Information pushing method and device, equipment, medium and program product

Abstract

The invention discloses an information pushing method and device, equipment, a medium and a program product, and relates to the technical field of computers.The method comprises the steps that at least two pieces of pushing information in a first time period are collected, the at least two pieces of pushing information are to-be-pushed information received by at least one client side, and the to-be-pushed information is sent to the client side; the first time period refers to a time period in a preset duration range before the current moment; obtaining a similarity analysis result between the ith piece of push information in the at least two pieces of push information and other pieces of push information, wherein i is a positive integer; based on the similarity analysis result corresponding to the ith piece of push information, determining a push mode of the ith piece of push information; the pushing mode comprises at least one of interception pushing, merging pushing and direct pushing; and pushing the at least two pieces of push information based on the push modes corresponding to the at least two pieces of push information. Pushing of redundant messages in the same time period can be reduced.

Description

Technical Field

[0001] Embodiments of the present application relate to the field of computer technology, and in particular, to an information push method, device, equipment, medium, and program product. Background Art

[0002] With the development of Internet technology, smart devices have become essential tools in users' daily lives. For example, mobile phones, computers, smart watches, etc. According to different usage requirements of users, different application programs or platforms are often installed on smart devices, and users can obtain Internet information or news from these platforms.

[0003] In the related art, each platform operates independently, and their respective information push systems do not interfere with each other. When a certain hot event occurs, each platform will intelligently push messages related to the hot event to users. For example, blogs, articles, press releases, etc. published by the media, in order to discover content that users may be interested in.

[0004] However, dynamic messages of the same hot event are often pushed to each platform separately due to algorithm differences and operation strategy differences of different platforms. Given the unique push mechanisms of each platform and the diverse browsing preferences of users, there are duplicate contents in the information pushed by multiple platforms for related contents of the same hot event, resulting in information redundancy, increasing the time consumed by users to view information, and affecting the efficiency and experience of users' information reception. Summary of the Invention

[0005] Embodiments of the present application provide an information push method, device, equipment, medium, and program product, which can reduce the push of redundant messages within the same period. The technical solutions are as follows:

[0006] On the one hand, an information push method is provided, and the method includes:

[0007] Collect at least two push messages within a first time period;

[0008] Obtain the similarity analysis result between the at least two push messages;

[0009] Based on the similarity analysis result, determine the push method for the at least two push messages; the push method includes at least one of intercepting push, merging push, and direct push;

[0010] Push at least one of the at least two push messages based on the push method of the at least two push messages.

[0011] On the other hand, an information push device is provided, and the device includes:

[0012] A collection module, configured to collect at least two push messages within a first time period;

[0013] A similarity analysis module, configured to obtain a similarity analysis result between the at least two push messages;

[0014] A push method determination module, configured to determine a push method for the at least two push messages based on the similarity analysis result; the push method includes at least one of intercepting push, merging push, and direct push;

[0015] A push module, configured to push at least one of the at least two push messages based on the push method of the at least two push messages.

[0016] In an optional embodiment, the similarity analysis module is further configured to, for the i-th push message among the at least two push messages, obtain a similarity between the i-th push message and other push messages, where i is a positive integer; obtain similarities between each of the at least two push messages and other push messages respectively, to obtain the similarity analysis result.

[0017] In an optional embodiment, the push message includes a first content and a second content;

[0018] The similarity analysis module is further configured to obtain a first similarity between the first content of the i-th push message and the first content of the other push messages; and obtain a second similarity between the second content of the i-th push message and the second content of the other push messages; obtain the similarity between the i-th push message and the other push messages based on the first similarity and the second similarity.

[0019] In an optional embodiment, the similarity analysis module is further configured to extract features from the first content of the i-th push message to obtain a first content feature representation of the i-th one; extract features from the first content of the other push messages to obtain a first content feature representation corresponding to the other push messages; calculate a similarity between the first content feature representation of the i-th one and the first content feature representation corresponding to the other push messages, to obtain the first similarity;

[0020] The similarity analysis module is further configured to extract features from the second content of the i-th push message to obtain a second content feature representation of the i-th one; extract features from the second content of the other push messages to obtain a second content feature representation corresponding to the other push messages; calculate a similarity between the second content feature representation of the i-th one and the second content feature representation corresponding to the other push messages, to obtain the second similarity.

[0021] In an alternative embodiment, the similarity analysis module is further configured to obtain the similarity between the i-th push message and the other push messages based on the weighted operation result between the first similarity and the second similarity.

[0022] In an alternative embodiment, the push mode determination module is further configured to, for the i-th push message among the at least two push messages, in response to the similarity between the i-th push message and the other push messages being less than a first threshold, determine that the push mode for the i-th push message is direct push, where i is a positive integer.

[0023] In an alternative embodiment, the push mode determination module is further configured to, for the i-th push message among the at least two push messages, in response to the similarity between the i-th push message and the first push message among the other push messages being greater than the first threshold and less than a second threshold, determine that the push mode for the i-th push message is combined push; wherein the push mode of the first push message is the combined push.

[0024] In an alternative embodiment, the push module is further configured to analyze the i-th push message and the first push message to determine a first similar part and a first different part in the i-th push message, and a second similar part and a second different part in the first push message; wherein the matching degree between the first similar part and the second similar part meets a preset combination requirement; the first different part refers to the part of the i-th push message other than the first similar part; the second different part refers to the part of the target push message other than the second similar part; perform a combination process on the first similar part and the second similar part to obtain a combined part; generate a combined push message corresponding to the i-th push message based on the combined part, the first different part, and the second different part; and push the combined push message corresponding to the i-th push message.

[0025] In an alternative embodiment, the push mode determination module is further configured to, for the i-th push message among the at least two push messages, in response to the similarity between the i-th push message and any one of the other push messages being greater than the second threshold, obtain a first quality score of the i-th push message; and determine the push mode for the i-th push message based on the first quality score.

[0026] In an optional embodiment, the push method determination module is further configured to obtain a second quality score of a push message whose similarity to the i-th push message is greater than the second threshold; and determine the push method for the i-th push message based on the first quality score and the second quality score.

[0027] In an optional embodiment, the push method determination module is further configured to, when the first quality score is greater than the second quality score, determine that the push method for the i-th push message is direct push, where the push method for the push message corresponding to the second quality score is intercept push; or, when the first quality score is less than the second quality score, determine that the push method for the i-th push message is intercept push, where the push method for the push message corresponding to the second quality score is direct push; or, when the first quality score is equal to the second quality score, determine that the push method for the push message that meets the preset selection condition among the i-th push message and the push message corresponding to the second quality score is direct push.

[0028] In an optional embodiment, the device further includes:

[0029] A display module, configured to display a first combined push message, where the first combined push message refers to a message obtained by combining at least two target push messages that meet the preset combination requirements among the at least two push messages; the first combined push message includes identifiers of clients corresponding to the at least two target push messages respectively, and the identifier of the client includes a first identifier of a first client; in response to receiving a trigger operation on the first identifier, display the target push message from the first client among the at least two target push messages.

[0030] In an optional embodiment, the at least two push messages are messages to be pushed received by at least one client, and the first time period refers to a time period within a preset duration range before the current moment;

[0031] The acquisition module is further configured to, in response to detecting that there is a current push message to be pushed in the at least one client at the current moment, obtain at least two historical push messages, where the at least two historical push messages refer to messages pushed by the at least one client at a historical moment;

[0032] The similarity analysis module is further configured to obtain a similarity analysis result between the current push message and the at least two historical push messages;

[0033] The push method determination module is further configured to determine the push method for the current push message based on the similarity analysis result corresponding to the current push message.

[0034] On the other hand, a computer device is provided, which includes a processor and a memory. At least one instruction, at least one program, a code set or an instruction set is stored in the memory. The at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by the processor to implement the information push method as described in any one of the embodiments of the present application above.

[0035] On the other hand, a computer-readable storage medium is provided. At least one instruction, at least one program, a code set or an instruction set is stored in the storage medium. The at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by a processor to implement the information push method as described in any one of the embodiments of the present application above.

[0036] On the other hand, a computer program product or a computer program is provided. The computer program product or the computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the information push method as described in any one of the above embodiments.

[0037] The beneficial effects brought by the technical solutions provided by the embodiments of the present application at least include:

[0038] Collect at least two push messages to be pushed within the same time period, analyze the push messages, determine the similarity between the push messages, convert complex information into a digital dimension for evaluation, determine whether there are duplicate messages according to the similarity analysis result, and determine the push method for each message. Targeted interception or merging of push messages through similarity is beneficial to integrating key information in the messages to be pushed, filtering redundant parts, avoiding the problem of information overload caused by pushing duplicate content to users, reducing the time consumed by users to view redundant information, and improving the efficiency of users to obtain information. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained according to these drawings.

[0040] Figure 1 is a schematic diagram of an information push system provided by an exemplary embodiment of the present application;

[0041] Figure 2It is a flowchart of an information push method provided by an exemplary embodiment of the present application;

[0042] Figure 3 It is a schematic diagram of the process for determining the similarity between two push messages provided by an exemplary embodiment of the present application;

[0043] Figure 4 It is a schematic diagram for determining the push method based on the similarity analysis result provided by an exemplary embodiment of the present application;

[0044] Figure 5 It is a schematic diagram for determining the push method based on the quality score provided by an exemplary embodiment of the present application;

[0045] Figure 6 It is a schematic diagram for merging push messages provided by an exemplary embodiment of the present application;

[0046] Figure 7 It is a schematic diagram of the process for deploying a multimodal large model in a terminal provided by an exemplary embodiment of the present application;

[0047] Figure 8 It is a flowchart of an information push method provided by another exemplary embodiment of the present application;

[0048] Figure 9 It is a structural block diagram of an information push device provided by an exemplary embodiment of the present application;

[0049] Figure 10 It is a structural block diagram of an information push device provided by another exemplary embodiment of the present application;

[0050] Figure 11 It is a structural block diagram of a computer device provided by an exemplary embodiment of the present application. Detailed implementation manners

[0051] To make the objectives, technical solutions, and advantages of the present application clearer, the following will further describe the embodiments of the present application in detail with reference to the accompanying drawings.

[0052] The terms used in the present application are only for the purpose of describing specific embodiments and are not intended to limit the present application. The singular forms "a", "the", and "said" used in the present application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0053] It should be noted that the information and data involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards of relevant countries and regions.

[0054] With the development of Internet technology, smart devices have become indispensable tools in users' daily lives, such as mobile phones, computers, smart watches, etc. According to different user needs, each user's smart device often installs different applications or platforms, and users can obtain Internet news or information from these platforms.

[0055] In related technologies, each application and platform operates independently of each other, and the information push systems they build are also isolated from each other, lacking an effective collaborative linkage mechanism. When a hot event breaks out, each platform will rely on its own independent algorithm logic and established operating strategies to intelligently push various messages closely related to the hot event to users, such as blogs, articles, and press releases published by the media, in order to accurately tap into users' potential interests and meet their personalized needs to the greatest extent.

[0056] However, due to the significant differences in algorithm models and distinct differences in operating strategies between different platforms, dynamic messages derived from the same hot event are often pushed to various platforms (applications) in a dispersed manner. Considering the unique push mechanisms of each platform and the diversity of user browsing habits and preferences, when pushing relevant content of the same hot event to users, a large amount of duplicate information is likely to appear between multiple platforms. This information redundancy phenomenon not only increases the time cost required for users to obtain effective information, but also greatly reduces the efficiency of users receiving information, seriously affecting the smooth experience of users obtaining information, and is not conducive to the efficient and accurate dissemination of information.

[0057] The present application provides an information push method, which can analyze the information to be pushed on multiple platforms in the same time period, intercept and push redundant information, merge and push at least two information with high content similarity, and directly push information with low content similarity to other information. It can reduce the repeated push of information on the same topic, improve the efficiency of users in obtaining effective information, and avoid users wasting time receiving repeated and redundant information.

[0058] The information push system involved in the embodiments of the present application is described. For illustration, please refer to Figure 1 The system involves a terminal 100, on which a variety of different clients (applications / platforms) run, and the terminal 100 displays the information pushed by the client on the screen of the terminal 100.

[0059] Exemplarily, during a first time period, it is detected that there are at least two push messages to be pushed for multiple clients. The data cleaning algorithm is used to preprocess the at least two push messages to obtain the push messages after noise removal.

[0060] The preset keywords are obtained, and based on the matching between the keywords and the push messages after noise removal, the push messages containing the keywords among the at least two push messages are filtered to obtain the filtered push messages.

[0061] If the number of the filtered push messages is one, then directly push this message. If the number of the filtered push messages is multiple, then analyze the filtered push messages to determine the push mode of each push message. Among them, the types of push modes include direct push, intercept push, and merge push. Calculate the similarity between each pair of push messages respectively.

[0062] For the i-th push message, if the similarity between the i-th push message and other push messages is less than the first threshold, then directly push the i-th push message, where i is a positive integer.

[0063] If the similarity between the i-th push message and the target push message among other push messages is greater than the first threshold and less than the second threshold, then merge the i-th push message and the target push message and then push.

[0064] If the similarity between the i-th push message and the target push message among other push messages is greater than the second threshold, then perform quality assessment and scoring on the i-th push message and the target push message, directly push the push message with the higher quality score, and intercept the push message with the lower quality score.

[0065] It should be noted that the duration of the first time period is a preset value. One day is divided into multiple time periods with the preset duration as the period, and the first time period is any one of the multiple time periods. For example, when dividing 24 hours with 60 seconds as the preset duration, 1440 time periods are obtained, and the first time period is the 100th time period among them.

[0066] In some embodiments, the push messages are collected in units of moments. After performing the above processing on at least two push messages to be pushed at the same moment, then push. For example, at the first moment, it is detected that there are 3 push messages to be pushed for 3 clients, and each client corresponds to 1 push message. Perform similarity analysis on these 3 push messages to determine that the push mode of the first push message is intercept push; the push modes of the second push message and the third push message are merge push, and after merging the second push message and the third push message, the merged push message is obtained.

[0067] The above-mentioned terminal 100 can be various forms of terminal devices such as mobile phones, tablet computers, desktop computers, portable laptops, smart TVs, vehicle-mounted terminals, smart home devices, etc., and the embodiments of the present application do not limit this.

[0068] It should be noted that the process of determining the push mode of the push information can also be executed by a server in communication with the terminal 100; or jointly executed by the terminal 100 and the server. The server can be an independent physical server, or 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 communications, middleware services, domain name services, security services, Content Delivery Network (CDN), and big data and artificial intelligence platforms.

[0069] In some embodiments, the above-mentioned server can also be implemented as a node in a blockchain system.

[0070] Combined with the above noun introduction and application scenarios, the information push method provided by the present application is described. This method can be executed by the server or the terminal, or jointly executed by the server and the terminal. In the embodiments of the present application, it is described by taking the method executed by the terminal as an example. As Figure 2 shown, Figure 2 is a flowchart of an information push method provided by an exemplary embodiment of the present application. The method includes the following steps.

[0071] Step 210, collect at least two pieces of push information within the first time period.

[0072] Among them, the at least two pieces of push information are the information to be pushed received by at least one client. While collecting the push information, obtain the information source of each piece of push information, that is, determine the client to which each piece of push information belongs.

[0073] The at least one client refers to the software / application program / platform installed in the terminal, which is used to provide various services for users and push hot information or news to users. The push information refers to the message actively sent by the client to the user. When the client runs in the background, it can push information to the user after obtaining the user's authorization. The push information is usually floatingly displayed at a specified position on the screen.

[0074] Exemplarily, the types of push messages include but are not limited to the following: (1) Information-based messages: For example, news clients push current affairs news, weather information, etc.; (2) Social-based messages: For example, social clients push the content / updates posted by associated accounts associated with the user's social account to the user; (3) Shopping-based messages: For example, e-commerce clients push information such as product promotion activities and user order status updates to the user; (4) Life-based messages: For example, various clients push festival-related information to users on the day of the festival.

[0075] Optionally, the push message contains content of at least one modality, including but not limited to: (1) Text content; (2) Image content; (3) Video content; (4) Audio content, etc.

[0076] Herein, the first time period refers to a time period within a preset time range before the current moment. Exemplarily, the preset time range refers to 10 seconds. The moment 10 seconds away from the current moment is the first moment, and the time period between the first moment and the current moment is the first time period. The duration of the first time period can be arbitrary.

[0077] In some embodiments, the terminal periodically collects push messages, divides the time within a day with a preset duration as the period, obtains multiple time periods, and the first time period is the time period where the current moment is located.

[0078] In some embodiments, the duration of the first time period is short and can be approximately regarded as a certain moment. Then, the push messages to be pushed received within the client are collected in units of moments. For example, if at least two push messages are detected at the first moment, the push messages at the first moment are collected as the push messages within the first time period.

[0079] Step 220, obtain the similarity analysis result between at least two push messages.

[0080] Optionally, for the i-th push message among at least two push messages, obtain the similarity between the i-th push message and other push messages, where i is a positive integer.

[0081] Obtain the similarity between each push message among at least two push messages and other push messages respectively to obtain the similarity analysis result.

[0082] That is, the similarity analysis result between at least two push messages includes the similarity between every two push messages. Taking the i-th push message as an example, calculate the similarity between the i-th push message and other push messages respectively. After performing the same steps for all push messages, integrate all the similarities to obtain the similarity analysis result.

[0083] Optionally, after collecting the push messages, at least two push messages are preprocessed to obtain at least two preprocessed push messages. The preprocessing is used to remove the noise in the at least two push messages so that the at least two push messages meet the preset format requirements.

[0084] Exemplarily, the steps of preprocessing the push messages include but are not limited to: removing HTML (Hyper Text Markup Language) tags / labels, special symbols, spaces, etc. from the push messages.

[0085] The following is an example of preprocessing a push message. The push message before preprocessing: An earthquake with magnitude 1, 2, and 3 occurred in a certain area where HTML tags were affected. @#!。 。

[0086] Among them, The content between is the push message, which contains noises such as "HTML tags", "number 1", "number 2", "number 3", special symbols "@#!。" and spaces " ".

[0087] After removing the noise, the preprocessed push message is as follows: An earthquake occurred in a certain area 。

[0088] In some embodiments, the user can customize the blocked words, block the push messages containing the blocked words, and perform similarity analysis after filtering out the push messages that do not contain the blocked words.

[0089] Optionally, obtain the preset keywords and match them with at least two push messages.

[0090] In response to at least one push message among the at least two push messages that matches the keyword, filter out at least one push message from the at least two push messages to obtain the filtered push messages. Perform the step of obtaining the similarity analysis result based on the filtered push messages.

[0091] Among them, the push method of at least one push message is intercepted push.

[0092] Exemplarily, the preset keywords include keyword A, keyword B, and keyword C. Among the at least two push messages, there are push message 1, push message 2, push message 3, and push message 4.

[0093] Among them, push message 1 contains keyword C, push message 2 does not contain any keyword, push message 3 does not contain any keyword, and push message 4 contains keyword A and keyword B.

[0094] Then, filter the push messages 1 and 4 that contain the keyword to obtain the filtered push messages 2 and 3, and determine the push method of the push messages 1 and 4 as interception push.

[0095] Optionally, the push message contains a first content and a second content, and the first content is different from the second content.

[0096] Divide the push message based on at least one of the following methods to obtain the first content and the second content: (1) Divide based on the information position of the content in the push message. For example, the first content is located at the position of the title in the push message, and the second content is located at the position of the body text in the push message; (2) Divide based on the presentation form of the content in the push message. For example, the presentation forms of the first content and the second content are different, and the font size / color / overlay effect, etc. of the first content are different from those of the second content; (3) Divide based on the content type to which the content belongs. For example, the first content is text content, and the second content is image content; (4) Divide based on the content length format. For example, the push message contains headings of different levels, and the content length and format of each heading are different, including the first content (main text heading) in the first format and the second content (sub-heading) in the second format, and the first format is different from the second format.

[0097] Exemplarily, the first content is located at the first information position in the push message, and the second content is located at the second information position in the push message. Among them, the first information position and the second information position are different, and the first information position and the second information position are positions determined based on the information structure of the push message.

[0098] Exemplarily, the first information position refers to the position of the title in the push message, and the second information position refers to the position of the body text content in the push message. The first content is the title content, and the second content is the body text content.

[0099] Obtain the first similarity between the first content of the i-th push message and the first content of other push messages. Also, obtain the second similarity between the second content of the i-th push message and the second content of other push messages.

[0100] Obtain the similarity between the i-th push message and other push messages based on the first similarity and the second similarity.

[0101] Exemplarily, extract features from the first content of the i-th push message to obtain the i-th first content feature representation.

[0102] Extract features from the first content of other push messages to obtain the first content feature representations corresponding to other push messages.

[0103] Calculate the similarity between the i-th first content feature representation and the first content feature representations corresponding to other push messages to obtain a first similarity.

[0104] That is, the first similarity analysis result is the result of analyzing the similarity between the i-th push message and other push messages in terms of the first content.

[0105] Exemplarily, the process of obtaining the second similarity analysis result is the same as the process of obtaining the first similarity analysis result described above.

[0106] Extract features from the second content of the i-th push message to obtain the i-th second content feature representation.

[0107] Extract features from the second content of other push messages to obtain the second content feature representations corresponding to other push messages.

[0108] Calculate the similarity between the i-th second content feature representation and the second content feature representations corresponding to other push messages respectively to obtain a second similarity.

[0109] Optionally, based on the weighted operation result between the first similarity and the second similarity, obtain the similarity between the i-th push message and other push messages.

[0110] For each message in other push messages, perform the above steps to calculate the similarity between the i-th push message and each message in other push messages respectively, and obtain a similarity analysis result.

[0111] By converting push messages into feature representations and calculating similarities, the content relevance between push messages can be determined, the multi-modal content (such as text, images) contained in the messages can be numericalized, the complex information can be converted into the digital dimension for evaluation, which is convenient for computer fast processing and can accurately measure the similarity between messages.

[0112] Exemplarily, take the calculation of the similarity between the i-th push message and the j-th other push message as an example for illustration.

[0113] Split the i-th push message to obtain the first content i and the second content i; split the j-th other push message to obtain the first content j and the second content j.

[0114] Input the first content i, the second content i, the first content j, and the second content j into the BERT (Bidirectional Encoder Representations from Transformers) model respectively. The BERT model converts them into corresponding feature representations: the first content feature representation α1 corresponding to the first content i, the second content feature representation β1 corresponding to the second content i, the first content feature representation α2 corresponding to the first content j, and the second content feature representation β2 corresponding to the second content j.

[0115] Calculate the first similarity S1 between α1 and α2 and the second similarity S2 between β1 and β2 through the following formula 1.

[0116]

[0117] Among them, formula 1 is used to calculate the cosine similarity between feature representation A and feature representation B. When calculating the first similarity S1, replace A and B with α1 and α2 respectively. When calculating the second similarity S2, replace A and B with β1 and β2 respectively.

[0118] Obtain the preset first weight Q1 and second weight Q2. Q1 and Q2 are arbitrary real numbers, which are used to indicate the importance of the first content and the second content in the push message respectively.

[0119] Perform a weighted operation on the first similarity and the second similarity based on the first weight and the second weight: Q1 * S1 + Q2 * S2 = Sij, to obtain the similarity Sij between the i-th push message and the j-th other push message.

[0120] Schematically, as Figure 3 shown, Figure 3 is a schematic diagram of the process of determining the similarity between two push messages.

[0121] Taking push message 1 and push message 2 as examples, split / divide push message 1 and push message 2, and respectively determine their first content and second content: the first content 1, the second content 1, the first content 2, and the second content 2.

[0122] Based on the preset keyword library 300, determine whether there is information containing the words in the keyword library 300 between push message 1 and push message 2. If so, intercept the push; if not, calculate the similarity between push message 1 and push message 2.

[0123] The first content and the second content of the push message 1 and the push message 2 are respectively input into the BERT model 301. After feature extraction by the BERT model 301, the feature representations corresponding to the first content 1, the second content 1, the first content 2, and the second content 2 are output: V11, V12, V21, and V22.

[0124] After calculating the cosine similarity 1 between V11 and V21 and the cosine similarity 2 between V12 and V22, a weighted operation is performed on the cosine similarity 1 and the cosine similarity 2 to obtain a similarity weighted sum 302 as the similarity between the push message 1 and the push message 2.

[0125] Exemplarily, if Q1 is 0.3, Q2 is 0.7, S1 is 0.5, and S2 is 0.3, then Sij = 0.3 * 0.5 + 0.7 * 0.3 = 0.15 + 0.21 = 0.36 is calculated.

[0126] By splitting the push message into different parts according to the structure and presentation form of the push message and assigning specified weights to the content at different positions, rather than directly using the entire push message as a unit to calculate the similarity, higher weights can be assigned to the important parts / key content in the push message, improving the accuracy of similarity calculation and the effect of determining the information push method based on similarity.

[0127] Step 230: Based on the similarity analysis result, determine the push methods for at least two push messages.

[0128] Optionally, the push methods include at least one of intercepting push, merging push, and direct push.

[0129] For the i-th push message, intercepting push means canceling the push of the i-th push message to the user; merging push means merging the i-th push message with other messages that meet the merging push requirements into one message and then pushing it to the user; direct push means pushing the original text of the i-th push message to the user.

[0130] Optionally, for the i-th push message among at least two push messages, in response to the similarity between the i-th push message and other push messages being less than the first threshold, determine the push method for the i-th push message as direct push, where i is a positive integer.

[0131] Exemplarily, the first threshold is 0, and the other push messages include push message A, push message B, and push message C. The similarities between the i-th push message and push message A, push message B, and push message C are -0.5, -0.9, and -0.3 respectively, all less than the first threshold 0. Determine the push method for the i-th push message as direct push.

[0132] By calculating the similarity between push messages and directly pushing the push messages with lower similarity to users, it can ensure that users can view differentiated content in a short time and avoid information overload for users.

[0133] Optionally, for the i-th push message among at least two push messages, in response to the similarity between the i-th push message and the first push message among other push messages being greater than the first threshold and less than the second threshold, it is determined that the push method for the i-th push message is merged push. Among them, the push method of the first push message is merged push.

[0134] Among them, the first threshold is less than the second threshold.

[0135] Exemplarily, the first threshold is 0, the second threshold is 0.5, and the other push messages include push message A, push message B, and push message C. The similarities between the i-th push message and push message A, push message B, and push message C are -0.2, 0.1, and -0.5 respectively. Among them, the similarity between push message B and the i-th push message is greater than the first threshold and less than the second threshold.

[0136] Then it is determined that push message B is the first push message, and after merging push message B and the i-th push message into one message, it is pushed.

[0137] In some embodiments, if there are at least two first push messages whose similarities with the i-th push message are greater than the first threshold and less than the second threshold, then at least two first push messages and the i-th push message can be merged into one message for pushing.

[0138] Exemplarily, the similarities between the i-th push message and push message A, push message B, and push message C are 0.2, 0.1, and -0.5 respectively. Among them, the similarities between push message A and push message B and the i-th push message are greater than the first threshold and less than the second threshold.

[0139] Then it is determined that push message A and push message B are the first push messages, and the push method is: after merging push message A, push message B, and the i-th push message into one message, it is pushed.

[0140] By merging at least two push messages with higher similarities, retaining the different parts of the information and merging the same parts of the information, it can streamline the presentation of push content, eliminate redundant parts, enable users to quickly extract key information, and improve the information acquisition efficiency. While ensuring that more comprehensive and complete information is pushed to users, it reduces the time cost for users to screen information and enhances the user browsing experience.

[0141] Optionally, for the i-th push message among at least two push messages, in response to the similarity between the i-th push message and any one of the other push messages being greater than a second threshold, obtain a first quality score of the i-th push message; determine the push method for the i-th push message based on the first quality score.

[0142] Analyze the i-th push message and the push messages whose similarity with the i-th push message is greater than the second threshold, obtain the first quality score of the i-th push message, and obtain the second quality score of the push messages whose similarity with the i-th push message is greater than the second threshold.

[0143] Determine the push method for the i-th push message based on the first quality score and the second quality score.

[0144] Exemplarily, the second threshold is 0.5, and the similarities between the i-th push message and push message A, push message B, and push message C are 0.7, 0.1, and -0.5 respectively. Since the similarity between push message A and the i-th push message is greater than the second threshold, determine that push message A is the push message corresponding to the second quality score. Determine the push method based on the quality scores of push message A and the i-th push message.

[0145] It should be noted that if there is a push message whose similarity with the i-th push message is equal to a first threshold, determine that the push method for the i-th push message is direct push or combined push; if there is a push message whose similarity with the i-th push message is equal to the second threshold, determine that the push method for the i-th push message is combined push or push based on the quality score.

[0146] Schematically, as Figure 4 shown, Figure 4 is a schematic diagram for determining the push method based on the similarity analysis result.

[0147] The similarity analysis result corresponding to the i-th push message includes the similarities 400 between the i-th push message and all other push messages respectively, and determine the push method based on the interval where the similarity 400 is located.

[0148] Exemplarily, if all the similarities 400 are within the interval [-1, 0], then it is determined that the push method corresponding to the i-th push message is direct push 401; if there is a first similarity among the similarities 400 within the interval (0, 0.5], then it is determined that the push method corresponding to the i-th push message is merged push 402, and at least two push messages corresponding to the first similarity are merged using the large multimodal model; if there is a second similarity among the similarities 400 within the interval (0.5, 1], then the quality scores of at least two push messages corresponding to the second similarity are determined based on the large multimodal model, it is determined that the push method for the push message with the highest quality score is direct push 401, and it is determined that the push method for the push messages with lower quality scores (here referring to other push messages lower than the highest quality score) is intercepted push 403.

[0149] Exemplarily, the format of the i-th push message and other push messages is a json string: ["information source", "text information", "non-text information"], and the large multimodal model (Large Multimodal Models, LMM) is used to evaluate two push messages, score the i-th push message and other push messages, and obtain the first quality score and the second quality score.

[0150] Among them, the large multimodal model is an intelligent model that integrates various modalities of data (such as text, images, audio, etc.) to comprehensively evaluate the quality of push messages and give corresponding quality scores. When scoring push messages, the large multimodal model extracts modal features such as text, images, and audio in the push messages for fusion analysis, and outputs scores according to the preset scoring criteria and weight allocation.

[0151] Among them, when the large multimodal model conducts multi-faceted evaluations on push messages, it includes but is not limited to the following aspects: information integrity, content richness, beneficiality, clarity of content, etc.

[0152] The value range of the quality scores output by the large multimodal model is 0 to 100. The higher the quality score, the more benefits the user receives from the push message and the better the integrity.

[0153] Exemplarily, in the case where the first quality score is greater than the second quality score, it is determined that the push method for the i-th push message is direct push, and among them, the push method for the push message corresponding to the second quality score is intercepted push.

[0154] Exemplarily, in the case where the first quality score is less than the second quality score, it is determined that the push method for the i-th push message is intercepted push, and among them, the push method for the push message corresponding to the second quality score is direct push.

[0155] That is, since the similarity between the i-th push message and the push message corresponding to the second quality score is high and the content is highly relevant, one of the push messages is selected for retention and the other is intercepted.

[0156] By using the quality scores generated by the multi-modal large model to screen and retain the push messages with higher scores among the push messages with high similarity, it is possible to comprehensively consider various modal contents in the push messages, multi-dimensionally mine the differences between the information, improve the objectivity and accuracy of quality assessment, screen and push the content with higher quality and more effective information for users, and improve the efficiency of users obtaining information.

[0157] Exemplarily, when the first quality score is equal to the second quality score, it is determined that the push method for the push message that meets the preset selection conditions among the i-th push message and the push message corresponding to the second quality score is direct push.

[0158] In some embodiments, the first quality score and the second quality score are the same, indicating that the information natures of the two push messages are the same and the content is balanced, so that the amount of information and the beneficiality obtained by the user are the same when pushed to the user. Optionally, one of the messages is selected for pushing and the other is intercepted.

[0159] Alternatively, one piece of information is selected from the i-th push message and the push message corresponding to the second quality score for retention and push based on the preset selection conditions.

[0160] Optionally, the preset selection conditions include, but are not limited to, the following: (1) The number of key information: For example, count the number of key information such as time, place, protagonist (person or animal), event background, result, etc. in the push message, and retain and push the push message with more key information to ensure the integrity and clarity of the push message; (2) The time order of receiving the push message: For example, retain and push the latest received push message between the two to ensure the timeliness of the push message; (3) The number of text characters in the push message: For example, retain and push the push message with fewer text characters to ensure the conciseness of the push message and make it more efficient for users to read; (4) The number of images included in the push message: For example, retain and push the push message with more images. Compared with text content, users are usually more sensitive to image or graphic elements. Selecting the information with more image elements for pushing can ensure the intuitive and easy-to-understand nature of the push message.

[0161] Schematically, as Figure 5 shown, Figure 5 is a schematic diagram for determining the push method based on the quality score.

[0162] Taking push message 1 and push message 2 as examples, the similarity between push message 1 and push message 2 is greater than the second threshold. Use a prompt (guidance text) to combine push message 1 and push message 2 to generate an instruction, input it into the multi-modal large model 500. The multi-modal large model 500 outputs the quality score 1 of push message 1 and the quality score 2 of push message 2, and based on the scores, an evaluation is made: the push method corresponding to the push message with a low quality score is intercepted push, and the push method corresponding to the push message with a high quality score is direct push.

[0163] In some embodiments, only the above process is used to judge the push method based on the similarity between two push messages. If there is a push message with a relatively high quality score, the push is temporarily retained, but it is necessary to further compare the push message with a relatively high quality score with all other push messages before determining whether to push.

[0164] In some embodiments, after obtaining the similarity analysis result corresponding to the i-th push message, the push method of the i-th push message can also be determined in the following way.

[0165] The similarity analysis result includes the similarities between the i-th push message and other push messages respectively. After taking the average value of the similarities, the average similarity is obtained, and the push method is determined based on the relationship between the average similarity and the preset threshold.

[0166] Optionally, in response to the average similarity between the i-th push message and other push messages being less than the first threshold, the push method is determined to be direct push.

[0167] Optionally, in response to the average similarity between the i-th push message and other push messages being greater than the first threshold and less than the second threshold, the push method is determined to be merged push. Select the m push messages with the highest similarity values from other push messages and merge them with the i-th push message for pushing, where m is a positive integer.

[0168] Optionally, in response to the average similarity between the i-th push message and other push messages being greater than the second threshold, select the n push messages with the highest similarity values from other push messages, determine the quality scores corresponding to the n push messages and the i-th push message respectively. If the quality score of the i-th push message is the highest, directly push the i-th push message. If the quality score of the i-th push message is lower than any of the n push messages, intercept the push of the i-th push message.

[0169] Step 240, push at least one push message among at least two push messages based on the push methods of the at least two push messages.

[0170] Among them, if the number of messages determined to be pushed among at least two push messages exceeds 2, the at least two push messages can be pushed in the following form to avoid the situation of user information overload caused by displaying multiple messages on the terminal screen at one time.

[0171] Exemplarily, sort based on the chronological order of collection of at least two push messages, and push the at least two push messages in sequence at preset time intervals. Or, after grouping and typesetting the at least two push messages, push them sequentially in units of a group of push messages.

[0172] For the messages that need to be merged and pushed among at least two push messages, the pushing process is as follows:

[0173] Optionally, the similarity between the i-th push message and the first push message meets the conditions for merged pushing, and the two are merged and pushed together.

[0174] Analyze the i-th push message and the first push message to determine the first similar part and the first different part in the i-th push message, and the second similar part and the second different part in the first push message.

[0175] Among them, the matching degree between the first similar part and the second similar part meets the preset merging requirements. The first different part refers to the part of the i-th push message other than the first similar part. The second different part refers to the part of the target push message other than the second similar part.

[0176] Perform a merging process on the first similar part and the second similar part to obtain a merged part.

[0177] Generate a merged push message corresponding to the i-th push message based on the merged part, the first different part, and the second different part.

[0178] Push the merged push message corresponding to the i-th push message.

[0179] Exemplarily, set a prompt for the i-th push message and the first push message, and input the i-th push message, the first push message, and the prompt into a pre-trained multi-modal large model. The prompt refers to the guiding text input to the multi-modal large model, which is used to instruct the model to generate and output specified content based on this.

[0180] Exemplarily, the formats of the i-th push message and the first push message before merging are json strings: ["information source", "text information", "non-text information"]. Obtain the "text information" and "non-text information" in the two messages, and set the prompt as follows: "There is a certain similarity between the above two messages. Please analyze the two messages, and without changing the original meaning, merge the similar parts of the two messages and integrate the dissimilar parts to ensure the accuracy and integrity of the information."

[0181] Among them, the i-th merged push message after merging combines the "text information" and "non-text information" in the two messages.

[0182] In some embodiments, when pushing the merged push message to the user, the user can view the original text of the push message before merging through the merged push message.

[0183] Optionally, display the first merged push message.

[0184] The first merged push message refers to a message obtained by merging at least two target push messages that meet the preset merging requirements among at least two push messages.

[0185] The first merged push message contains the identifiers of the clients corresponding to at least two target push messages respectively, and the identifier of the client contains the first identifier of the first client.

[0186] In response to receiving a trigger operation on the first identifier, display the target push message from the first client among at least two target push messages.

[0187] Exemplarily, the first merged push message is the information obtained by merging the push message A from the first client, the push message B from the second client, and the push message C from the third client.

[0188] The format of the first merged push message is as follows: ["first identifier", "second identifier", "third identifier", "merged information"], where the first identifier corresponds to the first client and is used to prompt the user to view the source / original text of the push message A after triggering; the second identifier corresponds to the second client and is used to prompt the user to view the source / original text of the push message B after triggering; the third identifier corresponds to the third client and is used to prompt the user to view the source / original text of the push message B after triggering.

[0189] In some embodiments, the first merged push message is obtained by merging at least two target push messages from the same client (such as the first client). The format of the first merged push message is as follows: ["First identifier", "Merged information"]. After triggering the first identifier, the original links corresponding to at least two target push messages can be expanded, and after triggering one of the original links, the original text of the corresponding target push message is displayed.

[0190] Alternatively, the format of the first merged push message is as follows: ["First identifier 1", "First identifier 2", "Merged information"], where the first identifier 1 and the first identifier 2 correspond to different target push messages, and the original text of the corresponding target push message is displayed after triggering.

[0191] Schematically, as Figure 6 shown, Figure 6 is a schematic diagram of a merged push message.

[0192] Taking push message 1 and push message 2 as examples for illustration, after splitting the text information and non-text information of push message 1 and push message 2 respectively, the prompt is used to combine the above text information and non-text information to generate an instruction input into the multi-modal large model 600. The multi-modal large model 600 merges the text information and non-text information parts to obtain the merged information 601, and combines the information sources (clients) of push message 1 and push message 2 respectively to generate the merged push message 602.

[0193] Schematically, as Figure 7 shown, Figure 7 is a schematic diagram of the process of deploying a multi-modal large model in a terminal.

[0194] The multi-modal large model 700 involved in this application is compiled and deployed on the terminal using the MLC-LLM framework (Machine Learning Compiler for Large-Language Models, a machine learning compiler for large language models). This framework can be applied to multiple platforms, can easily handle the compilation requirements of different intelligent devices, greatly simplifies the user's integration process, and can further optimize the model performance to adapt to its own use cases. According to the configuration of the terminal, select an appropriate quantization bit number (such as 8bit, 4bit) to quantize the multi-modal large model 700 so that it can run on the terminal, and then perform targeted compilation and deployment according to the terminal system to be deployed (such as IOS, Android, etc.).

[0195] In summary, the information push method provided by this application can collect at least two push messages to be pushed within the same time period, analyze the push messages, determine the similarity between the push messages, convert complex information into a digital dimension for evaluation, determine whether there are duplicate messages according to the similarity analysis result, and determine the push method for each message. By intercepting or merging push messages according to similarity, it is beneficial to integrate key information in the messages to be pushed, filter redundant parts, avoid the problem of information overload caused by pushing duplicate content to users, reduce the time consumed by users to view redundant information, and improve the efficiency of users to obtain information.

[0196] There may be similarities between the information pushed by different clients within the same time period or around the same theme. Therefore, the method of analyzing multiple push messages pushed within the same time period / same moment before pushing can avoid the repeated pushing of redundant information. In some embodiments, due to the different processing mechanisms of push messages by each client, the push time points for the same hot information may not be the same. To avoid pushing information on the same theme to users at different time periods, it is possible to analyze between the latest received push message and the historical push messages that have been pushed to users within the historical time period, and intercept the push messages that are highly similar to the historical push messages to prevent users from spending time browsing duplicate content.

[0197] Figure 8 It is a flowchart of an information push method provided by another exemplary embodiment of this application, and this method includes the following steps.

[0198] Step 810, in response to detecting that there is at least one current push message to be pushed in at least one client at the current moment, obtain at least two historical push messages.

[0199] Optionally, at least two historical push messages refer to the messages pushed by at least one client at a historical moment. The historical moment refers to the moment when information is pushed to the user within a historical time period within a preset time range from the current moment.

[0200] For example, taking 24 hours as the preset time range and the current moment as the end moment, a historical time period is divided. Within the historical time period, at least one client has pushed a total of 100 push messages to the user, and these 100 push messages are historical push messages.

[0201] Step 820, obtain the similarity analysis result between the current push message and at least two historical push messages.

[0202] Among them, the process of obtaining the similarity analysis result corresponding to the current push message refers to the description of step 220 above, and will not be elaborated here.

[0203] Optionally, the push information includes a first content at a first information position and a second content at a second information position.

[0204] Obtain a first similarity analysis result between the first content of the current push information and the first content of the historical push information. And, obtain a second similarity analysis result between the second content of the current push information and the second content of the historical push information.

[0205] Obtain a similarity analysis result based on the first similarity analysis result and the second similarity analysis result.

[0206] Exemplarily, extract features from the first content of the current push information to obtain a feature representation of the current first content.

[0207] Extract features from the first content of the historical push information to obtain a feature representation of the first content corresponding to the historical push information.

[0208] Obtain a first similarity analysis result based on the similarity between the current feature representation of the first content and the feature representations of the first content corresponding to the historical push information respectively.

[0209] Among them, the process of obtaining the second similarity analysis result is the same as above and will not be elaborated here.

[0210] Exemplarily, for the k-th historical push information in the historical push information, obtain the first similarity corresponding to the k-th historical push information from the first similarity analysis result, and, obtain the second similarity corresponding to the k-th historical push information from the second similarity analysis result. k is a positive integer.

[0211] Obtain the similarity between the current push information and the k-th historical push information in the historical push information based on the weighted operation result between the first similarity and the second similarity.

[0212] Obtain a similarity analysis result based on the similarity between the current push information and the historical push information.

[0213] Step 830, based on the similarity analysis result corresponding to the current push information, determine the push mode of the current push information.

[0214] Among them, the similarity analysis result corresponding to the current push information includes the respective similarities between the current push information and each historical push information. If there is any historical push information with a relatively high similarity to the current push information, then intercept and push the current information.

[0215] Optionally, in response to the similarity analysis result indicating that the similarities between the current push information and the historical push information are less than a first threshold, determine the push mode of the current push information as direct push.

[0216] That is, the similarity between the current push message and each historical push message is relatively low, and it is determined that the content theme of the current push message is different from the content themes of all push messages browsed by the user in the historical time period, and the current push message is directly pushed to the user.

[0217] Optionally, in response to the similarity analysis result indicating that the similarity between the current push message and the target push message among the historical push messages is greater than the first threshold and less than the second threshold, it is determined that the push method for the current push message is partial push.

[0218] That is, there is partial similarity between the current push message and the target push message, and there is partial identity between the content theme of the current push message and the content theme of the target push message browsed by the user in the historical time period. Only the parts where the current push message and the target push message are different are pushed, and the same parts are hidden.

[0219] Exemplarily, the current push message and the target push message are analyzed, and the target difference part that exists different from the target push message is determined in the current push message.

[0220] Based on the target difference part, the current push message is folded to obtain a folded push message corresponding to the current push message, where the folded push message is the information that shows the target difference part of the current push message in a prominent display form, and the information other than the target difference part in the current push message is folded / hidden.

[0221] For example, the content of the current push message contains 8 paragraphs, and the content of the 3rd, 4th, 5th, and 6th paragraphs is similar to the content of the target push message. Then, this part of the content is hidden, or this part of the content is summarized into a short content and then displayed. The content of the 1st, 2nd, 7th, and 8th paragraphs among the 8 paragraphs is highlighted. The folded push message is pushed to the user.

[0222] Among them, when performing partial push on the current push message, the source of the current push message is displayed in the folded push message, which is convenient for the user to view the complete current push message. Exemplarily, the folded push message contains the identifier of the client to which the current push message belongs. Triggering the identifier in the folded push message displays the original text of the current push message.

[0223] Optionally, in response to the similarity analysis result indicating that the similarity between the current push message and the second push message among the historical push messages is greater than the second threshold, it is determined that the push method for the current push message is intercept push.

[0224] That is, there is a high similarity between the current push message and the target push message, and the content theme of the current push message is the same as the content theme of the target push message that the user has browsed in the historical time period. The current push message is intercepted.

[0225] In summary, the information push method provided by this application takes into account the problem that the client does not respond to the timeliness of the push message. It performs a similarity analysis between the latest received push message and the historical push messages that have been pushed to the user in the historical time period, and determines the push method of the current push message based on the similarity analysis result. It can avoid pushing messages with the same theme to the user at different time periods, intercept push messages that are highly similar to the historical push messages, and prevent the user from spending time browsing duplicate content.

[0226] Figure 9 It is a structural block diagram of an information push device provided by an exemplary embodiment of this application. As Figure 9 shown, the device includes the following parts.

[0227] The acquisition module 910 is used to acquire at least two push messages within the first time period;

[0228] The similarity analysis module 920 is used to obtain the similarity analysis result between the at least two push messages;

[0229] The push method determination module 930 is used to determine the push method of the at least two push messages based on the similarity analysis result; the push method includes at least one of intercepting push, merging push, and directly pushing;

[0230] The push module 940 is used to push at least one of the at least two push messages based on the push method of the at least two push messages.

[0231] In an optional embodiment, the similarity analysis module 920 is further used to obtain the similarity between the i-th push message among the at least two push messages and other push messages, where i is a positive integer; obtain the similarity between each push message among the at least two push messages and other push messages respectively, to obtain the similarity analysis result.

[0232] In an optional embodiment, the push message includes a first content and a second content;

[0233] The similarity analysis module 920 is further configured to obtain a first similarity between the first content of the i-th push message and the first content of the other push messages; and obtain a second similarity between the second content of the i-th push message and the second content of the other push messages; and obtain the similarity between the i-th push message and the other push messages based on the first similarity and the second similarity.

[0234] In an alternative embodiment, the similarity analysis module 920 is further configured to extract features from the first content of the i-th push message to obtain a first content feature representation of the i-th; extract features from the first content of the other push messages to obtain a first content feature representation corresponding to the other push messages; calculate the similarity between the first content feature representation of the i-th and the first content feature representation corresponding to the other push messages to obtain the first similarity;

[0235] The similarity analysis module 920 is further configured to extract features from the second content of the i-th push message to obtain a second content feature representation of the i-th; extract features from the second content of the other push messages to obtain a second content feature representation corresponding to the other push messages; calculate the similarity between the second content feature representation of the i-th and the second content feature representation corresponding to the other push messages to obtain the second similarity.

[0236] In an alternative embodiment, the similarity analysis module 920 is further configured to obtain the similarity between the i-th push message and the other push messages based on the weighted operation result between the first similarity and the second similarity.

[0237] In an alternative embodiment, the push mode determination module 930 is further configured to, for the i-th push message among the at least two push messages, in response to the similarity between the i-th push message and the other push messages being less than a first threshold, determine the push mode of the i-th push message as direct push, where i is a positive integer.

[0238] In an alternative embodiment, the push mode determination module 930 is further configured to, for the i-th push message among the at least two push messages, in response to the similarity between the i-th push message and the first push message among the other push messages being greater than the first threshold and less than a second threshold, determine the push mode of the i-th push message as combined push; wherein the push mode of the first push message is the combined push.

[0239] In an optional embodiment, the push module 940 is further configured to analyze the i-th push message and the first push message to determine a first similar part and a first different part in the i-th push message, and a second similar part and a second different part in the first push message; wherein, the matching degree between the first similar part and the second similar part meets a preset merging requirement; the first different part refers to the part of the i-th push message other than the first similar part; the second different part refers to the part of the target push message other than the second similar part; perform a merging process on the first similar part and the second similar part to obtain a merged part; generate a merged push message corresponding to the i-th push message based on the merged part, the first different part, and the second different part; and push the merged push message corresponding to the i-th push message.

[0240] In an optional embodiment, the push mode determining module 930 is further configured to, for the i-th push message among the at least two push messages, in response to the similarity between the i-th push message and any one of the other push messages being greater than a second threshold, obtain a first quality score of the i-th push message; and determine a push mode for the i-th push message based on the first quality score.

[0241] In an optional embodiment, the push mode determining module 930 is further configured to obtain a second quality score of a push message whose similarity with the i-th push message is greater than the second threshold; and determine a push mode for the i-th push message based on the first quality score and the second quality score.

[0242] In an optional embodiment, the push mode determining module 930 is further configured to, when the first quality score is greater than the second quality score, determine that the push mode for the i-th push message is direct push, where the push mode of the push message corresponding to the second quality score is intercept push; or, when the first quality score is less than the second quality score, determine that the push mode for the i-th push message is intercept push, where the push mode of the push message corresponding to the second quality score is direct push; or, when the first quality score is equal to the second quality score, determine that the push mode of the push message that meets a preset selection condition among the i-th push message and the push message corresponding to the second quality score is direct push.

[0243] In an optional embodiment, as Figure 10 shown, the apparatus further includes:

[0244] A display module 950 is configured to display a first merged push message, where the first merged push message refers to a message obtained by merging at least two target push messages that meet a preset merging requirement among the at least two push messages; the first merged push message includes identifiers of clients corresponding to the at least two target push messages respectively, and the identifier of the client includes a first identifier of a first client; in response to receiving a trigger operation on the first identifier, the target push message from the first client among the at least two target push messages is displayed.

[0245] In an optional embodiment, the at least two push messages are messages to be pushed received by at least one client, and the first time period refers to a time period within a preset duration range before the current moment.

[0246] The acquisition module 910 is further configured to, in response to detecting that there is a current push message to be pushed in the at least one client at the current moment, acquire at least two historical push messages, where the at least two historical push messages refer to messages pushed by the at least one client at a historical moment.

[0247] The similarity analysis module 920 is further configured to obtain a similarity analysis result between the current push message and the at least two historical push messages.

[0248] The push mode determination module 930 is further configured to determine a push mode for the current push message based on the similarity analysis result corresponding to the current push message.

[0249] In summary, the information push device provided in this application can acquire at least two push messages to be pushed within the same time period, analyze the push messages, determine the similarity between the push messages, convert complex information into a digital dimension for evaluation, determine whether there is duplicate information according to the similarity analysis result, and determine the push mode for each piece of information. Targeted interception or merging of push messages through similarity is beneficial to integrating key information in the messages to be pushed, filtering redundant parts, avoiding the problem of information overload caused by pushing duplicate content to users, reducing the time consumed by users to view redundant information, and improving the efficiency of users to obtain information.

[0250] It should be noted that: for the information push device provided in the above embodiment, only the above division of each functional module is used for illustration. In practical applications, the above functions can be allocated to different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the information push device provided in the above embodiment and the information push method embodiment belong to the same concept, and the specific implementation process can be found in the method embodiment, which will not be elaborated here.

[0251] Figure 11 FIG.

[0251] shows a block diagram of a computer device 1100 provided by an exemplary embodiment of the present application. The computer device 1100 may be: a smart phone, a tablet computer, an MP3 player (Moving Picture Experts Group Audio Layer III), an MP4 (Moving Picture Experts Group Audio Layer IV) player, a laptop computer or a desktop computer. The computer device 1100 may also be referred to by other names such as a user device, a portable terminal, a laptop terminal, a desktop terminal, etc.

[0252] Generally, the computer device 1100 includes a processor 1101 and a memory 1102.

[0253] The processor 1101 may include one or more processing cores, such as a quad-core processor, an octa-core processor, etc. The processor 1101 may be implemented in at least one hardware form of DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), PLA (Programmable Logic Array). The processor 1101 may also include a main processor and a coprocessor. The main processor is a processor for processing data in the wake state, also known as the CPU (Central Processing Unit); the coprocessor is a low-power processor for processing data in the standby state. In some embodiments, the processor 1101 may be integrated with a GPU (Graphics Processing Unit), and the GPU is responsible for rendering and drawing the content to be displayed on the display screen. In some embodiments, the processor 1101 may also include an AI (Artificial Intelligence) processor, and the AI processor is used to process computational operations related to machine learning.

[0254] The memory 1102 may include one or more computer-readable storage media, which may be non-transitory. The memory 1102 may also include high-speed random access memory and non-volatile memory, such as one or more disk storage devices and flash storage devices. In some embodiments, the non-transitory computer-readable storage media in the memory 1102 is used to store at least one instruction, and the at least one instruction is to be executed by the processor 1101 to implement the information push method provided in the method embodiments of this application.

[0255] In some embodiments, the computer device 1100 further includes some other components 1103, and the types and quantities of the other components 1103 can be selected based on the functional requirements of the computer device 1100. Those skilled in the art can understand that Figure 11 the structure shown in does not constitute a limitation on the computer device 1100, and it may include more or fewer components than shown in the figure, or combine certain components, or adopt different component arrangements.

[0256] Optionally, the computer-readable storage medium may include: read-only memory (ROM), random access memory (RAM), solid state drives (SSD), or optical discs, etc. Among them, the random access memory may include resistive random access memory (ReRAM) and dynamic random access memory (DRAM). The serial numbers of the embodiments of this application above are only for description and do not represent the advantages or disadvantages of the embodiments.

[0257] The embodiments of this application also provide a computer device, which includes a processor and a memory. At least one instruction, at least one program, a code set, or an instruction set is stored in the memory, and the at least one instruction, the at least one program, the code set, or the instruction set is loaded and executed by the processor to implement the information push method as described in any one of the above embodiments of this application.

[0258] The embodiments of this application also provide a computer-readable storage medium. At least one instruction, at least one program, a code set, or an instruction set is stored in the storage medium, and the at least one instruction, the at least one program, the code set, or the instruction set is loaded and executed by a processor to implement the information push method as described in any one of the above embodiments of this application.

[0259] The embodiments of the present application also provide a computer program product or a computer program. The computer program product or the computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the information push method described in any one of the above embodiments.

[0260] Those of ordinary skill in the art can understand that all or part of the steps to implement the above embodiments can be completed by hardware, or can be completed by instructing relevant hardware through a program. The program can be stored in a computer-readable storage medium. The storage medium mentioned above can be a read-only memory, a magnetic disk, an optical disk, or the like.

[0261] The above are only optional embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. An information push method, characterized in that: The method comprises: Collect at least two push messages within the first time period; Obtaining a similarity analysis result between the at least two pieces of push information; Based on the similarity analysis result, determining a push mode for the at least two push messages; the push mode includes at least one of intercept push, merge push, and direct push; At least one of the at least two push messages is pushed based on the pushing mode of the at least two push messages.

2. The method according to claim 1, characterized in that The obtaining the similarity analysis result between the at least two pieces of push information includes: For the i-th push information among the at least two push information, obtaining the similarity between the i-th push information and other push information, where i is a positive integer; The similarity between each push message in the at least two push messages and the other push messages is obtained to obtain the similarity analysis result.

3. The method according to claim 2, characterized in that The push information includes first content and second content; The obtaining of the similarity between the i-th push information and other push information includes: Acquire a first similarity between the first content of the i-th push information and the first content of the other push information; and acquire a second similarity between the second content of the i-th push information and the second content of the other push information; The similarity between the i-th push information and other push information is obtained based on the first similarity and the second similarity.

4. The method according to claim 3, characterized in that: The obtaining a first similarity between the first content of the i-th push information and the first content of the other push information includes: Performing feature extraction on the first content of the i-th push information to obtain an i-th first content feature representation; Extracting features of the first content of the other pushed information to obtain a first content feature representation corresponding to the other pushed information; Calculating the similarity between the i-th first content feature representation and the first content feature representation corresponding to the other pushed information to obtain the first similarity; The obtaining the second similarity between the second content of the i-th push information and the second content of the other push information includes: Performing feature extraction on the second content of the i-th push information to obtain an i-th second content feature representation; Extracting features of the second content of the other push information to obtain a second content feature representation corresponding to the other push information; The similarity between the i-th second content feature representation and the second content feature representation corresponding to the other push information is calculated to obtain the second similarity.

5. The method according to claim 3, characterized in that: The obtaining the similarity between the i-th push information and other push information based on the first similarity and the second similarity includes: Based on a weighted calculation result between the first similarity and the second similarity, the similarity between the i-th push information and the other push information is obtained.

6. The method according to any one of claims 1 to 5, characterized in that: The determining of a push mode for the at least two pieces of push information includes: For the i-th push information among the at least two push information, in response to the similarity between the i-th push information and the other push information being less than a first threshold, determining that the push mode for the i-th push information is direct push, where i is a positive integer.

7. The method according to any one of claims 1 to 5, characterized in that: The determining of a push mode for the at least two pieces of push information includes: For the i-th push information among the at least two push information, in response to a similarity between the i-th push information and the first push information among the other push information being greater than a first threshold and less than a second threshold, determining that the push mode for the i-th push information is a combined push; The push mode of the first push information is the combined push.

8. The method according to claim 7, characterized in that The pushing of at least one of the at least two pieces of push information in a push manner based on the at least two pieces of push information includes: Analyze the i-th push information and the first push information to determine a first similar part and a first different part in the i-th push information, and a second similar part and a second different part in the first push information; wherein the matching degree between the first similar part and the second similar part meets the preset merging requirement; the first different part refers to the part of the i-th push information other than the first similar part; the second different part refers to the part of the target push information other than the second similar part; Merging the first similar part and the second similar part to obtain a merged part; generating a combined push information corresponding to the i-th push information based on the combined part, the first difference part and the second difference part; Push the combined push information corresponding to the i-th push information.

9. The method according to any one of claims 1 to 5, characterized in that: The determining of a push mode for the at least two pieces of push information includes: For an i-th push information among the at least two push information, in response to a similarity between the i-th push information and any push information among the other push information being greater than a second threshold, obtaining a first quality score of the i-th push information; A push mode for the i-th push information is determined based on the first quality score.

10. The method according to claim 9, characterized in that The determining, based on the first quality score, a push mode for the i-th push information includes: Obtaining a second quality score of the push information whose similarity to the i-th push information is greater than the second threshold; A push mode for the i-th push information is determined based on the first quality score and the second quality score.

11. The method according to claim 10, characterized in that The determining, based on the first quality score and the second quality score, a push mode for the i-th push information includes: When the first quality score is greater than the second quality score, determining that the push mode for the i-th push information is direct push, wherein the push mode for the push information corresponding to the second quality score is intercept push; or, When the first quality score is less than the second quality score, determining that the push mode for the i-th push information is intercept push, wherein the push mode for the push information corresponding to the second quality score is direct push; or, When the first quality score is equal to the second quality score, it is determined that the push mode for the push information that meets the preset selection condition among the i-th push information and the push information corresponding to the second quality score is direct push.

12. The method according to any one of claims 1 to 5, characterized in that: After pushing at least one of the at least two pieces of push information in a push manner based on the at least two pieces of push information, the method further includes: Displaying a first merged push information, wherein the first merged push information refers to a message for merging at least two target push information that meet a preset merging requirement among the at least two push information; the first merged push information includes identifiers of clients corresponding to the at least two target push information, respectively, and the identifiers of the clients include a first identifier of the first client; In response to receiving a trigger operation on the first identifier, the target push information from the first client among the at least two target push information is displayed.

13. The method according to any one of claims 1 to 5, characterized in that: The at least two pieces of push information are pieces of information to be pushed that are received by at least one client, and the first time period refers to a time period of a preset time range before the current moment; The method further comprises: In response to detecting that there is current push information to be pushed in the at least one client at the current moment, obtaining at least two pieces of historical push information, wherein the at least two pieces of historical push information refer to information pushed by the at least one client at a historical moment; Obtaining a similarity analysis result between the current push information and the at least two historical push information; Based on the similarity analysis result corresponding to the current push information, a push mode for the current push information is determined.

14. An information push device, characterized in that: The device comprises: A collection module, used for collecting at least two pieces of push information within a first time period; A similarity analysis module, used to obtain a similarity analysis result between the at least two pieces of push information; A push mode determination module, used to determine a push mode for the at least two push messages based on the similarity analysis result; the push mode includes at least one of intercept push, merge push, and direct push; The push module is used to push at least one of the at least two push messages based on the push mode of the at least two push messages.

15. A computer device, characterized in that: The computer device includes a processor and a memory, wherein the memory stores at least one program, and the at least one program is loaded and executed by the processor to implement the information push method according to any one of claims 1 to 13.

16. A computer-readable storage medium, characterized in that: The storage medium stores at least one program, and the at least one program is loaded and executed by the processor to implement the information push method as described in any one of claims 1 to 13.

17. A computer program product, characterized in that It comprises a computer program, which, when executed by a processor, implements the information push method as described in any one of claims 1 to 13.

Images