelligence technologies such as Natural Language Processing (NLP) and Machine Learning (ML).

[0095] Machine learning is a multidisciplinary field involving probability theory, statistics, approximation theory, convex analysis, and algorithm complexity theory. It specifically studies how computers can simulate or implement human learning behavior to acquire new knowledge or skills and reorganize existing knowledge structures to continuously improve their performance. Machine learning is the core of artificial intelligence and the fundamental way to endow computers with intelligence; its applications span all areas of artificial intelligence. Machine learning and deep learning typically include techniques such as artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and instructional learning.

[0096] Natural Language Processing (NLP) is an important area within computer science and artificial intelligence. It studies the theories and methods for enabling effective communication between humans and computers using natural language. NLP is a science that integrates linguistics, computer science, and mathematics. Therefore, research in this field involves natural language—the language people use in daily life—and thus it has a close connection with linguistic research. NLP technologies typically include text processing, semantic understanding, machine translation, question answering, and knowledge graphs.

[0097] For details, please see Figure 1 , Figure 1 This is a schematic diagram of a network architecture provided in an embodiment of this application. Figure 1 As shown, this network architecture may include a service server 2000 and a user terminal cluster. The user terminal cluster may specifically include one or more user terminals; the number of user terminals in the user terminal cluster is not limited here. Figure 1 As shown, the multiple user terminals may specifically include user terminal 3000a, user terminal 3000b, user terminal 3000c, ..., user terminal 3000n; user terminal 3000a, user terminal 3000b, user terminal 3000c, ..., user terminal 3000n can be directly or indirectly connected to the service server 2000 via wired or wireless communication, so that each user terminal can interact with the service server 2000 through the network connection.

[0098] Among them, the business server 2000 can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms.

[0099] Each user terminal in the user terminal cluster can include: smartphones, tablets, laptops, desktop computers, smart home devices, wearable devices, in-vehicle terminals, and other smart terminals with data push capabilities. It should be understood that, for example... Figure 1 Each user terminal in the user terminal cluster shown can be integrated with an application client. When the application client runs on each user terminal, it can interact with the aforementioned... Figure 1 The business servers 2000 shown interact with each other. Specifically, the application clients may include: in-vehicle clients, smart home clients, entertainment clients (e.g., game clients), multimedia clients (e.g., video clients), social clients, and information clients (e.g., news clients), etc.

[0100] For ease of understanding, the embodiments of this application may be described in detail below. Figure 1 From the multiple user terminals shown, one user terminal is selected as the target user terminal. For example, in the embodiments of this application, a user terminal can be selected as the target user terminal. Figure 1 The user terminal 3000a shown serves as the target user terminal, which may integrate an application client with data push functionality. In this case, the target user terminal can interact with the business server 2000 through this application client.

[0101] For ease of understanding, in this application embodiment, the information stream content (e.g., videos, text and images) recommended to a certain object (e.g., object Y1) can be collectively referred to as recommended content. In this application embodiment, the information stream content (e.g., videos, text and images) selected by an object (e.g., object Y1) in the application client that matches their interests can be collectively referred to as historical content. Historical content can include historical recommended content and historical retrieved content. Historical recommended content can be content passively received by object Y1; for example, historical recommended content can be content pushed to object Y1 by the business server 2000 based on object Y1's attribute preferences. Historical retrieved content can be content that object Y1 is interested in; for example, historical retrieved content can be content actively searched or viewed by object Y1. Here, attribute preferences can be emotional preferences.

[0102] For ease of understanding, in this application embodiment, the information stream content (e.g., video, text and images) uploaded by an object (e.g., object Y2) through an application client can be collectively referred to as sample content.

[0103] It is understood that in this application embodiment, the object that logs into the application client through first account information (e.g., account information 1) can be referred to as the first object (e.g., the aforementioned object Y1), the user terminal corresponding to the first object can be referred to as the first terminal, and the application client integrated and installed on the first terminal can be the first client. In this application embodiment, any user terminal in the aforementioned user terminal cluster can be selected as the first terminal. For example, in this application embodiment, user terminal 3000a in the aforementioned user terminal cluster can be selected as the first terminal. It should be understood that the first object in this application embodiment can be a user who receives information stream content (i.e., historical content) through the first client, i.e., a data receiver.

[0104] It is understood that in this application embodiment, the object that logs into the application client through second account information (e.g., account information 2) can be referred to as the second object (e.g., the aforementioned object Y2), and the user terminal corresponding to the second object can be referred to as the second terminal. The application client integrated and installed on the second terminal can be the second client. In this application embodiment, any user terminal in the aforementioned user terminal cluster can be selected as the second terminal. For example, in this application embodiment, user terminal 3000b in the aforementioned user terminal cluster can be used as the second terminal. It should be understood that the second object in this application embodiment can be a user who uploads information stream content (i.e., sample content) through the second client, i.e., a data uploader.

[0105] It should be understood that the first object in this application embodiment can act as both the data receiver and the data uploader. For example, the first object can become a data receiver through a first client in a first terminal, and the first object can also become a data uploader through a first client in a first terminal. Similarly, the second object in this application embodiment can act as both the data uploader and the data receiver. For example, the second object can become a data uploader through a second client in a second terminal, and the second object can also become a data receiver through a second client in a second terminal.

[0106] It should be understood that the first client in the first terminal can record the business object data (which can be viewing data) of object Y1 for the historical content provided (i.e., viewed) by the first object (e.g., object Y1 mentioned above), and then send the viewing data to the business server 2000. It is understood that the business server 2000 can obtain the historical content viewed by object Y1 within the business object period (which can be the viewing period) from the viewing data, extract valid content from this historical content, determine the weighted business object attribute value of object Y1 within the viewing period based on the content attribute value corresponding to the valid content (which can be a weighted viewing sentiment value), and then determine the sentiment tendency corresponding to the weighted viewing sentiment value as the first attribute tendency of the first object (which can be a first sentiment tendency). Therefore, when the weighted sentiment value satisfies the reverse attribute recommendation condition (the reverse attribute recommendation condition can be the reverse sentiment recommendation condition), the business server 2000 can obtain at least one target recommendation content for information flow recommendation of object Y1, wherein the second attribute tendency (the second attribute tendency can be the second sentiment tendency) associated with the at least one target recommendation content can be the opposite attribute tendency (the opposite attribute tendency can be the opposite sentiment tendency) of object Y1's first sentiment tendency during the viewing period.

[0107] It is understood that the data push method involved in the embodiments of this application can be applied to blockchain systems. Figure 1 The business server 2000 shown can be a blockchain node in a blockchain system. It should be understood that when a second client corresponding to a second object (e.g., object Y2 mentioned above) needs to write the sentiment value of sample content uploaded by the second object through the second client into the blockchain, it can send the sample content to a packaging node in the blockchain network. This allows the packaging node to perform attribute analysis (sentiment analysis) on the sample content, obtain the sentiment value of the sample content, and then broadcast the block containing the sentiment value to the consensus nodes in the blockchain network for block consensus. When these consensus nodes reach block consensus, the packaging node can then write the aforementioned block containing the sentiment value into the blockchain. It should be understood that by writing the sentiment value of the sample content into the blockchain, this embodiment allows the business server 2000 to quickly obtain the sentiment value of historical content in the reading data when it receives reading data sent by the first object (e.g., object Y1 mentioned above) through the first client. This historical content can include the sample content uploaded by object Y2.

[0108] For better understanding, please refer to [link / reference]. Figure 2 , Figure 2This is a schematic diagram illustrating a data interaction scenario provided in an embodiment of this application. For example... Figure 2 The server 20a shown can be the one described above. Figure 1 The corresponding business server 2000 in the embodiment, such as Figure 2 The user terminal 20b shown (i.e., the first terminal) can be the one described above. Figure 1 For ease of understanding, the embodiments of this application refer to any user terminal in the user terminal cluster of the corresponding embodiment. Figure 1 The user terminal 3000a shown is used as an example of the user terminal 20b to illustrate... Figure 2 The diagram illustrates the specific process of data interaction between server 20a and user terminal 20b. User terminal 20b has an application client (i.e., the first client) installed. This application client can be used to display recommended content pushed to the object corresponding to user terminal 20b, where the object corresponding to user terminal 20b can be object 20c (i.e., the first object).

[0109] like Figure 2 As shown, object 20c can execute a first trigger operation on the historical content in the recommended content of the application client of user terminal 20b. In this way, the application client can respond to the first trigger operation executed by object 20c on the historical content, record the viewing data of object 20c on the historical content based on the first trigger operation, and then send the recorded viewing data to server 20a. The historical content corresponding to the first trigger operation is the historical content viewed by object 20c.

[0110] It is understood that the application client can perform a validity check on the historical content corresponding to the first triggering operation. In this validity check, based on the viewing behavior recorded for that historical content, historical content corresponding to valid object behaviors (which can be valid viewing behaviors) is considered valid content, while historical content corresponding to invalid object behaviors (which can be invalid viewing behaviors) is considered invalid recommended content. Furthermore, the application client can generate viewing data to be sent to server 20a based on the valid and invalid viewing behaviors recorded for the historical content. Optionally, the application client may also generate viewing data to be sent to server 20a directly based on the viewing behavior recorded for the historical content, without performing a validity check on the historical content corresponding to the first triggering operation.

[0111] like Figure 2As shown, after receiving the viewing data sent by object 20c through the application client, server 20a can obtain the historical content 200a provided by object 20c within the viewing period from the viewing data, and then obtain the valid content 200b corresponding to the valid viewing behavior from the historical content 200a. The historical content 200a can include multiple historical content items, specifically including: historical content 3a, historical content 3b, historical content 3c, historical content 3d, ..., historical content 3m. The valid content 200b can include multiple valid content items; taking four valid content items as an example, these four valid content items can specifically include: historical content 3a (i.e., valid content 3a), historical content 3b (i.e., valid content 3b), historical content 3c (i.e., valid content 3c), and historical content 3d (i.e., valid content 3d).

[0112] It is understood that when the browsing data sent by the application client includes valid and invalid browsing behaviors corresponding to historical content, server 20a can directly obtain the valid content 200b corresponding to the valid browsing behaviors from historical content 200a. Optionally, when the browsing data sent by the application client does not include valid and invalid browsing behaviors corresponding to historical content, server 20a can perform validity judgment on the historical content in historical content 200a. In the validity judgment, based on the browsing behaviors recorded for these historical contents, these historical contents are divided into historical content corresponding to valid browsing behaviors and invalid recommended content corresponding to invalid browsing behaviors, and then the valid content corresponding to the valid browsing behaviors (i.e., Figure 2 The valid content shown is 200b).

[0113] like Figure 2 As shown, server 20a can obtain the content attribute value (which can be a reading sentiment value) corresponding to each valid piece of content in valid content 200b from the content database (e.g., content database 200c). For example, the content attribute value corresponding to valid content 3a can be content attribute value 31a, the content attribute value corresponding to valid content 3b can be content attribute value 31b, the content attribute value corresponding to valid content 3c can be content attribute value 31c, and the content attribute value corresponding to valid content 3d can be content attribute value 31d. In this embodiment, content attribute values ​​31a, 31b, 31c, and 31d can be collectively referred to as the object attribute value of object 20c (which can be a user reading sentiment value).

[0114] It is understandable that, such as Figure 2 The content database 200c shown may include multiple databases, which may specifically include... Figure 2The databases 30a, 30b, ..., 30n are shown. This means that content database 200c can be used to store data to be recommended with viewing sentiment values. For example, database 30a can be used to store data to be recommended corresponding to viewing sentiment value Z1, database 30b can be used to store data to be recommended corresponding to viewing sentiment value Z2, ..., database 30n can be used to store data to be recommended corresponding to viewing sentiment value Z3.

[0115] like Figure 2 As shown, server 20a can determine the weighted business object attribute value (i.e., weighted reading sentiment value) of object 20c within the reading period based on content attribute values ​​31a, 31b, 31c, and 31d. Specifically, server 20a can determine the sentiment tendency corresponding to the weighted reading sentiment value as the first attribute tendency (i.e., the first sentiment tendency), and the opposite sentiment tendency of the first sentiment tendency can be the second attribute tendency (i.e., the second sentiment tendency). Further, server 20a can obtain target recommended content matching the second sentiment tendency from content database 200c. Here, the reading sentiment value of the target recommended content can be the reverse attribute value (here, the reverse attribute value can be the reverse sentiment value), which is determined based on the weighted reading sentiment value. In other words, server 20a can obtain content to be recommended with a reverse sentiment value from content database 200c and use the obtained content to be recommended as the target recommended content matching the second sentiment tendency.

[0116] like Figure 2 As shown, after obtaining the target recommendation content to be recommended to the application client of the user terminal 20b, the server 20a can return the target recommendation content to the application client so that the target recommendation content can be displayed to the object 20c in the application client.

[0117] Therefore, this embodiment of the application can determine in real time the weighted sentiment value indicated by the effective content in the historical content viewed by the first object, and identify the emotional state (i.e., the first sentiment tendency) of the first object after viewing the effective content based on the sentiment value. Then, it can proactively push differentiated content to the first object. This differentiated content is the target recommended content that matches the opposite sentiment tendency (i.e., the second sentiment tendency) of the first sentiment tendency. Clearly, by pushing target recommended content (e.g., positive-tense information flow content) to the first object, it can prevent the first object from viewing content with a single sentiment (e.g., negative-tense information flow content) for a long time, thereby reasonably guiding and adjusting the first object's sentiment tendency, and thus improving the flexibility of information flow content recommendation.

[0118] The specific implementation method for data interaction between the application client in server 20a and user terminal 20b can be found in the following... Figures 3-13b The corresponding embodiment describes the data interaction between the server and the first client.

[0119] Further, please see Figure 3 , Figure 3 This is a flowchart illustrating a data push method provided in an embodiment of this application. The method can be executed by a server, by a first client, or by both a server and a first client. The server can be one of the aforementioned... Figure 2 The corresponding server 20a in the implementation can be the first client as described above. Figure 2 The corresponding application client in this embodiment. For ease of understanding, this embodiment uses the method being executed by a server as an example for explanation. The data push method may include the following steps S101-S104:

[0120] Step S101: Receive business object data uploaded by the first object through the first client;

[0121] It is understood that business object data can include historical content provided by the first object within a historical time period. The server can obtain the historical content corresponding to the business object period (i.e., the historical content provided by the first object within the viewing period) from the historical content corresponding to the historical time period. Thus, the viewing data can include historical content provided by the first object outside the viewing period. Optionally, the viewing data can include historical content provided by the first object within the viewing period, thus the viewing data may not include historical content provided by the first object outside the viewing period.

[0122] Step S102: Obtain the valid content corresponding to the valid object behavior from the historical content;

[0123] It is understandable that the historical content provided by the first object within the viewing period may include valid content corresponding to valid object behaviors and invalid recommended content corresponding to invalid object behaviors (here, invalid object behaviors can be invalid viewing behaviors). Valid and invalid viewing behaviors are determined by the object behaviors (here, object behaviors can be viewing behaviors) recorded by the first object regarding the historical content within the business object period. In other words, valid viewing behaviors are determined by the viewing behaviors recorded by the first object regarding valid content within the viewing period, and invalid viewing behaviors are determined by the viewing behaviors recorded by the first object regarding invalid recommended content within the viewing period.

[0124] Understandably, the server can determine the validity of historical content based on the first object's recorded viewing behavior within the viewing period, obtain behavioral analysis results, and then extract valid content corresponding to valid viewing behavior from the historical content based on the behavioral analysis results. Optionally, it is also understandable that the first client can determine the validity of historical content based on the first object's recorded viewing behavior, obtain behavioral analysis results, and then divide the historical content into valid content and invalid recommended content based on the behavioral analysis results. The client then uploads the viewing data carrying the valid content and invalid recommended content to the server, allowing the server to directly extract the valid content corresponding to valid viewing behavior from the historical content of the viewing data.

[0125] Historical content may include, but is not limited to, text and image content and video content. It is understood that if historical content includes text and image content, then the business object data also includes the object behavior corresponding to the text and image content. This object behavior includes the content duration corresponding to the first object (i.e., the viewing duration of the text and image content). Therefore, the specific process of the server judging the validity of the text and image content can be described as follows: The server can obtain the number of words in the text and image content for the first object within the content duration (the number of words can be the number of words viewed), and determine the content speed of the first object for the text and image content based on the content duration and the number of words (the content speed can be the viewing speed). Further, the server can obtain the content type of the text and image content and determine the dynamic adjustment coefficient corresponding to the text and image content based on the content type. Further, the server can obtain the base speed associated with the text and image content, adjust the base speed based on the dynamic adjustment coefficient, use the adjusted base speed as the dynamic speed, and compare the content speed with the dynamic speed to obtain the speed comparison result. Furthermore, if the speed comparison result indicates that the content speed is less than the dynamic speed, the server can determine that the object behavior corresponding to the text and image content is a valid object behavior, and will select the text and image content that corresponds to the valid object behavior from the text and image content as valid content.

[0126] It is understood that the server can obtain the minimum speed associated with the text and image content. If the speed comparison result indicates that the reading speed is less than the dynamic speed and the reading speed is greater than the minimum speed, the server can determine that the reading behavior corresponding to the text and image content is a valid reading behavior, and will select the text and image content corresponding to the valid reading behavior as valid content. In other words, if the speed comparison result indicates that the reading speed is less than the dynamic speed, the server can determine the text and image content corresponding to the reading speed less than the dynamic speed as intermediate recommended content, and then obtain the minimum speed associated with the text and image content. If the speed comparison result indicates that the reading speed is greater than the minimum speed, the server can determine the intermediate recommended content corresponding to the reading speed greater than the minimum speed as valid content. It should be understood that the embodiments of this application do not limit the specific value of the minimum speed.

[0127] It is understood that the dynamic adjustment coefficient can be used to represent the overall reading difficulty of the text and image content. The higher the overall reading difficulty of the text and image content, the smaller the dynamic adjustment coefficient; the lower the overall reading difficulty of the text and image content, the larger the dynamic adjustment coefficient. The dynamic adjustment coefficient can be determined by the classification and tag information of the text and image content. For example, when the content type of the text and image content is classical Chinese, the overall reading difficulty corresponding to this type is high, requiring weighting based on the first dynamic coefficient on the base speed. Here, the first dynamic coefficient can be a dynamic adjustment coefficient less than 1, for example, 0.5. As another example, when the content type of the text and image content is story-based, the overall reading difficulty corresponding to this type is low, requiring weighting based on the second dynamic coefficient on the base speed. Here, the second dynamic coefficient can be a dynamic adjustment coefficient greater than 1, for example, 1.5. Therefore, by dynamically weighting changes for different content types of text and image content, the rationality of the dynamic speed can be dynamically ensured. It should be understood that the embodiments of this application do not limit the specific value of the dynamic adjustment coefficient.

[0128] It is understood that the base speed represents a basic numerical value for effective reading. This basic value is used to represent a common reading speed standard, which in turn represents the basic number of words read within a basic time period. The relationship between base speed, basic number of words, and basic time period can be expressed as: Base speed = Basic number of words / Basic time period. For example, when the basic time period is 1 minute (i.e., 60 seconds) and the basic number of words is 300, the base speed can be determined to be 5 (i.e., 5 words per second). It should be understood that the embodiments of this application do not limit the specific value of the base speed. Similarly, the formula for determining the reading speed based on the reading time and the number of words read can be expressed as: Reading speed = Number of words read / Reading time period.

[0129] Optionally, the reading behavior corresponding to the text and image content also includes the number of times the first object (i.e., reading the text and image content) has been viewed and the completion rate of the text and image content. The number of views can represent the number of clicks made by the first object on the text and image content. The completion rate is determined by the number of words read by the first object within the reading time and the total number of words in the text and image content. Optionally, it is understood that the server can obtain a frequency threshold associated with the text and image content, compare the number of views with the frequency threshold to obtain a first comparison result. If the first comparison result indicates that the number of views is greater than the frequency threshold, then the reading behavior corresponding to the text and image content is determined to be a valid reading behavior. Optionally, it is understood that the server can obtain a completion threshold associated with the text and image content, compare the completion rate of the text and image content with the completion rate threshold to obtain a second comparison result. If the second comparison result indicates that the completion rate is greater than the completion rate threshold, then the reading behavior corresponding to the text and image content is determined to be a valid reading behavior.

[0130] Optionally, and understandably, the server can also determine the valid viewing value of the text and image content based on the number of times the content is viewed and the completion rate. Then, it can obtain a valid threshold associated with the content and compare the valid viewing value with the threshold. If the valid viewing value is greater than the threshold, the viewing behavior is determined to be a valid viewing behavior. Optionally, and understandably, the server can also filter valid content corresponding to valid viewing behaviors from the text and image content based on speed comparison results, a first comparison result, and a second comparison result.

[0131] It is understandable that if historical content includes video content, then the viewing data also includes the viewing behavior corresponding to the video content. This viewing behavior includes the completion rate of the first user's viewing of the video content and the number of times the video content was viewed. Therefore, the server can determine the validity of video content based on its completion rate and the number of times it was viewed. It should be understood that the specific process of the server determining the validity of video content based on its completion rate and the number of times it was viewed can be found in the description above regarding the determination of the validity of text and image content based on its viewing count and completion rate; it will not be repeated here.

[0132] It is understandable that the historical content refers to the content to be recommended in the content database corresponding to the first client. This content database contains sample content uploaded by the second object through the second client. The specific process of the server storing the sample content can be described as follows: The server (i.e., the information flow backend) can obtain the content details of the sample content, perform attribute analysis (i.e., sentiment judgment) on the content details, and obtain the first attribute value (which can be the first sentiment value) and the second attribute value (which can be the second sentiment value) of the content details. Based on the first and second attribute values, the server determines the content attribute value of the sample content. The server can then add the first and second sentiment values ​​to obtain the viewing sentiment value of the sample content. Furthermore, the server can store the content details, content attribute values, first attribute value, and second attribute value as content to be recommended in the content database.

[0133] When the sample content is text and image content, the content details may include, but are not limited to, the subject information (i.e., title), content text (i.e., body text), category information, and tag information. It is understood that the tag information may include first tag information and second tag information. The first tag information may be the tag information entered by the second object in the second client, and the second tag information may be the tag information obtained by the server after performing tag analysis on the content text.

[0134] Optionally, when the sample content is video content, the content details may include, but are not limited to, the subject information, classification information, tag information, voice information, subtitle information, and audio-visual information of the sample content. Specifically, the voice information can be recognized by the server using ASR (Automatic Speech Recognition), and the subtitle information can be recognized by the server using OCR (Optical Character Recognition).

[0135] Specifically, based on the sentiment value of the sample content, the sentiment tendency corresponding to the sample content can be determined. If the sentiment tendency corresponding to the sentiment value of the sample content is positive, then the sentiment tendency corresponding to the sample content is positive; if the sentiment tendency corresponding to the sentiment value of the sample content is negative, then the sentiment tendency corresponding to the sample content is negative.

[0136] Optionally, the server can also obtain the object identifier of the second object corresponding to the sample content, and store the object identifier of the second object together with the content details of the sample content in the content database.

[0137] For easier understanding, please refer to Figure 4, Figure 4 This is a schematic diagram illustrating a process for uploading sample content provided in an embodiment of this application. For example... Figure 4 The client shown can be a second client for uploading sample content, and this second client can be integrated into the above. Figure 1 Any user terminal in the user terminal cluster shown, for example, user terminal 3000b; Figure 4 The content producer shown can be a second object corresponding to the second client, for example, object Y2; such as Figure 4 The server shown can be a server that receives sample content, and this server can be one of the aforementioned... Figure 1 The business server shown is 2000. Optionally, the server side can also be understood as the server program deployed on the server.

[0138] like Figure 4 As shown, the content producer can execute step S41 to edit content (i.e., sample content) on the client. Then, the client can execute step S42 to display the edited sample content on its screen. Further, the client can execute step S43 to upload the edited sample content to the server. Upon receiving the sample content uploaded by the content producer through the client, the server can execute step S44 to perform sentiment value calculation (i.e., sentiment analysis) on the received sample content, obtaining a first sentiment value, a second sentiment value, and a viewing sentiment value. These first sentiment value, second sentiment value, and viewing sentiment value can be collectively referred to as the sentiment value parameters of the sample content. The server can store these sentiment value parameters in the content database. Specifically, after receiving the sample content, the server can obtain content detail information from the sample content and perform sentiment value calculation on this information to achieve the sentiment value calculation for the sample content.

[0139] like Figure 4 As shown, after the server calculates the sentiment value parameters of the sample content using sentiment values, it can execute step S45 to return the sentiment value parameters to the client, so that the client can display the sentiment value parameters of the sample content (i.e., viewing sentiment value, first sentiment value, and second sentiment value) on the display screen, and display the sentiment value parameter details. In this way, the content producer can execute step S46 to view the parameter details in the client. The parameter details can be used to display the first matching word associated with the first sentiment value and the second matching word associated with the second sentiment value in the content details information.

[0140] Understandably, the first and second matching word segments can be used to determine the first and second sentiment values ​​of the sample content, respectively. Therefore, content creators can edit the sample content again on the client side, that is, modify the first or second matching word segments in the sample content, so that the server can recalculate the sentiment value of the modified sample content to obtain a new first sentiment value and a new second sentiment value, thereby achieving the purpose of modifying the sentiment value parameters.

[0141] For ease of understanding, the specific process of the content producer (i.e., the second object) editing the sample content in the client (i.e., the second client) can be found in [reference needed]. Figure 5 , Figure 5 This is a schematic diagram illustrating a scenario for uploading sample content provided in an embodiment of this application. For example... Figure 5 The object 50a shown can be the one described above. Figure 4 The content producer in the corresponding embodiment; such as Figure 5 The first upload interface 50b and the second upload interface 50c shown can be Figure 4 The content upload interface of the client at different times in the corresponding embodiment.

[0142] like Figure 5 As shown, object 50a can perform a first input operation on the first upload interface 50b to input the theme information and content text of the sample content into the first upload interface 50b, for example, theme information 5a and content text 5b, where theme information 5a can be "Let's eat zongzi together!". Further, object 50a can perform an interface switching operation (e.g., a swipe operation) on the first upload interface 50b to switch the client to a second upload interface 50c, and then perform a second input operation on the second upload interface 50c to input the category information and tag information of the sample content into the second upload interface 50c, for example, category information 5c and tag information 5d, where category information 5c can be "movie" and tag information 5d can be "zongzi".

[0143] Furthermore, object 50a can perform a content publishing operation on the second upload interface 50c, uploading the topic information 5a, content text 5b, category information 5c, and tag information 5d as content details of sample content (e.g., sample content H) to the server. For example, as Figure 5 As shown, the second upload interface 50c may include a content upload control 50d, and the object 50a may perform a content publishing operation on the content upload control 50d.

[0144] In this context, it is understandable that when object 50a uploads topic information 5a, content text 5b, classification information 5c, and tag information 5d as sample content to the server, it needs to agree to relevant rights (i.e., standardized protocols) so that the server can perform sentiment analysis on the sample content uploaded by object 50a. In other words, the sentiment tendency of the content text (i.e., sample content) will be systematically evaluated (i.e., sentiment analysis) and uploaded to the content library (i.e., content database).

[0145] It should be understood that the first input operation, the second input operation, the interface switching operation, and the content publishing operation may include contact operations such as clicking, long pressing, and swiping, or non-contact operations such as voice and gestures. This application does not limit these operations.

[0146] For ease of understanding, the specific process by which the content producer (i.e., the second object) views parameter details on the client (i.e., the second client) can be found in [reference needed]. Figure 6 , Figure 6 This is a schematic diagram illustrating a scenario for viewing sample content provided in an embodiment of this application. For example... Figure 6 The object 60a shown can be the one described above. Figure 4 The content producer in the corresponding embodiment; such as Figure 6 The parameter details interfaces 6a and 6b shown can be... Figure 4 The parameter details interface in the client of the corresponding embodiment.

[0147] It is understood that object 60a can perform a display trigger operation (e.g., a click operation) on the emotion display control 60b in the client to output parameter details interface 6a in the client. This parameter details interface 6a may include sample parameters of the sample content uploaded by the second object (e.g., the sample content H mentioned above), such as... Figure 6 As shown, the sample parameters here may include the sentiment value corresponding to the overall sentiment tendency (i.e., the viewing sentiment value), the sentiment value corresponding to the negative sentiment tendency (e.g., the first sentiment value), the sentiment value corresponding to the positive sentiment tendency (e.g., the second sentiment value), the first detail control 60d corresponding to the negative sentiment tendency, the second detail control 60c corresponding to the positive sentiment tendency, and the graphical display corresponding to the overall sentiment tendency. It can be understood that object 60a can perform a hide trigger operation (e.g., a click operation) on the sentiment display control 60b to hide the sample parameters of the sample content H in the parameter details interface 6a on the client side.

[0148] like Figure 6As shown, object 60a can perform a details viewing operation on the second details control 60c. In this way, the client can respond to the details viewing operation performed by object 60a by outputting the content text of sample content H and the matching word corresponding to the positive tendency (e.g., the second matching word) in the client, thus switching the parameter details interface 6a to the parameter details interface 6b. For example, the second matching word here can be the second matching word 60e, which can be "AAAA". Similarly, object 60a can perform a details viewing operation on the first details control 60d to output the content text of sample content H and the matching word corresponding to the negative tendency (e.g., the first matching word) in the client. The display form of the first or second matching word in sample content H can be a selected form, a bold form, or a standalone display form; this embodiment does not limit the display form.

[0149] like Figure 6 As shown, the parameter details interface 6b may include a content modification control 60f. Object 60a can perform content modification operations on this content modification control 60f to further edit the sample content H on the client side. It can be understood that by providing content modification functionality, object 60a can further edit and adjust extreme text in the sample content H, thereby adjusting the reading sentiment value, first sentiment value, and second sentiment value corresponding to the sample content H.

[0150] It should be understood that the specific process of the server performing sentiment analysis on content details information can be described as follows: The server can obtain the content text from the content details information of the sample content and the attribute analysis strategy associated with the sample content (here, the attribute analysis strategy can be a sentiment analysis strategy, i.e., a sentiment calculation algorithm). The attribute analysis strategy includes an attribute dictionary analysis strategy (here, the attribute dictionary analysis strategy can be a sentiment dictionary analysis strategy). Further, the server can perform word segmentation on the content text based on the attribute dictionary analysis strategy to obtain the first text segment of the content text. Stop word processing is then applied to the first text segment, and the stop word processed first text segment is used as the second text segment of the content text. Further, the server can obtain the first word dictionary and the second word dictionary associated with the content text based on the attribute dictionary analysis strategy. The second text segment is then matched with the first word dictionary and the second word dictionary respectively to obtain the string matching results. Further, the server can determine the first attribute value and the second attribute value of the content details information based on the attribute matching values ​​in the string matching results (here, the attribute matching values ​​can be sentiment matching values). The first attribute value and the second attribute value are then summed to obtain the content attribute value of the sample content.

[0151] It can be understood that the word segmentation process is to recombine a continuous sequence of characters (i.e., the content text) into a sequence of words (i.e., the first text word segmentation) according to certain specifications. For example, when the content text is: "I am a student", the server can perform word segmentation on the content text, and the first text word segmentation of the content text can be: "I", "am", "a", "student". It should be understood that the word segmentation algorithm used in the word segmentation process can be a word segmentation method based on string matching, a word segmentation method based on understanding, or a word segmentation method based on statistics. The embodiments of the present application do not limit the word segmentation algorithm used in the word segmentation process.

[0152] It can be understood that stop words can be understood as words with high occurrence frequencies and little practical significance. This type of words mainly includes modal particles, adverbs, prepositions, conjunctions, etc. For example, when the first text word segmentation is: "I", "am", "a", "student", the server can perform stop word processing (i.e., remove stop words) on the first text word segmentation, and the second text word segmentation of the content text can be: "a", "student". It should be understood that stop words can be stored in a stop word dictionary. The embodiments of the present application do not limit the stop word dictionary used in the stop word processing.

[0153] Optionally, the server can also directly perform word segmentation on the content detail information to obtain the first text word segmentation of the content detail information, instead of obtaining the content text from the content detail information and then performing word segmentation on the content text.

[0154] The attribute dictionary analysis strategy is used to instruct the acquisition of a first word dictionary associated with the first attribute tendency and a second word dictionary associated with the second attribute tendency. It should be understood that the specific process of the server performing string matching based on the second text segmentation and the sentiment dictionary (which can include both the first and second word dictionaries) can be described as follows: When the server acquires the first word dictionary associated with the content text based on the attribute dictionary analysis strategy, it searches for strings matching the second text segmentation in the first word dictionary. If a string matching the second text segmentation is found in the first word dictionary, it is used as the first matching segmentation corresponding to the second text segmentation. Further, the server can acquire the first weight information corresponding to the first matching segmentation, perform a second accumulation process on the first weight information, and obtain the first matching value associated with the first attribute tendency. Further, when the server acquires the second word dictionary associated with the content text based on the attribute dictionary analysis strategy, it searches for strings matching the second text segmentation in the second word dictionary. If a string matching the second text segmentation is found in the second word dictionary, it is used as the second matching segmentation corresponding to the second text segmentation. Furthermore, the server can obtain the second weight information corresponding to the second matched word segment, perform a second accumulation process on the second weight information, and obtain a second matching value associated with the second attribute tendency. Furthermore, the server can obtain the string matching result based on the first matching value and the second matching value.

[0155] It should be understood that the sentiment dictionary in this application embodiment can be the CNKI sentiment dictionary or the BosonNLP sentiment dictionary, and this application embodiment does not limit the specific type of sentiment dictionary. The sentiment dictionary may also include a negation word dictionary and a degree adverb dictionary.

[0156] Optionally, if no string matching the second text segment is found in the first word dictionary, the server does not obtain the first matching value associated with the first sentiment tendency, and directly obtains the string matching result based on the second matching value. Similarly, optionally, if no string matching the second text segment is found in the second word dictionary, the server does not obtain the second matching value associated with the second sentiment tendency, and directly obtains the string matching result based on the first matching value.

[0157] Specifically, when the first sentiment tendency is negative, the first matching value is negative (i.e., the first weight information is negative), and the first word dictionary is a negative word dictionary; when the second sentiment tendency is positive, the second matching value is positive (i.e., the second weight information is positive), and the second word dictionary is a positive word dictionary. It can be understood that the sentiment tendency corresponding to the sample content is determined by the first and second matching values. When the absolute value of the first matching value is greater than the absolute value of the second matching value, the sentiment tendency corresponding to the sample content is associated with the first matching value, that is, the sentiment tendency corresponding to the sample content is the first sentiment tendency; when the absolute value of the first matching value is less than the absolute value of the second matching value, the sentiment tendency corresponding to the sample content is associated with the second matching value, that is, the sentiment tendency corresponding to the sample content is the second sentiment tendency.

[0158] It is understandable that the sample word segments in the sentiment dictionary (i.e., the first sample word segment in the first word dictionary and the second sample word segment in the second word dictionary) can have different weight information. The server can use the weight information corresponding to the first sample word as the first weight information of the first matching word, and the weight information corresponding to the second sample word as the second weight information of the second matching word. The weight information can be used to represent the degree to which a sample word belongs to its corresponding sentiment tendency. For example, the first word dictionary may include first sample word F1 and first sample word F2. If the degree to which first sample word F1 belongs to the first sentiment tendency is greater than the degree to which first sample word F2 belongs to the first sentiment tendency, then the absolute value of the first weight information Q1 corresponding to first sample word F1 is greater than the absolute value of the first weight information Q2 corresponding to first sample word F2.

[0159] It can be understood that the second accumulation process can perform addition operations on each first weight information in the first weight information, or on each second weight information in the second weight information. For example, when there are two first matching words, these two first matching words can be the first sample word F1 (i.e., the first matching word F1) and the first sample word F2 (i.e., the first matching word F2) mentioned above, and the first matching value corresponding to the first matching word can be equal to the first weight information Q1 plus the first weight information Q2.

[0160] Optionally, and understandably, the server may also disregard the first weight information of the first matching word and the second weight information of the second matching word, and instead determine the first matching value based on the number of the first matching words, or the second matching value based on the number of the second matching words. In other words, when the first word dictionary is a negative word dictionary and the second word dictionary is a positive word dictionary, the server can treat the first weight information as -1 and the second weight information as +1. For example, the first matching value corresponding to the first matching words F1 and F2 mentioned above can be equal to -2.

[0161] It should be understood that sentiment analysis strategies include model learning analysis strategies. Therefore, the specific process by which the server performs sentiment analysis on content details based on a model learning analysis strategy can be described as follows: The server can obtain a target network model associated with the content text based on the model learning analysis strategy. This target network model is obtained by training an initial network model based on training text content and sample labels. Further, the server can input the content text into the target network model, and the target network model extracts features from the content text to obtain the text attribute features corresponding to the content text (here, text attribute features can be text sentiment features). Further, the server can use the classifier in the target network model to determine the matching degree between the text attribute features and the sample attribute features (here, sample attribute features can be sample sentiment features) in the classifier (i.e., the classifier in the target network model), and based on the matching degree, determine a first predicted value associated with a first attribute tendency and a second predicted value associated with a second attribute tendency. Further, the server can update the string matching results based on the first and second predicted values.

[0162] It can be understood that there can be multiple sample sentiment features, specifically two, which can be a first sample sentiment feature and a second sample sentiment feature. The matching degree between the text sentiment feature and the first sample sentiment feature can be called the first matching degree, and the matching degree between the text sentiment feature and the second sample sentiment feature can be called the second matching degree. Based on the first matching degree, a first predicted value associated with the first sentiment tendency can be determined, and based on the second matching degree, a second predicted value associated with the second sentiment tendency can be determined. For example, if the first matching degree is 0.7 and the second matching degree is 0.3, then the first predicted value can be -7 and the second predicted value can be +3.

[0163] It is understood that the feature extraction methods in the target network model can be Information Gain (IG), Chi-square (CHI), and Document Frequency (DF), etc. It should be understood that the embodiments of this application do not limit the specific type of feature extraction method.

[0164] It is understood that the classification method used by the classifier (i.e., the sentiment classifier) ​​in the target network model can be the center vector classification method, the K-Nearest-Neighbor (KNN) classification method, the Bayesian classifier, the support vector machine, the conditional random field, the maximum entropy classifier, etc. It should be understood that the embodiments of this application do not limit the specific type of classification method.

[0165] Understandably, the server can directly obtain the string matching result based on the first matching value and the second matching value. In this case, the sentiment matching value in the string matching result is the first matching value and the second matching value. Optionally, the server can also directly obtain the string matching result based on the first predicted value and the second predicted value. In this case, the sentiment matching value in the string matching result is the first predicted value and the second predicted value. Optionally, the server can also obtain the string matching result based on the first matching value, the second matching value, the first predicted value, and the second predicted value. In this case, the sentiment matching value in the string matching result can be determined by the first matching value, the second matching value, the first predicted value, and the second predicted value.

[0166] For ease of understanding, this embodiment uses the sentiment matching values ​​in the string matching results—first and second matching values—as examples to illustrate how to obtain the first sentiment value based on the first matching value and how to obtain the second sentiment value based on the second matching value. It is understood that the server can determine the negative threshold (e.g., -10) corresponding to the first sentiment tendency and the positive threshold (e.g., +10) corresponding to the second sentiment tendency. Because the first matching value can be less than the negative threshold or the second matching value can be greater than the positive threshold, the server can use a sentiment value mapping algorithm to map the first and second matching values ​​to the range between the negative and positive thresholds, respectively.

[0167] The specific execution process of the sentiment value mapping algorithm can be as follows: The server can obtain the minimum first matching value and the maximum second matching value from the content to be recommended in the content database, and compare the absolute value of the minimum first matching value with the absolute value of the maximum second matching value to obtain an absolute value comparison result. Further, if the absolute value comparison result indicates that the absolute value of the minimum first matching value is greater than the absolute value of the maximum second matching value, the server can map the minimum first matching value to a negative threshold. Further, the server can obtain the first sentiment value corresponding to the first matching value based on the negative threshold, the minimum first matching value, and the first matching value; the server can obtain the second sentiment value corresponding to the second matching value based on the negative threshold, the minimum first matching value, and the second matching value. It should be understood that the embodiments of this application do not limit the specific algorithm type of the sentiment value mapping algorithm.

[0168] Optionally, if the absolute value comparison result indicates that the absolute value of the smallest first matching value is less than or equal to the absolute value of the largest second matching value, the server can map the largest second matching value to a positive threshold. Further, the server can obtain a first sentiment value corresponding to the first matching value based on the positive threshold, the largest second matching value, and the first matching value; the server can obtain a second sentiment value corresponding to the second matching value based on the positive threshold, the largest second matching value, and the second matching value.

[0169] It should be understood that the specific process by which the server trains the initial network model to obtain the target network model can be described as follows: The server can obtain training text content and sample labels for the training text content. The sample labels may include a first sample label corresponding to a first sentiment tendency and a second sample label corresponding to a second sentiment tendency. Further, the server can input the training text content into the initial network model, and extract features from the training text content through the initial network model to obtain training sentiment features corresponding to the training text content. Further, the server can iteratively train the initial network model based on the sample labels, training sentiment features, and the classifier in the initial network model, and use the iteratively trained initial network model as the target network model.

[0170] Step S103: Take the content attribute value corresponding to the valid content as the object attribute value of the first object, determine the weighted business object attribute value of the first object in the business object cycle based on the object attribute value, and determine the attribute tendency corresponding to the weighted business object attribute value as the first attribute tendency of the first object.

[0171] Specifically, the server can use the content attribute value corresponding to the valid content as the object attribute value of the first object, and perform a first accumulation process on the object attribute value to obtain the accumulated content attribute value (here, the accumulated content attribute value can be the accumulated viewing sentiment value). Further, the server can obtain the content quantity of the valid content (here, the content quantity can be the number of views), and determine the weighted business object attribute value of the first object within the business object's lifecycle based on the accumulated content attribute value and the content quantity. The weighted business object attribute value is obtained by dividing the accumulated content attribute value by the content quantity. Further, the server can determine the attribute tendency corresponding to the weighted business object attribute value as the first attribute tendency of the first object.

[0172] The server's first accumulation of user viewing sentiment values ​​can be understood as performing an addition operation on each user viewing sentiment value within the user viewing sentiment value set. Optionally, the specific process of the server performing the first accumulation of user viewing sentiment values ​​to obtain the accumulated viewing sentiment value can also be described as follows: If the user viewing sentiment value includes a first object sentiment value associated with a first sentiment tendency, the server can obtain the first object sentiment value associated with the first sentiment tendency from the user viewing sentiment value set, and perform an addition operation on each first object sentiment value in the first object sentiment value set to obtain the first accumulated sentiment value. Further, if the user viewing sentiment value includes a second object sentiment value associated with a second sentiment tendency, the server can obtain the second object sentiment value associated with the second sentiment tendency from the user viewing sentiment value set, and perform an addition operation on each second object sentiment value in the second object sentiment value set to obtain the second accumulated sentiment value. Further, the server can obtain the accumulated viewing sentiment value based on the first accumulated sentiment value and the second accumulated sentiment value (e.g., performing an addition operation on the first accumulated sentiment value and the second accumulated sentiment value).

[0173] It is understandable that when the weighted sentiment value is negative, the corresponding sentiment tendency is negative. In this case, the server can determine the negative tendency as the first sentiment tendency of the first object. Optionally, when the weighted sentiment value is positive, the corresponding sentiment tendency is positive. In this case, the server can determine the positive tendency as the first sentiment tendency of the first object. For ease of understanding, this application embodiment uses a negative sentiment tendency corresponding to the weighted sentiment value as an example for explanation.

[0174] It is understood that the reverse attribute recommendation conditions in this application embodiment can be used to determine the execution path of the server in automatic recommendation mode and custom recommendation mode. If the weighted reading sentiment value meets the reverse attribute recommendation conditions, the server can execute the following step S104. Optionally, if the weighted reading sentiment value does not meet the reverse attribute recommendation conditions, the server can distribute recommended content to the first object based on the original recommendation algorithm.

[0175] In automatic recommendation mode, the server can obtain reverse attribute recommendation conditions corresponding to the business object's time period. These reverse attribute recommendation conditions include a first attribute threshold (which can be a first sentiment threshold) associated with a first attribute tendency and a second attribute threshold (which can be a second sentiment threshold) associated with a second attribute tendency. Further, when the first attribute threshold is less than the second attribute threshold, the server can compare the weighted business object attribute value with both the first and second attribute thresholds to obtain a threshold comparison result. Further, if the threshold comparison result indicates that the weighted business object attribute value is less than the first attribute threshold or greater than the second attribute threshold, the server can determine that the weighted business object attribute value meets the reverse attribute recommendation conditions.

[0176] Optionally, if the threshold comparison result indicates that the weighted reading sentiment value is greater than or equal to the first sentiment threshold and the weighted reading sentiment value is less than or equal to the second sentiment threshold, the server can determine that the weighted reading sentiment value does not meet the reverse sentiment recommendation conditions, obtain the associated recommendation content that matches the first sentiment tendency, and send the associated recommendation content to the first client.

[0177] Similarly, optionally, when the first sentiment threshold is greater than the second sentiment threshold, the server can compare the weighted reading sentiment value with the first and second sentiment thresholds to obtain a threshold comparison result. Further, if the threshold comparison result indicates that the weighted reading sentiment value is greater than the first sentiment threshold or less than the second sentiment threshold, the server can determine that the weighted reading sentiment value meets the reverse sentiment recommendation conditions. It should be understood that this application embodiment does not limit the magnitude of the first and second sentiment thresholds; for ease of understanding, this application embodiment uses the example of the first sentiment threshold being less than the second sentiment threshold for illustration.

[0178] Understandably, the server can use the average of the first and second sentiment thresholds as the sentiment balance threshold in the automatic recommendation mode. Therefore, the reverse sentiment recommendation conditions can also include the sentiment balance threshold associated with the first object. The first and second sentiment thresholds are opposites of each other, and the sentiment balance threshold can be equal to zero (i.e., 0).

[0179] Optionally, in the custom recommendation mode, the server can obtain the reverse attribute recommendation conditions corresponding to the business object's lifecycle. These reverse sentiment recommendation conditions include an attribute configuration threshold associated with the first object (which can be a sentiment configuration threshold) and an attribute fluctuation value associated with the attribute configuration threshold (which can be a sentiment fluctuation value). Further, the server can determine a first absolute difference between the weighted business object attribute value and the attribute configuration threshold based on the sentiment configuration threshold and the sentiment fluctuation value. Further, if the first absolute difference is greater than the attribute fluctuation value, the server can determine that the weighted business object attribute value meets the reverse attribute recommendation conditions corresponding to the business object's lifecycle.

[0180] Optionally, if the first absolute difference is less than or equal to the sentiment fluctuation value, the server can determine that the weighted reading sentiment value does not meet the reverse sentiment recommendation conditions corresponding to the reading period, and obtain the associated recommendation content that matches the first sentiment tendency, and send the associated recommendation content to the first client.

[0181] Understandably, if the weighted reading sentiment value is greater than the sentiment configuration threshold, then the first absolute difference is equal to the weighted reading sentiment value minus the sentiment configuration threshold; alternatively, if the weighted reading sentiment value is less than the sentiment configuration threshold, then the first absolute difference is equal to the sentiment configuration threshold minus the weighted reading sentiment value.

[0182] Step S104: If the weighted business object attribute value meets the reverse attribute recommendation condition corresponding to the business object period, then obtain the target recommendation content that matches the second attribute tendency and send the target recommendation content to the first client.

[0183] Specifically, if the weighted business object attribute value meets the reverse attribute recommendation condition corresponding to the business object's cycle, the server can obtain the attribute balance value (which can be the sentiment balance value) from the reverse attribute recommendation condition, and determine the reverse attribute value corresponding to the weighted business object attribute value based on the attribute balance value. The second absolute difference between the weighted business object attribute value and the attribute balance value is equal to the second absolute difference between the reverse attribute value and the attribute balance value. Further, the server can use the content to be recommended with the reverse attribute value as the target recommended content matching the second attribute tendency, and send the target recommended content to the first client. The second attribute tendency is the opposite of the first attribute tendency.

[0184] When the first client receives the target recommended content returned by the server, it can display the target recommended content to the first object in the first client, that is, output the target recommended content in the first client so that the first object can perform a viewing operation on the target recommended content. At this time, the target recommended content can also be called recommended content.

[0185] Specifically, when the recommendation mode corresponding to the first client is automatic recommendation mode, the emotional balance value can be an emotional balance threshold. Optionally, when the recommendation mode corresponding to the first client is custom recommendation mode, the emotional balance value can be an emotional configuration threshold. It should be understood that the absolute value of the first emotional threshold or the absolute value of the second emotional threshold in automatic recommendation mode can be equal to the emotional fluctuation value. This application embodiment does not limit the specific values ​​of the first emotional threshold, the second emotional threshold, and the emotional fluctuation value.

[0186] Understandably, the second absolute difference can represent the absolute value of the difference between the weighted sentiment value and the sentiment balance value, or the absolute value of the difference between the reverse sentiment value and the sentiment balance value. When the weighted sentiment value is less than the sentiment balance value, the reverse sentiment value is greater than the sentiment balance value; conversely, when the weighted sentiment value is greater than the sentiment balance value, the reverse sentiment value is less than the sentiment balance value. For example, the sentiment balance value can be equal to N, the weighted sentiment value can be equal to (N-3), and the reverse sentiment value can be equal to (N+3). Or, for another example, the sentiment balance value can be equal to N, the weighted sentiment value can be equal to (N+3), and the reverse sentiment value can be equal to (N-3).

[0187] Optionally, if the weighted sentiment value does not meet the reverse sentiment recommendation conditions corresponding to the viewing period, the server can obtain related recommended content that matches the first sentiment tendency and send the related recommended content to the first client.

[0188] For ease of understanding, the execution flow in automatic recommendation mode can be found in [link to documentation]. Figure 7a , Figure 7a This is a flowchart illustrating a scenario where the automatic recommendation mode is not satisfied, as provided in an embodiment of this application. Figure 7a The client shown can be a first client for sending viewing data, and this first client can be integrated into a first terminal, which can be the aforementioned... Figure 1 Any user terminal in the user terminal cluster shown, for example, user terminal 3000a; Figure 7a The object shown can be the first object corresponding to the first client, for example, object Y1; such as Figure 7a The server shown can be a server that receives viewing data, and this server can be the one described above. Figure 1 The business server shown is 2000.

[0189] like Figure 7aAs shown, the object can execute step S51, performing effective content reading on the client side. The historical content corresponding to the effective content reading is considered effective content. Optionally, the object can also perform invalid content reading on the client side. The historical content corresponding to the invalid content reading is considered invalid recommended content. In this way, the client can execute step S52, recording the object's reading activity and identifying the sentiment value received by the object; that is, the client can obtain the reading sentiment value corresponding to the historical content viewed from the server. Further, the client can execute step S53, uploading the object data (i.e., the reading data) to the server.

[0190] It is understood that the client can determine the validity of the viewed content; therefore, the viewed data can include valid content. Alternatively, it is understood that the server can determine the validity of the viewed content; therefore, the viewed data can include the viewed content.

[0191] like Figure 7a As shown, after receiving the object data uploaded by the client, the server can record the object's periodic data, that is, obtain the valid content within the viewing period from the object data uploaded by the client and determine the weighted viewing sentiment value associated with the valid content. Further, the server can execute step S54. In step S54, if the absolute value of the weighted viewing sentiment value does not exceed the threshold (i.e., the weighted viewing sentiment value does not meet the reverse sentiment recommendation condition), then the server continues to recommend content according to the original recommendation method (i.e., recommend content according to the original recommendation logic, i.e., recommend related recommended content), so that the client executes step S55, presenting the content of the recommended object (i.e., the related recommended content) on the client's display screen. Here, the threshold can be the absolute value of a first sentiment threshold or the absolute value of a second sentiment threshold, where the absolute value of the first sentiment threshold is equal to the absolute value of the second sentiment threshold.

[0192] For ease of understanding, the execution flow in automatic recommendation mode can be found in [link to documentation]. Figure 7b , Figure 7b This is a flowchart illustrating an automatic recommendation mode provided in an embodiment of this application. Wherein, Figure 7b The object shown can be Figure 7a The object shown, Figure 7b The client shown can be Figure 7a The client shown Figure 7b The server shown can be Figure 7a The server shown.

[0193] like Figure 7b For the specific process of steps S61-S63 shown above, please refer to the above. Figure 7aThe descriptions of steps S51-S53 in the corresponding embodiments will not be repeated here. It is understood that the server can execute step S64. In step S64, if the absolute value of the weighted sentiment value exceeds a threshold (i.e., the weighted sentiment value meets the reverse sentiment recommendation condition), it begins to search for content in the content library (i.e., the content database) for recommendation (i.e., recommending target content), so that the client executes step S65, displaying the content of the reverse sentiment value of the recommended object (i.e., the target recommended content) on the client's display screen.

[0194] In this context, it is understandable that an object can select target recommended content that interests it on the client's display screen. After the object effectively views the selected target recommended content, the weighted viewing sentiment value associated with the object will change. Therefore, when the absolute value of the new weighted viewing sentiment value does not exceed the threshold, the server can return the recommendation logic to the original content recommendation logic.

[0195] For easier understanding, the execution flow in the custom recommendation mode can be found in [link to documentation]. Figure 8a , Figure 8a This is a flowchart illustrating a scenario where a custom recommendation mode is not satisfied, as provided in an embodiment of this application. Wherein, Figure 8a The object shown can be Figure 7a The object shown, Figure 8a The client shown can be Figure 7a The client shown Figure 8a The server shown can be Figure 7a The server shown.

[0196] like Figure 8a As shown, the object can execute step S71, setting an emotion value (i.e., emotion configuration threshold) on the client and performing effective content reading. For example, the emotion configuration threshold can be N, where N can be equal to 2. Then, the client can execute step S72, recording the emotion value of the content viewed by the object in real time, and then execute step S73, uploading the object data to the server. The specific process of the client executing steps S72-S73 can be found above. Figure 7a The descriptions of steps S52-S53 in the corresponding embodiments will not be repeated here.

[0197] like Figure 8aAs shown, after receiving the object data uploaded by the client, the server can determine the weighted sentiment value associated with the valid content and proceed to step S74. In step S74, if the weighted sentiment value meets the sentiment configuration threshold set by the object (i.e., the weighted sentiment value does not meet the reverse sentiment recommendation condition), the content recommendation will continue to be performed according to the original recommendation method (i.e., recommend related recommended content), so that the client can proceed to step S75 and display the content of the recommended object (i.e., related recommended content) on the client's display screen.

[0198] Specifically, it can be understood that the weighted reading sentiment value conforming to the sentiment configuration threshold set by the object can be interpreted as: the first absolute difference between the weighted reading sentiment value and the sentiment configuration threshold is less than or equal to the sentiment fluctuation value. The sentiment fluctuation value can represent the sentiment fluctuation value that fluctuates above or below the sentiment configuration threshold. The weighted reading sentiment value conforming to the sentiment configuration threshold set by the object can also be understood as: the weighted reading sentiment value being greater than or equal to the sentiment configuration threshold and fluctuating downwards, and the weighted reading sentiment value being less than or equal to the sentiment configuration threshold and fluctuating upwards.

[0199] For easier understanding, the execution flow in the custom recommendation mode can be found in [link to documentation]. Figure 8b , Figure 8b This is a flowchart illustrating a customized recommendation mode provided in an embodiment of this application. Wherein, Figure 8a The object shown can be Figure 7a The object shown, Figure 8b The client shown can be Figure 7a The client shown Figure 8b The server shown can be Figure 7a The server shown.

[0200] like Figure 8b For the specific process of steps S81-S83 shown above, please refer to the above. Figure 8a The descriptions of steps S71-S73 in the corresponding embodiments will not be repeated here. It is understood that the server can execute step S84. In step S84, if the weighted sentiment value does not meet the sentiment configuration threshold set for the object (i.e., the weighted sentiment value meets the reverse sentiment recommendation condition), then it begins to search for content in the content library (i.e., the content database) for recommendation (i.e., recommending the target content), so that the client executes step S85, displaying the content of the reverse sentiment value of the recommended object (i.e., the target recommended content) on the client's display screen.

[0201] In this context, it is understandable that an object can select target recommended content that interests it on the client's display screen. After the object effectively views the selected target recommended content, the sentiment value of the viewed content will change. Therefore, when the new sentiment value of the viewed content indicates that the weighted sentiment value meets the sentiment configuration threshold set by the object, the server can return the recommendation logic to the original content recommendation logic.

[0202] Therefore, this application embodiment can obtain effective content corresponding to effective object behavior from the information flow content viewed by the first object within the business object cycle by recording the browsing of information flow content (i.e., historical content) of an object (e.g., a first object) in the information flow product. Then, based on the weighted business object attribute value indicated by the effective content, the attribute tendency of the first object after reading the effective content can be identified. In this application embodiment, the identified attribute tendency can be used as the first attribute tendency of the first object. It should be understood that after the first attribute tendency of the first object is identified, when the weighted business object attribute value meets the reverse sentiment recommendation condition, target recommendation content matching the opposite attribute tendency (i.e., the second attribute tendency) can be recommended to the first object. This means that this application embodiment can actively push differentiated information flow content (e.g., target recommendation content with a positive tendency) when the first object has been browsing information flow content with a single attribute tendency (e.g., effective content with a negative tendency) for a long time, thereby providing the first object with a sense of trust in the comprehensive and sound information flow product, and thus improving the flexibility of information flow content recommendation.

[0203] Further, please see Figure 9 , Figure 9 This is a flowchart illustrating a data push method provided in an embodiment of this application. The method can be executed by a server, by a first client, or by both a server and a first client. The server can be one of the aforementioned... Figure 2 The corresponding server 20a in the implementation can be the first client as described above. Figure 2 The corresponding application client in this embodiment. For ease of understanding, this embodiment will be described using the example of the method being jointly executed by the server and the first client. The data push method may include the following steps S201-S207:

[0204] Step S201: In response to the first trigger operation performed by the first object on the historical content in the first client, record the business object data of the first object on the historical content based on the first trigger operation, and send the recorded business object data to the server.

[0205] Among them, historical content includes valid content corresponding to valid object behavior, which is determined by the object behavior recorded by the first object in relation to historical content within the business object cycle.

[0206] It is understandable that the number of historical contents corresponding to the first triggering operation can be multiple, and the triggering operations performed by the first object on multiple historical contents can be collectively referred to as the first triggering operation.

[0207] Understandably, before the first client records the first object's viewing data of historical content based on the first trigger operation, the first object needs to authorize the first client to have the permission to record the first object's viewing data of historical content. The viewing data can include the historical content provided by the first object within the viewing period, as well as the viewing behavior recorded by the first object regarding the historical content within the viewing period. The historical content here can include, but is not limited to, text and image content and video content, and the viewing behavior here can include the viewing duration and time of the first object's viewing of historical content (e.g., text and image content, video content).

[0208] For easier understanding, please refer to Figure 10 , Figure 10 This is a schematic diagram illustrating a scenario of authorized statistical analysis provided in an embodiment of this application. For example... Figure 10 The object 10a shown can be the first object corresponding to the first client, for example, object Y1; as shown Figure 10 The client display interface 100a shown can be the interface in the first terminal corresponding to object 10a. The client display interface 100a of the first terminal can include multiple application clients (i.e., multiple application clients can be integrated and installed in the first terminal), and these multiple application clients can include a first client 10b. The server corresponding to this first client 10b can be the one described above. Figure 1 The business server shown is 2000.

[0209] like Figure 10As shown, object 10a can perform a client launch operation (e.g., a click operation) on the first client 10b in the client display interface 100a. This allows the first terminal to open the first client 10b and output the content recommendation interface 100b within it. It is understood that when the client launch operation performed by object 10a on the first client 10b is the first time object 10a launches the first client 10b, the first client 10b can output a sub-interface independent of the content recommendation interface 100b. This sub-interface may include permission prompt information, such as permission prompt information 100c, which could be something like, "To help you with more comprehensive content recommendations, please allow statistical analysis of your reading content and reading behavior. The analysis results are only visible to yourself."

[0210] like Figure 10 As shown, the sub-interface may also include an allow control and a deny control. When object 10a performs an allow operation on the allow control, it can authorize the first client 10b; when object 10a performs a deny operation on the deny control, it can choose not to authorize the first client 10b.

[0211] It should be understood that client-side initiation, permission, and denial operations can include contact operations such as clicking, long-pressing, and swiping, as well as non-contact operations such as voice and gestures; this application does not limit these operations.

[0212] Step S202: Receive business object data uploaded by the first object through the first client;

[0213] It should be understood that the unit corresponding to the reading period in the embodiments of this application can be seconds, minutes, hours, days, weeks, months, years, etc., and the embodiments of this application do not limit the unit here. For example, the reading period can be 3 hours, or the reading period can be 7 days.

[0214] Step S203: Obtain the valid content corresponding to the valid object behavior from the historical content;

[0215] Step S204: Take the content attribute value corresponding to the valid content as the object attribute value of the first object, determine the weighted business object attribute value of the first object in the business object cycle based on the object attribute value, and determine the attribute tendency corresponding to the weighted business object attribute value as the first attribute tendency of the first object.

[0216] It should be understood that this application embodiment uses the example of two categories of attribute tendencies corresponding to the weighted reading sentiment value for illustration. These two types of sentiment tendencies may include a first sentiment tendency and a second sentiment tendency, which are opposite to each other. Optionally, the sentiment tendency corresponding to the weighted reading sentiment value may also be no sentiment tendency (i.e., no tendency). For example, when the weighted reading sentiment value is equal to zero (i.e., 0), the sentiment tendency corresponding to the weighted reading sentiment value is no sentiment tendency.

[0217] It should be understood that in the embodiments of this application, the first emotional tendency can be a positive tendency, and when the first emotional tendency is a positive tendency, the second emotional tendency is a negative tendency; conversely, in the embodiments of this application, the first emotional tendency can be a negative tendency, and when the first emotional tendency is a negative tendency, the second emotional tendency is a positive tendency. For ease of understanding, the embodiments of this application will be described using the example of a negative emotional tendency as the first emotional tendency.

[0218] Step S205: If the weighted business object attribute value meets the reverse attribute recommendation condition corresponding to the business object cycle, then obtain the target recommendation content that matches the second attribute tendency and send the target recommendation content to the first client.

[0219] The number of target recommended content items can be one or more. When the server obtains multiple target recommended content items, the server can send multiple target recommended content items to the first client at once, or it can send multiple target recommended content items to the first client multiple times. This application does not limit this.

[0220] Optionally, if the weighted sentiment value satisfies the reverse sentiment recommendation conditions corresponding to the viewing period, the server can incorporate sentiment value intervention into the recommendation process, i.e., increase the weight of the sentiment value recommendation mode. For example, in the recommendation mode without sentiment value intervention, the server can retrieve interest-based recommended content matching the first object from the content database based on the object profile of the first object. Alternatively, in the recommendation mode with sentiment value intervention, the server sets the weight of the sentiment value recommendation mode to the sentiment recommendation weight (e.g., 50%), so the weight of the object profile recommendation mode can be the profile recommendation weight (e.g., 50%).

[0221] Therefore, the server can retrieve interest-based and target-based recommended content matching the first object from the content database based on the object profile and second sentiment tendency. The target-based recommended content is selected from the content database based on sentiment recommendation weights, while the interest-based recommended content is selected based on the profile recommendation weights. This allows for more diverse emotional content recommendations to the first object from a comprehensive and healthy information perspective. The meta-sentimental content here can include the object profile and second sentiment tendency; the object profile can be determined by hobbies, personality, cognition, worldview, etc. Thus, this embodiment of the application can prevent the object from excessively viewing content with negative emotions, allowing the object to obtain more comprehensive content from an emotional perspective.

[0222] In this embodiment of the application, when the server sets the aforementioned sentiment recommendation weight to 100%, if the weighted reading sentiment value meets the reverse sentiment recommendation condition corresponding to the reading period, the server can obtain target recommended content that matches the first object. That is, this embodiment of the application uses a sentiment recommendation weight of 100% as an example for explanation. It should be understood that this embodiment of the application does not limit the specific value of the sentiment recommendation weight.

[0223] Understandably, if the weighted sentiment value satisfies the reverse sentiment recommendation conditions corresponding to the reading period, the server can determine the reverse sentiment value corresponding to the weighted sentiment value based on the sentiment balance value in the reverse sentiment recommendation conditions. Then, the content to be recommended with the reverse sentiment value is used as the target recommendation content matching the second sentiment tendency, and the target recommendation content is sent to the first client. Specifically, when the first sentiment tendency is positive, the second sentiment tendency is negative; when the first sentiment tendency is negative, the second sentiment tendency is positive.

[0224] Optionally, the weighted sentiment value can also have the same sentiment tendency as the negative sentiment value; that is, the weighted sentiment value, sentiment balance value, and negative sentiment value all have a primary sentiment tendency. Specifically, when the weighted sentiment value, sentiment balance value, and negative sentiment value are all negative, the primary sentiment tendency can be negative; when the weighted sentiment value, sentiment balance value, and negative sentiment value are all positive, the primary sentiment tendency can be positive. Furthermore, the server can use the content to be recommended with the negative sentiment value as the target recommended content that matches the primary sentiment tendency, and then send the target recommended content to the first client.

[0225] Understandably, in the custom recommendation mode, the first object can adjust the sentiment configuration threshold associated with it within the first client. The specific process of adjusting the sentiment configuration threshold can be described as follows: the first client can obtain the historical viewing information interface associated with it. This historical viewing information interface displays the first object's weighted viewing sentiment value, historical content, and threshold adjustment controls within the viewing period. Further, in response to a second triggering operation performed by the first object on the threshold adjustment controls, the first client can output a strategy adjustment interface. Further still, in response to a third triggering operation performed by the first object on the strategy adjustment interface, the first client can send the sentiment configuration threshold corresponding to the third triggering operation to the server, so that the server can determine the first absolute difference between the weighted viewing sentiment value and the sentiment configuration threshold based on the sentiment configuration threshold and the sentiment fluctuation value associated with it.

[0226] For easier understanding, please refer to Figure 11 , Figure 11 This is a schematic diagram illustrating a scenario for adjusting an emotion configuration threshold, as provided in an embodiment of this application. Figure 11 The object 11a shown can be Figure 10 The object 10a shown is... Figure 11 The historical viewing information interface 110a shown can be the historical viewing information interface in the first client, which can be... Figure 10 The first client 10b is shown. The policy adjustment interface 110b and policy adjustment interface 110c can be the policy adjustment interfaces of the first client at different times.

[0227] like Figure 11 As shown, the historical viewing information interface 110a may include: the weighted viewing sentiment value of object 11a in the viewing period (i.e., the sentiment value corresponding to the overall sentiment tendency of the content), historical content, threshold adjustment control (e.g., threshold adjustment control 11b), first accumulated sentiment value (i.e., the sentiment value corresponding to the negative tendency), second accumulated sentiment value (i.e., the sentiment value corresponding to the positive tendency), first sentiment value corresponding to historical content, second sentiment value corresponding to historical content, and period selection control associated with the viewing period (e.g., the period selection control corresponding to "within 1 week").

[0228] The number of historical content items can be multiple, specifically X items, where X can be a positive integer. These X items can include, but are not limited to, historical content X1, historical content X2, historical content X3, and historical content X4. For example, the first sentiment value corresponding to historical content X1 can be -5, and the second sentiment value corresponding to historical content X1 can be +2.

[0229] Among them, Figure 11 In the historical viewing information interface 110a shown, object 11a can periodically view the sentiment value parameters of each piece of historical content it has viewed, as well as the sentiment value parameters of all historical content, thereby making a judgment on the sentiment of the content it receives. In addition, object 11a can also personalize the sentiment value logic of subsequent content recommendations based on individual circumstances (i.e., adjust the sentiment configuration threshold).

[0230] like Figure 11 As shown, object 11a can perform a second trigger operation on threshold adjustment control 11b. In response, the first client can output a strategy adjustment interface 110b. The strategy adjustment interface 110b may include an emotion configuration control, such as emotion configuration control 11c. Furthermore, object 11a can perform a third trigger operation on emotion configuration control 11c. In response, the first client can switch the strategy adjustment interface 110b to the strategy adjustment interface 110c, and adjust the emotion configuration control 11c from its first position in strategy adjustment interface 110b to its second position in strategy adjustment interface 110c. The first and second positions can be collectively referred to as the emotion configuration position associated with the first object. Further, the first client can use the emotion configuration threshold corresponding to the second position as the emotion configuration threshold associated with object 11a, and then send the emotion configuration threshold to the server corresponding to the first client. For example, the emotional configuration threshold corresponding to the first position can be 0, and the emotional configuration threshold corresponding to the second position can be +5.

[0231] Optionally, the strategy adjustment interface 110b may include an emotion configuration input box. Object 11a may perform a third trigger operation (e.g., an input operation) on the emotion configuration input box. In this way, the first client may respond to the third trigger operation performed by object 11a on the emotion configuration input box, obtain the emotion configuration threshold entered by object 11a in the emotion configuration input box, and then send the emotion configuration threshold to the server.

[0232] like Figure 11As shown, the strategy adjustment interface of the first client may also include sentiment prompts. These sentiment prompts may include negative, balanced, and positive prompts. For example, a negative prompt may be "Negative," a balanced prompt may be "Balanced," and a positive prompt may be "Positive." It is understood that when the sentiment configuration position is between a negative and a balanced prompt, the sentiment configuration threshold indicated by that position is associated with a first sentiment tendency; when the sentiment configuration position is between a positive and a balanced prompt, the sentiment configuration threshold indicated by that position is associated with a second sentiment tendency.

[0233] Step S206: Receive target recommended content associated with valid content sent by the server based on business object data;

[0234] It should be understood that the data interaction process between the first object, the second object, and the server in the embodiments of this application can be found in [reference needed]. Figure 12 , Figure 12 This is a schematic diagram illustrating a data interaction process provided in an embodiment of this application. For example... Figure 12 The content producer shown may include, but is not limited to, the second object corresponding to the second client, for example, object Y2; such as Figure 12 The content consumer shown may include, but is not limited to, the first object corresponding to the first client, for example, object Y1; such as Figure 12 The content library platform and data platform shown can be collectively referred to as the content platform, which can be deployed on a server.

[0235] like Figure 12 As shown, content producers (e.g., the second object) can upload text content (i.e., sample content) to the server (i.e., the content platform) through a second client. Upon receiving the sample content, the server can standardize its storage based on the content library platform. This standardization can include content standardization and sentiment standardization. Content standardization can include basic information about the sample content, such as the author (the object identifier of the second object), title (i.e., topic information), body text (i.e., content text), category (i.e., category information), and tags (i.e., tag information). Sentiment standardization can include the negative value (i.e., the first sentiment value), the positive value (i.e., the second sentiment value), and the overall sentiment value (i.e., the viewing sentiment value) of the sample content. Furthermore, after receiving the sample content, the server also needs to conduct sample review.

[0236] like Figure 12As shown, the server can obtain recommended content from the content library platform for distribution to content consumers (e.g., the first object), and distribute the recommended content to the first object. Different content consumers receive different recommended content. Thus, the first object can choose whether to consume the content distributed by the server (the content consumed is the historical content), and then send object data back based on its viewed historical content, i.e., send viewing data to the server through the first client.

[0237] like Figure 12 As shown, after receiving viewing data uploaded by a content consumer (e.g., the first object) through the first client, the server can perform consumption data analysis on the viewing data based on the data platform to store the resulting object's consumption data. This object's consumption data can include the first object's consumption data. Specifically, the first object's consumption data can include the first object's consumption patterns for different types and different sentiment values. The different types are determined by the classification and tags of historical content, and the sentiment values ​​are determined by the negative, positive, and overall sentiment values ​​of historical content.

[0238] like Figure 12 As shown, the server can retrieve consumption data associated with sample content (e.g., sample content H) uploaded by a content producer (e.g., the second object) from the object's consumption data, in order to display to the second object the consumption status of other objects for the sample content H. This consumption status may include, but is not limited to, the number of views, likes, favorites, and comments. These other objects can be one or more content consumers, and these one or more objects may include the first object.

[0239] Step S207: Display the target recommended content to the first object in the first client.

[0240] In this context, the first object can select target recommended content that interests it from the first client. For example, the target recommended content that the first object is interested in could be target recommended content A. It is understood that when the first object selects target recommended content A from the first client (i.e., the first object performs the first trigger operation for target recommended content A), the first client can send viewing data for target recommended content A to the server based on target recommended content A (at this time, target recommended content A can be understood as historical content A).

[0241] This means that the server can obtain new valid content corresponding to valid reading behavior within the new reading period from the updated historical content, and then determine the new weighted reading sentiment value of the first object within the new reading period based on the reading sentiment value corresponding to the new valid content. Further, if the new weighted reading sentiment value satisfies the reverse sentiment recommendation conditions corresponding to the new reading period, the server can obtain new target recommendation content that matches the new second sentiment tendency corresponding to the new first sentiment tendency of the new weighted reading sentiment value. Optionally, if the new weighted reading sentiment value does not satisfy the reverse sentiment recommendation conditions corresponding to the new reading period, the server can obtain new associated recommendation content that matches the new first sentiment tendency and send the new associated recommendation content to the first client. Here, the new first sentiment tendency can be the same as or different from the first sentiment tendency; this application does not impose any restrictions on this.

[0242] It should be understood that the reading period in the embodiments of this application can be understood as the first reading period, and the new reading period in the embodiments of this application can be understood as the second reading period. The second reading period can be the next reading period after the first reading period. The server can determine the reading period to which the historical content belongs based on the reading time corresponding to the historical content. For example, the historical content viewed by the first object in the first client may include historical content A1, historical content A2, historical content A3, and historical content A4. The server can determine historical content A1, historical content A2, and historical content A3 as historical content within the first reading period, and the server can determine historical content A2, historical content A3, and historical content A4 as historical content within the second reading period.

[0243] Optionally, and understandably, when the first object does not select the target recommended content sent by the server in the first client, the first object's first sentiment tendency does not change. The server can obtain another new target recommended content that matches the second sentiment tendency after a certain time interval, and then send the new target recommended content to the first client.

[0244] In this context, the first client can present the target recommended content to the first target in various exposure scenarios, for example, Figure 13a The corresponding scenarios for directly recommending target content and Figure 13b The corresponding indirect recommendation target content recommendation scenario.

[0245] For easier understanding, please refer to Figure 13a , Figure 13a This is a schematic diagram illustrating a scenario of directly recommending target content, as provided in an embodiment of this application. Wherein, Figure 13a The object 13a shown can be Figure 10 The object 10a shown is... Figure 13a The content recommendation interfaces 130a and 130b shown can be the content recommendation interfaces in the first client corresponding to object 10a. This first client can be... Figure 10 The first client 10b is shown. The content recommendation interface 130a and content recommendation interface 130b can be the content recommendation interfaces of the first client at different times.

[0246] The content recommendation interface 130a can include multiple recommended contents. Here, we take an example where the content recommendation interface 130a includes three recommended contents, all of which are historical content viewed by object 13a. These three historical contents can include: historical content A1, historical content A2, and historical content A3 (not shown in the diagram). For example... Figure 13a As shown, object 13a can perform a first recommendation operation (e.g., a swipe operation) on the content recommendation interface 130a. In this way, the first client can respond to the first recommendation operation, output the target recommended content 13b in the application client, and switch the content recommendation interface 130a to the content recommendation interface 130b. The target recommended content 13b may include target recommended content A4 and target recommended content A5.

[0247] It is understandable that historical content A1, historical content A2, and historical content A3 can be the information flow content corresponding to the first sentiment tendency, while target recommended content A4 and target recommended content A5 can be the information flow content corresponding to the second sentiment tendency.

[0248] For easier understanding, please refer to Figure 13b , Figure 13b This is a schematic diagram illustrating a scenario of indirectly recommending target content, as provided in an embodiment of this application. Wherein, Figure 13b The object 13a shown is Figure 13a The object 13a shown is... Figure 13b The content recommendation interface shown is 130a. Figure 13a The content recommendation interface shown is 130a.

[0249] like Figure 13b As shown, object 13a can perform a second recommendation operation on historical content A1 in the content recommendation interface 130a. The first client can then respond to this second recommendation operation (e.g., by clicking). At this time, the first client can output the topic information and content text of historical content A1, and can also output target recommended content 13c, switching the content recommendation interface 130a to content recommendation interface 130c. The target recommended content 13c may include target recommended content A6.

[0250] It is understandable that historical content A1 and historical content A2 can be the information flow content corresponding to the first sentiment tendency, and the target recommended content A6 can be the information flow content corresponding to the second sentiment tendency.

[0251] It should be understood that Figure 13a The first recommended operation in the corresponding embodiment and Figure 13b The second recommended operation in the corresponding embodiment may include contact operations such as clicking, long pressing, and swiping, or non-contact operations such as voice and gestures. This application does not limit the scope of the operation.

[0252] Therefore, this application embodiment can statistically analyze the absolute value and changes of the reading sentiment of an object (e.g., a first object) in real time, and then adjust the content recommendation based on the first object's emotional state regarding the content it reads. Specifically, information flow content can be recommended to the first object based on the absolute value of the reading sentiment, and the recommended information flow content can be continuously adjusted based on changes in reading sentiment. It is understood that recommending different information flow content to the first object at different times ensures that the sentiment value seen by the first object within the reading period remains balanced, that is, ensuring that the weighted reading sentiment value indicated by the effective content provided to the first object within the reading period remains balanced. Furthermore, the reverse recommendation algorithm used in this application embodiment can determine the authenticity and validity of the first object's reading sentiment within a specific time period, thereby determining the first object's reading sentiment based on effective content, and thus achieving efficient reverse content recommendation. It is understood that this reverse content recommendation can reasonably guide and adjust the first object's emotional state regarding the content it reads, thereby preventing the first object's emotions from becoming too extreme, enhancing the object's trust in the comprehensive and sound information flow product, and thus improving the flexibility of information flow content recommendation.

[0253] Further, please see Figure 14 , Figure 14 This is a schematic diagram of the structure of a data push device provided in an embodiment of this application. The data push device 10 may include: a data receiving module 101, a content acquisition module 102, an attribute determination module 103, and a content sending module 104; further, the data push device 10 may also include: a condition acquisition module 105, a threshold comparison module 106, a first determination module 107, a difference determination module 108, a second determination module 109, an attribute analysis module 110, and a content storage module 111;

[0254] The data receiving module 101 is used to receive business object data uploaded by the first object through the first client; the business object data includes historical content provided by the first object during the business object period;

[0255] The content acquisition module 102 is used to acquire valid content corresponding to valid object behaviors from historical content; the valid object behaviors are determined by the object behaviors recorded by the first object for historical content within the business object cycle;

[0256] If the historical content includes text and images, then the business object data also includes the object behavior corresponding to the text and images; the object behavior corresponding to the text and images includes the content duration corresponding to the first object.

[0257] The content acquisition module 102 includes: a speed determination unit 1021, a coefficient determination unit 1022, a speed comparison unit 1023, and a content filtering unit 1024;

[0258] The speed determination unit 1021 is used to obtain the number of words in the content of the first object for the text and image content within the content duration, and determine the content speed of the first object for the text and image content based on the content duration and the number of words in the content.

[0259] The coefficient determination unit 1022 is used to obtain the content type of the graphic content and determine the dynamic adjustment coefficient corresponding to the graphic content based on the content type of the graphic content.

[0260] The speed comparison unit 1023 is used to obtain the basic speed associated with the text and image content, adjust the basic speed based on the dynamic adjustment coefficient, use the adjusted basic speed as the dynamic speed, compare the content speed with the dynamic speed, and obtain the speed comparison result.

[0261] The content filtering unit 1024 is used to determine that the object behavior corresponding to the text and image content is a valid object behavior if the speed comparison result indicates that the content speed is less than the dynamic speed, and to select the text and image content that corresponds to the valid object behavior from the text and image content as valid content.

[0262] The specific implementation methods of the speed determination unit 1021, the coefficient determination unit 1022, the speed comparison unit 1023, and the content filtering unit 1024 can be found in the above description. Figure 3 The description of step S102 in the corresponding embodiments will not be repeated here.

[0263] The attribute determination module 103 is used to take the content attribute value corresponding to the valid content as the object attribute value of the first object, determine the weighted business object attribute value of the first object in the business object cycle based on the object attribute value, and determine the attribute tendency corresponding to the weighted business object attribute value as the first attribute tendency of the first object.

[0264] The attribute determination module 103 includes: a first processing unit 1031, a second processing unit 1032, and a tendency determination unit 1033;

[0265] The first processing unit 1031 is used to take the content attribute value corresponding to the valid content as the object attribute value of the first object, perform a first accumulation process on the object attribute value, and obtain the accumulated content attribute value.

[0266] The second processing unit 1032 is used to obtain the content quantity of valid content and determine the weighted business object attribute value of the first object within the business object cycle based on the accumulated content attribute value and the content quantity.

[0267] The tendency determination unit 1033 is used to determine the attribute tendency corresponding to the weighted business object attribute value as the first attribute tendency of the first object.

[0268] The specific implementation methods of the first processing unit 1031, the second processing unit 1032, and the tendency determination unit 1033 can be found in the above description. Figure 3 The description of step S103 in the corresponding embodiments will not be repeated here.

[0269] The content sending module 104 is used to obtain target recommended content that matches the second attribute tendency if the weighted business object attribute value meets the reverse attribute recommendation condition corresponding to the business object period, and send the target recommended content to the first client so that the first client can display the target recommended content to the first object; the second attribute tendency is the opposite attribute tendency of the first attribute tendency.

[0270] The content sending module 104 includes: a reverse determination unit 1041 and a content matching unit 1042;

[0271] The reverse determination unit 1041 is used to obtain the attribute balance value in the reverse attribute recommendation condition if the weighted business object attribute value meets the reverse attribute recommendation condition corresponding to the business object period, and determine the reverse attribute value corresponding to the weighted business object attribute value based on the attribute balance value; the second absolute difference between the weighted business object attribute value and the attribute balance value is equal to the second absolute difference between the reverse attribute value and the attribute balance value.

[0272] The content matching unit 1042 is used to send the content to be recommended with the reverse attribute value as the target recommended content that matches the second attribute tendency to the first client.

[0273] The specific implementation methods of the reverse determination unit 1041 and the content matching unit 1042 can be found in the above description. Figure 3 The description of step S104 in the corresponding embodiment will not be repeated here.

[0274] Optionally, the condition acquisition module 105 is used to acquire the reverse attribute recommendation conditions corresponding to the business object cycle; the reverse attribute recommendation conditions include a first attribute threshold associated with a first attribute tendency and a second attribute threshold associated with a second attribute tendency.

[0275] The threshold comparison module 106 is used to compare the weighted business object attribute value with the first attribute threshold and the second attribute threshold when the first attribute threshold is less than the second attribute threshold, and obtain the threshold comparison result.

[0276] The first determining module 107 is used to determine that the weighted business object attribute value meets the reverse attribute recommendation condition if the threshold comparison result indicates that the weighted business object attribute value is less than the first attribute threshold or the weighted business object attribute value is greater than the second attribute threshold.

[0277] Optionally, the difference determination module 108 is used to obtain the attribute configuration threshold associated with the first object and the attribute fluctuation value associated with the attribute configuration threshold, and to determine the first absolute difference between the weighted business object attribute value and the attribute configuration threshold;

[0278] The second determining module 109 is used to determine that if the first absolute difference is greater than the attribute fluctuation value, the weighted business object attribute value satisfies the reverse attribute recommendation condition corresponding to the business object period.

[0279] Optionally, the historical content is the content to be recommended in the content database corresponding to the first client; the content database contains sample content uploaded by the second object through the second client;

[0280] The attribute analysis module 110 is used to obtain the content details information of the sample content, perform attribute analysis on the content details information to obtain the first attribute value and the second attribute value of the content details information, and determine the content attribute value of the sample content based on the first attribute value and the second attribute value.

[0281] The attribute analysis module 110 includes: a strategy acquisition unit 1101, a word segmentation determination unit 1102, a result determination unit 1103, and a summation processing unit 1104; optionally, the attribute analysis module 110 may further include: a model acquisition unit 1105, a feature extraction unit 1106, a numerical prediction unit 1107, and a result update unit 1108.

[0282] The strategy acquisition unit 1101 is used to acquire the content text in the content details information of the sample content and the attribute analysis strategy associated with the sample content; the attribute analysis strategy includes the attribute dictionary analysis strategy.

[0283] The word segmentation determination unit 1102 is used to perform word segmentation on the content text based on the attribute dictionary analysis strategy, obtain the first text segment of the content text, perform stop word processing on the first text segment, and use the first text segment after stop word processing as the second text segment of the content text.

[0284] The result determination unit 1103 is used to obtain a first word dictionary and a second word dictionary associated with the content text based on the attribute dictionary analysis strategy, and to perform string matching between the second text segmentation and the first word dictionary and the second word dictionary to obtain the string matching result.

[0285] Among them, the attribute dictionary analysis strategy is used to indicate the acquisition of a first word dictionary associated with the first attribute tendency and a second word dictionary associated with the second attribute tendency;

[0286] The result determination unit 1103 includes: a first matching subunit 11031, a first accumulation subunit 11032, a second matching subunit 11033, a second accumulation subunit 11034, and a result determination subunit 11035;

[0287] The first matching subunit 11031 is used to search for a string that matches the second text segment in the first word dictionary when the first word dictionary associated with the content text is obtained based on the attribute dictionary analysis strategy. If a string that matches the second text segment is found in the first word dictionary, the found string is used as the first matching segment corresponding to the second text segment.

[0288] The first accumulation subunit 11032 is used to obtain the first weight information corresponding to the first matching word segmentation, perform a second accumulation process on the first weight information, and obtain the first matching value associated with the first attribute tendency.

[0289] The second matching subunit 11033 is used to search for a string that matches the second text segment in the second word dictionary when the second word dictionary associated with the content text is obtained based on the attribute dictionary analysis strategy. If a string that matches the second text segment is found in the second word dictionary, the found string is used as the second matching segment corresponding to the second text segment.

[0290] The second accumulation subunit 11034 is used to obtain the second weight information corresponding to the second matching word segmentation, perform the second accumulation process on the second weight information, and obtain the second matching value associated with the second attribute tendency.

[0291] The result determination subunit 11035 is used to obtain the string matching result based on the first matching value and the second matching value.

[0292] The specific implementation methods of the first matching subunit 11031, the first accumulation subunit 11032, the second matching subunit 11033, the second accumulation subunit 11034, and the result determination subunit 11035 can be found above. Figure 3 The description of step S102 in the corresponding embodiments will not be repeated here.

[0293] The summation processing unit 1104 is used to determine the first attribute value and the second attribute value of the content details information based on the attribute matching value in the string matching result, and to sum the first attribute value and the second attribute value to obtain the content attribute value of the sample content.

[0294] Optionally, attribute analysis strategies include model learning analysis strategies;

[0295] The model acquisition unit 1105 is used to acquire a target network model associated with the content text based on a model learning analysis strategy; the target network model is obtained by training an initial network model based on the training text content and sample labels;

[0296] The feature extraction unit 1106 is used to input the content text into the target network model, and extract features from the content text through the target network model to obtain the text attribute features corresponding to the content text.

[0297] The numerical prediction unit 1107 is used to determine the matching degree between the text attribute features and the sample attribute features in the classifier of the target network model, and to determine the first predicted value associated with the first attribute tendency and the second predicted value associated with the second attribute tendency based on the matching degree.

[0298] The result update unit 1108 is used to update the string matching result based on the first predicted value and the second predicted value.

[0299] The specific implementation methods of the strategy acquisition unit 1101, word segmentation determination unit 1102, result determination unit 1103, summation processing unit 1104, model acquisition unit 1105, feature extraction unit 1106, numerical prediction unit 1107, and result update unit 1108 can be found above. Figure 3 The description of step S102 in the corresponding embodiments will not be repeated here.

[0300] The content storage module 111 is used to store content details, content attribute values, first attribute values, and second attribute values ​​as content to be recommended in the content database.

[0301] The specific implementation methods of the data receiving module 101, content acquisition module 102, attribute determination module 103, content sending module 104, condition acquisition module 105, threshold comparison module 106, first determination module 107, difference determination module 108, second determination module 109, attribute analysis module 110, and content storage module 111 can be found in the above description. Figure 3 The descriptions of steps S101-S104 in the corresponding embodiments will not be repeated here. Furthermore, the beneficial effects of using the same method will also not be repeated.

[0302] Further, please see Figure 15 , Figure 15 This is a schematic diagram of the structure of a data push device provided in an embodiment of this application. The data push device 20 may include: a data sending module 201, a content receiving module 202, and a content display module 203;

[0303] The data sending module 201 is used to respond to a first trigger operation performed by the first object on historical content in the first client, record business object data of the first object on the historical content based on the first trigger operation, and send the recorded business object data to the server; the historical content includes valid content corresponding to valid object behavior; the valid object behavior is determined by the object behavior recorded by the first object on the historical content within the business object period;

[0304] The content receiving module 202 is used to receive target recommended content associated with valid content sent by the server based on business object data; the target recommended content is recommended content that matches the second attribute tendency; the second attribute tendency is the opposite attribute tendency of the first attribute tendency; the first attribute tendency is the attribute tendency corresponding to the weighted business object attribute value of the first object within the business object period; the weighted business object attribute value is determined based on the object attribute value of the first object; the object attribute value is determined by the content attribute value corresponding to the valid content;

[0305] The content display module 203 is used to display the target recommended content to the first object in the first client.

[0306] The specific implementation methods of the data sending module 201, the content receiving module 202, and the content display module 203 can be found in the above description. Figure 9 The descriptions of steps S201-S207 in the corresponding embodiments will not be repeated here. Furthermore, the beneficial effects of using the same method will also not be repeated.

[0307] Further, please see Figure 16 , Figure 16 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Figure 16 As shown, the computer device 1000 may include a processor 1001, a network interface 1004, and a memory 1005. Furthermore, the computer device 1000 may also include a user interface 1003 and at least one communication bus 1002. The communication bus 1002 is used to enable communication between these components. The user interface 1003 may include a display screen and a keyboard; optionally, the user interface 1003 may also include a standard wired interface or a wireless interface. Optionally, the network interface 1004 may include a standard wired interface or a wireless interface (such as a Wi-Fi interface). The memory 1005 may be high-speed RAM or non-volatile memory, such as at least one disk storage device. Optionally, the memory 1005 may also be at least one storage device located remotely from the processor 1001. Figure 16 As shown, the memory 1005, which is a computer-readable storage medium, may include an operating system, a network communication module, a user interface module, and a device control application.

[0308] In such Figure 16 In the computer device 1000 shown, the network interface 1004 provides network communication functions; the user interface 1003 is mainly used to provide an input interface for users; and the processor 1001 can be used to call the device control application stored in the memory 1005.

[0309] It should be understood that the computer device 1000 described in the embodiments of this application can perform the foregoing... Figure 3 or Figure 9 The description of the data push method in the corresponding embodiments can also be executed as described above. Figure 14 In the corresponding embodiment, the data push device 10 or Figure 15 The description of the data push device 20 in the corresponding embodiments will not be repeated here. Furthermore, the beneficial effects of using the same method will also not be repeated.

[0310] Furthermore, it should be noted that this application embodiment also provides a computer-readable storage medium, which stores a computer program executed by the aforementioned data push device 10 or data push device 20. The computer program includes program instructions, and when the processor executes the program instructions, it can execute the aforementioned... Figure 3 or Figure 9The description of the data push method in the corresponding embodiments is already provided and will not be repeated here. Furthermore, the beneficial effects of using the same method will also not be repeated. For technical details not disclosed in the computer-readable storage medium embodiments related to this application, please refer to the description of the method embodiments of this application.

[0311] Furthermore, it should be noted that this application also provides a computer program product or computer program, which may include computer instructions, which may be stored in a computer-readable storage medium. The processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor may execute the computer instructions, causing the computer device to perform the aforementioned actions. Figure 3 or Figure 9 The description of the data push method in the corresponding embodiments is already provided and will not be repeated here. Furthermore, the beneficial effects of using the same method will also not be repeated. For technical details not disclosed in the computer program products or computer program embodiments related to this application, please refer to the description of the method embodiments of this application.

[0312] For further details, please see Figure 17 , Figure 17 This is a schematic diagram of a data push system provided in an embodiment of this application. The data push system 300 may include a data push device 300a and a data push device 300b. The data push device 300a can be the one described above. Figure 14 The data push device 10 in the corresponding embodiment can be understood to be integrated into the above-mentioned... Figure 2 The server 20a in the corresponding embodiment will not be described again here. The data push device 300b can be the one described above. Figure 15 The data push device 20 in the corresponding embodiment can be understood to be integrated into the above-mentioned data push device 300b. Figure 2 The user terminal 20b in the corresponding embodiment will not be described again here. Furthermore, the beneficial effects of using the same method will also not be described again. For technical details not disclosed in the data push system embodiments of this application, please refer to the description of the method embodiments of this application.

[0313] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0314] The above-disclosed embodiments are merely preferred embodiments of this application and should not be construed as limiting the scope of this application. Therefore, any equivalent variations made in accordance with the claims of this application shall still fall within the scope of this application.

Claims

1. A data push method, characterized in that, The method is executed by the server and includes: The system receives viewing data uploaded by a first object through a first client; the viewing data includes historical content viewed by the first object within the business object cycle; the historical content includes text and image content included in at least one information stream content that is distributed to the first object and actively selected by the first object for consumption by viewing. The effective content corresponding to the effective object behavior determined by the validity judgment is obtained from the historical content; the effective object behavior is determined at least by the object behavior when the number of views recorded by the first object for the historical content within the business object period is greater than the number of views threshold and / or the view completion rate is greater than the completion rate threshold; the number of views is used to characterize the number of clicks by the first object on the at least one information flow content viewed, and the view completion rate is determined by the number of words viewed and the total number of words in the text and image content by the first object within the view duration; The content attribute value corresponding to the effective content is used as the object attribute value of the first object. Based on the object attribute value, the weighted business object attribute value of the first object in the business object period is determined. The attribute tendency corresponding to the weighted business object attribute value is determined as the first attribute tendency of the first object. If the weighted business object attribute value satisfies the reverse attribute recommendation condition corresponding to the business object period, then the attribute balance value in the reverse attribute recommendation condition is obtained. When the reverse attribute value corresponding to the weighted business object attribute value is determined through the attribute balance value, the content to be recommended with the reverse attribute value is taken as the target recommended content that matches the second attribute tendency. The target recommended content is sent to the first client so that the first client can display the target recommended content to the first object. The second attribute tendency is the opposite attribute tendency of the first attribute tendency.

2. The method according to claim 1, characterized in that, If the historical content includes the graphic content, then the business object data also includes the object behavior corresponding to the graphic content; the object behavior corresponding to the graphic content includes the content duration corresponding to the first object; The step of retrieving valid content corresponding to valid object behaviors determined by validity judgment from the historical content includes: The first object is given the number of words in the text and image content within the content duration. Based on the content duration and the number of words, the content speed of the first object in relation to the text and image content is determined. Obtain the content type of the image and text content, and determine the dynamic adjustment coefficient corresponding to the image and text content based on the content type of the image and text content; Obtain the base speed associated with the text and image content, adjust the base speed based on the dynamic adjustment coefficient, use the adjusted base speed as the dynamic speed, compare the content speed with the dynamic speed, and obtain the speed comparison result; If the speed comparison result indicates that the content speed is less than the dynamic speed, then the object behavior corresponding to the text and image content is determined to be a valid object behavior, and the text and image content corresponding to the valid object behavior selected from the text and image content is taken as valid content.

3. The method according to claim 1, characterized in that, The step of using the content attribute value corresponding to the valid content as the object attribute value of the first object, determining the weighted business object attribute value of the first object within the business object period based on the object attribute value, and determining the attribute tendency corresponding to the weighted business object attribute value as the first attribute tendency of the first object includes: The content attribute value corresponding to the valid content is used as the object attribute value of the first object, and the object attribute value is subjected to a first accumulation process to obtain the accumulated content attribute value. Obtain the content quantity of the valid content, and determine the weighted business object attribute value of the first object within the business object period based on the accumulated content attribute value and the content quantity; The attribute tendency corresponding to the weighted business object attribute value is determined as the first attribute tendency of the first object.

4. The method according to claim 1, characterized in that, The method further includes: Obtain the reverse attribute recommendation conditions corresponding to the business object cycle; the reverse attribute recommendation conditions include a first attribute threshold associated with the first attribute tendency and a second attribute threshold associated with the second attribute tendency; When the first attribute threshold is less than the second attribute threshold, the weighted business object attribute value is compared with the first attribute threshold and the second attribute threshold to obtain the threshold comparison result; If the threshold comparison result indicates that the weighted business object attribute value is less than the first attribute threshold or the weighted business object attribute value is greater than the second attribute threshold, then it is determined that the weighted business object attribute value meets the reverse attribute recommendation condition.

5. The method according to claim 1, characterized in that, The method further includes: Obtain the attribute configuration threshold associated with the first object and the attribute fluctuation value associated with the attribute configuration threshold, and determine the first absolute difference between the weighted business object attribute value and the attribute configuration threshold; If the first absolute difference is greater than the attribute fluctuation value, then it is determined that the weighted business object attribute value satisfies the reverse attribute recommendation condition corresponding to the business object period.

6. The method according to claim 1, characterized in that, If the weighted business object attribute value satisfies the reverse attribute recommendation condition corresponding to the business object period, then the attribute balance value in the reverse attribute recommendation condition is obtained. When the reverse attribute value corresponding to the weighted business object attribute value is determined through the attribute balance value, the content to be recommended with the reverse attribute value is taken as the target recommended content matching the second attribute tendency, and the target recommended content is sent to the first client, including: If the weighted business object attribute value satisfies the reverse attribute recommendation condition corresponding to the business object period, then the attribute balance value in the reverse attribute recommendation condition is obtained, and the reverse attribute value corresponding to the weighted business object attribute value is determined based on the attribute balance value; the second absolute difference between the weighted business object attribute value and the attribute balance value is equal to the second absolute difference between the reverse attribute value and the attribute balance value. The content to be recommended that has the reverse attribute value is taken as the target recommended content that matches the second attribute tendency, and the target recommended content is sent to the first client.

7. The method according to claim 1, characterized in that, The historical content refers to the content to be recommended in the content database corresponding to the first client; The content database contains sample content uploaded by the second object through the second client; The method further includes: Obtain the content details information of the sample content, perform attribute analysis on the content details information to obtain the first attribute value and the second attribute value of the content details information, and determine the content attribute value of the sample content based on the first attribute value and the second attribute value; The content details, the content attribute values, the first attribute value, and the second attribute value are stored as content to be recommended in the content database.

8. The method according to claim 7, characterized in that, The steps of obtaining the content details information of the sample content, performing attribute analysis on the content details information to obtain a first attribute value and a second attribute value of the content details information, and determining the content attribute value of the sample content based on the first attribute value and the second attribute value include: The content text in the content details information of the sample content and the attribute analysis strategy associated with the sample content are obtained; the attribute analysis strategy includes an attribute dictionary analysis strategy. The content text is segmented based on the attribute dictionary analysis strategy to obtain the first text segment of the content text. The first text segment is then processed for stop words, and the first text segment after stop word processing is used as the second text segment of the content text. Based on the attribute dictionary analysis strategy, a first word dictionary and a second word dictionary associated with the content text are obtained. The second text is segmented and matched with the first word dictionary and the second word dictionary respectively to obtain the string matching result. Based on the attribute matching values ​​in the string matching results, the first attribute value and the second attribute value of the content details information are determined, and the first attribute value and the second attribute value are summed to obtain the content attribute value of the sample content.

9. The method according to claim 8, characterized in that, The attribute dictionary analysis strategy is used to instruct the acquisition of a first word dictionary associated with the first attribute tendency and a second word dictionary associated with the second attribute tendency; The method involves obtaining a first word dictionary and a second word dictionary associated with the content text based on the attribute dictionary analysis strategy, and then performing string matching between the second text segmentation and the first word dictionary and the second word dictionary to obtain the string matching results, including: When a first word dictionary associated with the content text is obtained based on the attribute dictionary analysis strategy, a string matching the second text segment is searched in the first word dictionary. If a string matching the second text segment is found in the first word dictionary, the found string is used as the first matching segment corresponding to the second text segment. Obtain the first weight information corresponding to the first matched word segmentation, and perform a second accumulation process on the first weight information to obtain the first matching value associated with the first attribute tendency; When a second word dictionary associated with the content text is obtained based on the attribute dictionary analysis strategy, a string matching the second text segment is searched in the second word dictionary. If a string matching the second text segment is found in the second word dictionary, the found string is used as the second matching segment corresponding to the second text segment. Obtain the second weight information corresponding to the second matching word segmentation, perform a second accumulation process on the second weight information, and obtain a second matching value associated with the second attribute tendency; Based on the first matching value and the second matching value, the string matching result is obtained.

10. The method according to claim 8, characterized in that, The attribute analysis strategy includes a model learning analysis strategy; The method further includes: The target network model associated with the content text is obtained based on the model learning and analysis strategy described above; the target network model is obtained by training the initial network model based on the training text content and sample labels. The content text is input into the target network model, and the target network model performs feature extraction on the content text to obtain the text attribute features corresponding to the content text; The matching degree between the text attribute features and the sample attribute features in the classifier is determined by the classifier in the target network model, and a first predicted value associated with the first attribute tendency and a second predicted value associated with the second attribute tendency are determined based on the matching degree. The string matching result is updated based on the first predicted value and the second predicted value.

11. A data push method, characterized in that, The method is executed by the first client and includes: In response to a first trigger operation performed by a first object on historical content in the first client, the system records the first object's viewing data on the historical content based on the first trigger operation and sends the recorded viewing data to the server. The historical content includes text and image content included in at least one information stream content distributed to the first object and actively selected by the first object for consumption through viewing. The at least one information stream content is used to obtain valid content corresponding to valid object behavior determined by validity judgment. The valid object behavior is determined at least by the object behavior when the number of views recorded by the first object on the historical content within the business object cycle is greater than a number threshold and / or the viewing completion rate is greater than a completion rate threshold. The number of views is used to characterize the number of clicks by the first object on the viewed at least one information stream content, and the viewing completion rate is determined by the number of words viewed and the total number of words in the text and image content within the viewing time. The system receives target recommended content associated with the valid content, sent by the server based on the business object data. The target recommended content is content to be recommended that matches a second attribute tendency and has a reverse attribute value. The second attribute tendency is the opposite attribute tendency of the first attribute tendency. The first attribute tendency is the attribute tendency corresponding to the weighted business object attribute value of the first object within the business object period. The weighted business object attribute value is determined based on the object attribute value of the first object. The object attribute value is determined by the content attribute value corresponding to the valid content. The reverse attribute value is determined by the attribute balance value obtained from the reverse attribute recommendation conditions when the weighted business object attribute value satisfies the reverse attribute recommendation conditions corresponding to the business object period. The target recommended content is displayed to the first object in the first client.

12. A data push device, characterized in that, The device operates on a server and includes: The data receiving module is used to receive viewing data uploaded by the first object through the first client; the viewing data includes the historical content viewed by the first object within the business object cycle; the historical content includes text and image content included in at least one information flow content that is distributed to the first object and actively selected by the first object to be consumed by viewing. The content acquisition module is used to acquire valid content corresponding to valid object behaviors determined by validity judgment from the historical content; the valid object behaviors are determined at least by the object behaviors recorded by the first object for the historical content within the business object period when the number of views is greater than a number threshold and / or the view completion rate is greater than a completion rate threshold; the number of views is used to characterize the number of clicks by the first object on the at least one information flow content viewed, and the view completion rate is determined by the number of words read by the first object on the content text of the graphic content and the total number of words within the view duration; The attribute determination module is used to take the content attribute value corresponding to the valid content as the object attribute value of the first object, determine the weighted business object attribute value of the first object in the business object period based on the object attribute value, and determine the attribute tendency corresponding to the weighted business object attribute value as the first attribute tendency of the first object. The content sending module is configured to, if the weighted service object attribute value satisfies the reverse attribute recommendation condition corresponding to the service object period, obtain the attribute balance value in the reverse attribute recommendation condition, and when the reverse attribute value corresponding to the weighted service object attribute value is determined through the attribute balance value, use the content to be recommended with the reverse attribute value as the target recommended content that matches the second attribute tendency, and send the target recommended content to the first client so that the first client can display the target recommended content to the first object; the second attribute tendency is the opposite attribute tendency of the first attribute tendency.

13. A data push device, characterized in that, The device operates in the first client and includes: A data sending module is configured to respond to a first trigger operation performed by a first object on historical content in the first client, record the first object's viewing data on the historical content based on the first trigger operation, and send the recorded viewing data to the server; the historical content includes text and image content included in at least one information stream content distributed to the first object and actively selected by the first object for consumption by viewing, and the at least one information stream content is used to obtain valid content corresponding to valid object behavior determined by validity judgment; the valid object behavior is determined at least by the object behavior when the number of views recorded by the first object on the historical content within the business object cycle is greater than a number threshold and / or the viewing completion rate is greater than a completion rate threshold; the number of views is used to characterize the number of clicks by the first object on the viewed at least one information stream content, and the viewing completion rate is determined by the number of words read by the first object on the text and image content and the total number of words within the viewing time; The content receiving module is used to receive target recommended content associated with the valid content, sent by the server based on the business object data; the target recommended content is content to be recommended that matches a second attribute tendency and has a reverse attribute value; the second attribute tendency is the opposite attribute tendency of the first attribute tendency; the first attribute tendency is the attribute tendency corresponding to the weighted business object attribute value of the first object within the business object period; the weighted business object attribute value is determined based on the object attribute value of the first object; the object attribute value is determined by the content attribute value corresponding to the valid content; the reverse attribute value is determined by the attribute balance value in the reverse attribute recommendation condition obtained when the weighted business object attribute value satisfies the reverse attribute recommendation condition corresponding to the business object period; The content display module is used to display the target recommended content to the first object in the first client.

14. A computer device, characterized in that, include: Processor and memory; The processor is connected to a memory, wherein the memory is used to store a computer program, and the processor is used to invoke the computer program to cause the computer device to perform the method according to any one of claims 1-11.

15. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program adapted to be loaded and executed by a processor to cause a computer device having the processor to perform the method according to any one of claims 1-11.

16. A computer program product, characterized in that, Includes a computer program / instruction that, when executed by a processor, implements the method according to any one of claims 1-11.

Images