A method for object recommendation in live broadcast

Abstract

The present application discloses a method for recommending an object in a live broadcast, which belongs to the field of electronic digital data processing, including displaying a target object in a live broadcast room corresponding to an anchor; obtaining a target live broadcast duration in real time; obtaining first interaction information when the target live broadcast duration reaches a first duration; starting a first live broadcast mode when the first interaction information satisfies a first interaction condition, and displaying multiple interactive tags in the live broadcast room in the first live broadcast mode; obtaining second interaction information when the target live broadcast duration reaches a second duration; inputting the second interaction information into a target model to obtain an initial time of a countdown effect; displaying a countdown effect in the live broadcast room, and displaying a timeout reminder message in the live broadcast room when the time in the countdown effect reaches zero, and the timeout reminder message is used to prompt the anchor to stop recommending the target object. The present application optimizes the two-way user experience of the anchor and the audience in the live broadcast room when recommending a target object.

Description

Technical Field

[0001] The present application relates to the field of electronic digital data processing, and in particular to a method for recommending objects during live broadcasting. Background Art

[0002] In video applications, more and more users can act as anchors and interact with a large number of video application users by broadcasting live in their own live broadcast rooms. Anchors can interact with a lot of information through live broadcasts, such as recommending target objects, showing their own talents, making personal comments, etc. However, when anchors recommend target objects, video applications in related technologies only provide live broadcast-related services to anchors, but cannot provide intelligent target object recommendation services, and lack technical solutions to guide anchors to make reasonable target object recommendations. Summary of the invention

[0003] The embodiment of the present application provides a method for recommending objects during live broadcasting, which can provide an intelligent target object recommendation service for the anchor and guide the anchor to make reasonable target object recommendations.

[0004] According to one aspect of an embodiment of the present application, a method for recommending an object in a live broadcast is provided, the method comprising:

[0005] Display the target object in the anchor’s corresponding live broadcast room;

[0006] Acquire the target live broadcast duration in real time, where the target live broadcast duration is the duration that the live broadcast room has occupied to recommend the target object;

[0007] When the target live broadcast duration reaches a first duration, first interaction information is obtained, where the first interaction information indicates an interaction status of the live broadcast room corresponding to the anchor within the first duration;

[0008] When the first interaction information satisfies the first interaction condition, a first live broadcast mode is started, and in the first live broadcast mode, the following operations are performed:

[0009] Displaying a plurality of interactive tags in the live broadcast room, wherein the interactive tags are explanation topics involved in the explanation process of the target object and are used to guide the audience in the live broadcast room to send corresponding bullet comments;

[0010] When the target live broadcast duration reaches a second duration, obtaining second interaction information, where the second interaction information indicates the interaction status of the live broadcast room within a preset time period, where the preset time period is obtained by subtracting the second duration from the first duration;

[0011] Inputting the second interaction information into the target model to obtain the initial time of the countdown effect;

[0012] A countdown effect is displayed in the live broadcast room. When the time in the countdown effect reaches zero, a timeout reminder message is displayed in the live broadcast room. The timeout reminder message is used to prompt the anchor to stop recommending the target object.

[0013] In one embodiment, the first interaction information includes an average number of viewers in the first duration and a total amount of bullet comments in the first duration; the method further includes:

[0014] When the average number of viewers is greater than the first number threshold, and the total amount of barrage is greater than the first barrage threshold, it is determined that the first interaction condition is met.

[0015] In one embodiment, the second interaction information includes a bullet screen set and each user account that has existed in the live broadcast room within the preset period, and the bullet screen set includes a plurality of first bullet screens carrying any interactive tag within the preset period and a plurality of second bullet screens not carrying any interactive tag; the target model is used to perform the following operations:

[0016] Counting the total number of user accounts;

[0017] For each of the user accounts, extract the historical behavior record corresponding to the user account; if the historical behavior record meets the preset behavior requirement, mark the user account with a first mark;

[0018] For each of the first bullet comments, if the first bullet comment includes at least two interactive tags, a second tag is added to the user account that posted the first bullet comment;

[0019] For each of the user accounts, if the barrage collection contains at least two barrages posted by the user account, a third mark is added to the user account;

[0020] The initial time is determined according to the total number of the user accounts and the marking status of each of the user accounts.

[0021] In one embodiment, the target model further performs the following operations:

[0022] For each of the interactive tags, the number of first bullet comments carrying the interactive tag is counted to obtain the number of first targets corresponding to the interactive tag;

[0023] Input each of the interactive tags into a preset extraction model to obtain a corresponding tag representation;

[0024] For each of the second bullet comments, the second bullet comments are input into the preset extraction model to obtain a corresponding bullet comment representation; a target label representation with the smallest distance from the bullet comment representation is determined among the label representations; if the semantic distance between the target label representation and the bullet comment representation is less than a preset semantic threshold, the second bullet comment is used as an associated bullet comment corresponding to the interactive label corresponding to the target label representation;

[0025] For each interactive tag, count the number of bullet comments associated with the interactive tag to obtain the number of second targets corresponding to the interactive tag;

[0026] For each of the interactive tags, the first target quantity corresponding to the interactive tag and the second target quantity corresponding to the interactive tag are added to obtain a reference quantity corresponding to the interactive tag;

[0027] Based on the reference number corresponding to each of the interactive tags and the initial time, the explanation time limit corresponding to each of the interactive tags is determined.

[0028] In one embodiment, the method further comprises:

[0029] According to the explanation time limit corresponding to each interactive tag, the explanation deadline corresponding to the interactive tag is determined; if the anchor is explaining the content corresponding to the interactive tag and the current time exceeds the deadline corresponding to the interactive tag, a corresponding reminder message is sent to the client corresponding to the anchor.

[0030] In one embodiment, determining the explanation time limit corresponding to each interactive tag based on the reference number corresponding to each interactive tag and the initial time includes:

[0031] According to the order in which the anchor explains the target object, the interactive tags are sorted to obtain an interactive tag sequence;

[0032] According to the reference number corresponding to each of the interactive tags in the interactive tag sequence, a time limit ratio sequence is obtained, wherein each element in the time limit ratio sequence indicates the proportion of the corresponding explanation time limit in the initial time;

[0033] Based on the time limit percentage sequence and the initial time, the explanation time limit corresponding to each of the interactive tags is determined.

[0034] In one embodiment, the method further comprises:

[0035] As the host explains, the interactive tags in the interactive tag sequence are lit up in sequence;

[0036] After the explanation of a certain interactive tag is completed, stop displaying the corresponding interactive tag.

[0037] In one embodiment, the method further comprises:

[0038] When the average number of viewers is less than a second number threshold, and the total amount of barrage is less than a second barrage threshold, determining that the first interaction information satisfies a second interaction condition;

[0039] When the first interaction information satisfies the second interaction condition, a target prompt is issued to the anchor, where the target prompt is used to indicate that the anchor is allowed to start the second live broadcast mode to reduce network resources occupied by the live broadcast.

[0040] According to one aspect of an embodiment of the present application, a computer-readable storage medium is provided, in which at least one instruction, at least one program, a code set or an instruction set is stored. 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 above-mentioned method for object recommendation in live broadcast.

[0041] The technical solution provided in the embodiments of the present application can bring the following beneficial effects:

[0042] The embodiment of the present application proposes a method for recommending objects during live broadcasting. The method for recommending objects during live broadcasting provides the following intelligent target object recommendation services:

[0043] (1) Automatic live broadcast mode switching: Intelligently select the appropriate live broadcast mode based on the interactive information obtained when displaying the target object during the live broadcast. When the interactive information indicates that the live broadcast room is very popular, the first live broadcast mode is used. The first live broadcast mode supports smart live broadcast, which includes precise control of the recommended duration of the core content of the target object and precise control of the duration of each explanation topic in the explanation process of the target object. When the interactive information indicates that the live broadcast room is not very popular, the live broadcast room is supported to switch to the second live broadcast mode. The second live broadcast mode allows the network resources occupied by the live broadcast to be reduced, so that the anchor can complete other things with the saved network resources.

[0044] (2) Propose a precise control scheme for the duration of core content recommendation for the target object: It is not good if the duration of core content recommendation for the target object is too long or too short. If it is too long, users will lose interest and reduce the popularity of the live broadcast room. If it is too short, it will be difficult to achieve the purpose of explanation and recommendation. This application uses a countdown effect combined with the display of timeout reminder information to encourage the anchor to complete the recommendation of the target object based on a reasonable core content recommendation duration.

[0045] (3) Propose an accurate modeling scheme for the duration of core content recommendations for the target object: By displaying interactive tags, guide users to participate in the interactions related to the target object's various explanation topics in advance, and pre-collect the information generated by these interactions to determine how long it is most appropriate to recommend live broadcasts to the target object, thereby planning a reasonable core content recommendation duration for the anchor.

[0046] (4) Propose a precise control plan for the duration of each explanation topic of the target object: the total recommended duration of the core content of the target object is limited. The explanation duration corresponding to each interactive tag is determined based on the interactive information related to each interactive tag. This application constrains the host to complete the explanation of the corresponding explanation topic within a reasonable time limit through display control of the interactive tag combined with sending reminder messages.

[0047] (5) The control of various durations in this application is carried out by displaying relevant information, and does not forcibly interfere with the host's live broadcast plan. Therefore, smart live broadcast focuses on guidance and providing smart assistance, and never forcibly intervenes in the host's live broadcast process. When displaying various information in this application, corresponding closing controls are provided for the host, allowing the host to close this information at any time. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0049] Figure 1 is a flow chart of a method for recommending objects during live broadcasting provided by an embodiment of the present application;

[0050] Figure 2 This is a live broadcast room diagram provided by an embodiment of the present application. Figure 1 ;

[0051] Figure 3 This is a live broadcast room diagram provided by an embodiment of the present application. Figure 2 ;

[0052] Figure 4 This is a live broadcast room diagram provided by an embodiment of the present application. Figure 3 ;

[0053] Figure 5 This is a schematic diagram of a method for determining a time limit provided by an embodiment of the present application. DETAILED DESCRIPTION

[0054] Please refer to Figure 1, which shows a flow chart of a method for recommending objects in a live broadcast provided by an embodiment of the present application. The method can be applied to a video application, and the video application can include a server and a client. The method can include the following steps:

[0055] S101. Display the target object in the live broadcast room corresponding to the anchor.

[0056] The target object can be any product introduced by the anchor.

[0057] S102. Acquire the target live broadcast duration in real time, where the target live broadcast duration is the duration that the live broadcast room has occupied to recommend the target object.

[0058] Live broadcast is a continuous process. From the moment the host starts to display the target object, the live broadcast duration recorded in real time is the target live broadcast duration. For example, when the target live broadcast duration is 1 minute, it means that the display of the target object has lasted for 1 minute. Display is also a recommendation.

[0059] S103. When the target live broadcast duration reaches a first duration, first interaction information is obtained, where the first interaction information indicates an interaction status of the live broadcast room corresponding to the anchor within the first duration.

[0060] The first duration is relatively short, which may be 2 minutes, or may be set according to the host's requirements. For example, the interaction information obtained 2 minutes before recommending the target object is used as the first interaction information. The first interaction information includes the average number of viewers in the first duration and the total amount of bullet comments in the first duration.

[0061] S104. When the first interaction information satisfies the first interaction condition, start the first live broadcast mode, and perform the following operations in the first live broadcast mode: display a plurality of interaction tags in the live broadcast room, wherein the interaction tags are explanation topics involved in the explanation process of the target object and are used to guide the audience in the live broadcast room to send corresponding barrages; when the target live broadcast duration reaches a second duration, obtain second interaction information, wherein the second interaction information indicates the interaction status of the live broadcast room within a preset time period, and the preset time period is obtained based on the difference between the second duration and the first duration; input the second interaction information into the target model to obtain the initial time of the countdown effect; display the countdown effect in the live broadcast room, and when the time in the countdown effect reaches zero, display a timeout reminder message in the live broadcast room, wherein the timeout reminder message is used to prompt the anchor to stop recommending the target object.

[0062] Specifically, when the average number of viewers is greater than the first number threshold, and the total amount of barrage is greater than the first barrage threshold, it is determined that the first interaction condition is met. This application does not limit the first number threshold and the first barrage threshold, which can be set by the anchor or automatically defaulted, and does not constitute an implementation obstacle. It can be considered that when the first interaction condition is met, the popularity of the live broadcast room is high. At this time, the anchor can be shown a control that can enter the first live broadcast mode. If the anchor clicks, the first live broadcast mode is started. Of course, the anchor can also turn off the first live broadcast mode at any time.

[0063] In the first live broadcast mode, multiple interactive tags are displayed in the live broadcast room, and the interactive tags are part or all of the explanation topics involved in the explanation process of the target object. Figure 2 , which shows the live broadcast room Figure 1 . The target object is a certain product, and the recommendation process of this product involves four explanation topics, namely "preferential plan", "product material", "product efficacy" and "brand story". These four explanation topics are preset by the anchor before the live broadcast, and the anchor will mainly recommend the target object around these four topics. When the target object is first recommended, the anchor will not directly start explaining the content of these four topics, but will conduct some routine explanations, or chat and interact with the audience in the live broadcast room. During this process, the video application will quickly determine whether to enter the first live broadcast mode, and display the interactive tags of these four topics after entering.

[0064] The audience in the live broadcast room can send corresponding barrages in response to these four interactive tags and speak freely, so that the subsequent anchor can be more targeted when explaining the corresponding topics, and better control the length of the explanation and select the content of the explanation.

[0065] The video application obtains the second interactive information based on the bullet comments sent by the audience in the live broadcast room, and can determine the initial time of the countdown effect based on the second interactive information. The initial time corresponds to the recommended duration of the core content of the target object. The explanation content corresponding to each of the four topics is the core content. After showing the countdown effect, the anchor enters the explanation of the explanation topic corresponding to the specific interactive tag.

[0066] Please refer to Figure 3 , which shows the live broadcast room Figure 2 . If the initial time is 20 minutes, the countdown starts from 20 minutes. The host can see how much time is left to explain the target object at any time, and the audience in the live broadcast room can also understand the progress of the explanation. When the countdown reaches 0 or before the countdown reaches 0, the host will explain the next product, and there is no need to send a timeout reminder message. After the recommendation of the target object is completed, the first live broadcast mode can be automatically exited. By the way, no matter which live broadcast mode the host uses, he can actively exit at any time.

[0067] like Figure 4 As shown, it shows the live broadcast room schematic Figure 3 If the host is still explaining the target object when the countdown reaches 0, a timeout reminder message is displayed in the live broadcast room, and the timeout reminder message is used to prompt the host to stop recommending the target object. This application does not interrupt the host's recommendation process. The whole process uses the method of sending relevant reminder messages to avoid mandatory influence on the host's recommendation process.

[0068] In the present application, barrages can be collected within a period of time after the interactive tag is displayed, and second interactive information can be obtained based on the collection results. For example, the second duration is 8 minutes and the first duration is 2 minutes. Then the preset time period begins after 2 minutes of live broadcast, and the preset time period lasts until the end of the 7th minute. That is, the preset time period is obtained based on the difference between the second duration and the first duration.

[0069] This application does not limit the specific method of inputting the second interaction information into the target model to obtain the initial time of the countdown effect. The target model is a model for calculating the recommended duration of the core content. In one embodiment, the second interaction information includes a barrage set and each user account that has existed in the live broadcast room during the preset period of time. The barrage set includes multiple first barrages carrying any interactive tags during the preset period of time and multiple second barrages that do not carry any interactive tags; the target model is used to perform the following operations:

[0070] (1) Counting the total number of user accounts;

[0071] (2) for each of the user accounts, extracting a historical behavior record corresponding to the user account; if the historical behavior record meets a preset behavior requirement, marking the user account with a first mark;

[0072] For example, if a user account has a record of consumption in the live broadcast room in the past month, it is considered to meet the preset behavior requirements.

[0073] (3) For each of the first bullet comments, if the first bullet comment includes at least two interactive tags, a second tag is added to the user account that posted the first bullet comment;

[0074] (4) For each of the user accounts, if the barrage collection contains at least two barrages posted by the user account, a third mark is added to the user account;

[0075] (5) Determining the initial time according to the total number of the user accounts and the status of the marking of each of the user accounts.

[0076] The embodiments of the present application do not limit the specific method of determining the initial time according to the total amount of the user accounts and the labeling of each of the user accounts. For example, it can be determined by fitting historical data or set based on manual experience, which does not constitute an implementation obstacle. In a specific implementation scheme, the number of all first tags, second tags, and third tags can be added together to obtain a tag statistic value, and the tag statistic value can be added to the total amount of the user accounts to obtain an interactive reference amount. This tag statistic directly performs a positive correlation flattening fit on the probability of the group formed by each user account purchasing the target object, which not only models the probability of each user account purchasing the target object, but also directly obtains the probability of the group purchasing the target object. For example, if a user account has the first tag, the second tag, and the third tag at the same time, and the purchase probability of the user account is very high, then the user account contributes "3" to the tag statistic value; if a user account has the first tag and the third tag at the same time, and the purchase probability of the user account is also very high, then the user account contributes "2" to the tag statistic value; if a user account has only one tag, and the purchase probability of the user account is ordinary, then the user account contributes "1" to the tag statistic value. Therefore, the probability of a group purchasing the target object can be represented simply by the tag statistic value. This is a flattened fitting. There is no need to calculate the probability of each user account purchasing the target object. Just calculate a tag statistic value directly, which significantly speeds up. The tag statistic value is added to the total amount of the user account to obtain the interactive reference amount. The mapping relationship between the interactive reference amount of the history modeling and the historical core content recommendation duration of the target object is directly based on artificial experience. The initial time can be determined based on the mapping relationship. This mapping relationship can be represented by a table or a curve through artificial experience, and does not constitute an implementation obstacle.

[0077] In one embodiment, please refer to Figure 5 , which shows a schematic diagram of explaining the time limit determination method. The target model also performs the following operations:

[0078] S501. For each interactive tag, count the number of first bullet comments carrying the interactive tag to obtain the number of first targets corresponding to the interactive tag;

[0079] Each first bullet screen can carry at least one interactive tag. For example, a first bullet screen can be "'Discount plan', 'Product material' Is this product pure cotton? How can I buy it at the cheapest price?" This first bullet screen reflects the topic that the live broadcast audience who sent the first bullet screen wants the anchor to focus on. "Discount plan" and "Product material" are the interactive tags actively selected by the live broadcast audience. This first bullet screen contributes "1" to the first target number of the interactive tag "Discount plan" and also contributes "1" to the first target number of the interactive tag "Product material".

[0080] S502. Input each of the interactive tags into a preset extraction model to obtain a corresponding tag representation; for each of the second bullet comments, input the second bullet comment into the preset extraction model to obtain a corresponding bullet comment representation; determine the target tag representation with the smallest distance from the bullet comment representation among the tag representations; if the semantic distance between the target tag representation and the bullet comment representation is less than a preset semantic threshold, use the second bullet comment as the associated bullet comment corresponding to the interactive tag corresponding to the target tag representation;

[0081] The preset extraction model is a model that can extract representational information of interactive tags. Any model with feature extraction function in the prior art can be used as the preset extraction model without posing any implementation obstacle.

[0082] The embodiment of the present application does not limit the method of obtaining the semantic distance, and the existing technology can be used. There is no limit on the preset semantic threshold, and it does not constitute an implementation obstacle. It can be considered that step S501 is that the live broadcast room audience classifies the barrage sent by themselves to see which interactive tag it belongs to. And step S502 is that the live broadcast room audience did not classify the barrage sent by themselves, so it is automatically classified by semantic distance to see which interactive tag it belongs to. Of course, if a second barrage is semantically different from each interactive tag, it will not be used as an associated barrage of any interactive tag.

[0083] S503. For each interactive tag, count the number of barrages associated with the interactive tag to obtain the second target number corresponding to the interactive tag; for each interactive tag, add the first target number corresponding to it and the second target number corresponding to it to obtain the reference number corresponding to the interactive tag;

[0084] The reference quantity indicates the number of barrages related to the corresponding interactive tag, including barrages directly classified by the live broadcast room audience and associated barrages obtained based on semantic classification.

[0085] S504. Based on the reference number corresponding to each of the interactive tags and the initial time, determine the explanation time limit corresponding to each of the interactive tags.

[0086] Specifically, the method of determining the explanation time limit corresponding to each interactive tag based on the reference number corresponding to each interactive tag and the initial time includes: sorting the interactive tags according to the order of the host's explanation of the target object to obtain an interactive tag sequence; obtaining a time limit proportion sequence according to the reference number corresponding to each interactive tag in the interactive tag sequence, wherein each element in the time limit proportion sequence indicates the proportion of the corresponding explanation time limit in the initial time; determining the explanation time limit corresponding to each interactive tag based on the time limit proportion sequence and the initial time.

[0087] For example, if the reference quantities of "preferential plan", "product material", "product efficacy" and "brand story" are rounded to an integer ratio of 1:2:2:1, that is, the number of parameters of "product material" and "product efficacy" is similar, which is basically twice the number of parameters of "preferential plan" and "brand story", then the initial time can be roughly divided into 6 parts, and the explanation time limits of "preferential plan", "product material", "product efficacy" and "brand story" respectively account for 1 part, 2 parts, 2 parts and 1 part. Of course, the proportions and explanation time limits can be approximately rounded. This process is not elaborated and does not constitute an implementation obstacle.

[0088] Along with the host's explanation process, the interactive tags in the interactive tag sequence can be lit up in order in response to the host's operation; after the explanation of a certain interactive tag is completed, the corresponding interactive tag will stop being displayed. According to the explanation time limit corresponding to each interactive tag, the explanation deadline corresponding to the interactive tag is determined; if the host is explaining the content corresponding to the interactive tag and the current time exceeds the deadline corresponding to the interactive tag, a corresponding reminder message is sent to the client corresponding to the host. The reminder message is used to remind the host that the explanation of the explanation topic corresponding to the corresponding interactive tag has timed out, so that the host can adjust the recommendation progress in a timely and flexible manner.

[0089] In one embodiment, when the average number of viewers is less than the second number threshold, and the total amount of barrage is less than the second barrage threshold, it is determined that the first interaction information meets the second interaction condition; this application does not limit the second number threshold and the second barrage threshold, and does not constitute an implementation obstacle. It can be considered that when the second interaction condition is met, the live broadcast room is less popular. When the first interaction information meets the second interaction condition, a target prompt is issued to the anchor, and the target prompt is used to indicate that the anchor is allowed to start the second live broadcast mode to reduce the network resources occupied by the live broadcast.

[0090] If the popularity of the live broadcast room is very low, the anchor can choose to reduce the occupied network resources. This application is not mandatory, but only prompts the anchor to have an additional choice. The anchor can use the saved network resources for other purposes at will. This application does not limit this, nor is it the focus of this application, and does not constitute an obstacle to the implementation of this application. This application also does not limit the specific scheme for reducing the network resources occupied by the live broadcast in the second live broadcast mode. For example, the live video can be compressed to reduce the network resources occupied by the live video transmission. Traditional video compression methods have a great impact on video quality and may have a great adverse impact on the live broadcast effect. In this regard, this application provides a new video compression method that significantly reduces the network resources occupied by the live broadcast with little impact on the video quality. The method compresses the original live video stream generated by the anchor in real time to obtain a compressed video stream, transmits the compressed video stream to replace the original live video stream to reduce the occupied network resources, and decompresses the compressed video stream on the client corresponding to the audience in the live broadcast room, and plays the decompression result. During real-time compression, the following compression operation is performed on each image of the original live video stream:

[0091] (10) Adaptively and scalably compress the image to obtain a compressed image. Divide the compressed image into blocks to obtain multiple image blocks. Of course, this application does not elaborate on the block division method, which does not constitute an implementation obstacle.

[0092] (20) Perform the following operations on each image block:

[0093] The image block is divided into multiple image regions. This application does not elaborate on the method of dividing the image regions, which does not constitute an implementation obstacle. The middle image region of the multiple image regions is determined as the source image region, and a similarity threshold is determined. The size of the similarity threshold can be set according to actual conditions, which does not constitute an implementation obstacle.

[0094] For each image region except the source image region in the image block, the following operations are performed in turn: a copy image region that is completely identical to the image region is generated, and the color weight of the copy image region is replaced with the color weight of the source image region to obtain a reference image region; adaptive scalable texture decompression is performed on the image region and the corresponding reference image region respectively to obtain a first image region and a second image region respectively; the similarity between the first image region and the second image region is calculated, and if the similarity is greater than the similarity threshold, the reference image region is used to replace the image region in the image block, and the similarity threshold is set to the similarity.

[0095] (30) Performing Zstd lossless compression on each of the image blocks to obtain a final compression result corresponding to the image, thereby completing the compression of the image.

[0096] In an exemplary embodiment, a computer-readable storage medium is further provided, wherein the storage medium stores at least one instruction, at least one program, code set or instruction set, and when the at least one instruction, at least one program, code set or instruction set is executed by a processor, the method for recommending objects in a live broadcast is implemented. The method includes:

[0097] Display the target object in the anchor’s corresponding live broadcast room;

[0098] Acquire the target live broadcast duration in real time, where the target live broadcast duration is the duration that the live broadcast room has occupied to recommend the target object;

[0099] When the target live broadcast duration reaches a first duration, first interaction information is obtained, where the first interaction information indicates an interaction status of the live broadcast room corresponding to the anchor within the first duration;

[0100] When the first interaction information satisfies the first interaction condition, a first live broadcast mode is started, and in the first live broadcast mode, the following operations are performed:

[0101] Displaying a plurality of interactive tags in the live broadcast room, wherein the interactive tags are explanation topics involved in the explanation process of the target object and are used to guide the audience in the live broadcast room to send corresponding bullet comments;

[0102] When the target live broadcast duration reaches a second duration, obtaining second interaction information, where the second interaction information indicates the interaction status of the live broadcast room within a preset time period, where the preset time period is obtained by subtracting the second duration from the first duration;

[0103] Inputting the second interaction information into the target model to obtain the initial time of the countdown effect;

[0104] A countdown effect is displayed in the live broadcast room. When the time in the countdown effect reaches zero, a timeout reminder message is displayed in the live broadcast room. The timeout reminder message is used to prompt the anchor to stop recommending the target object.

[0105] In one embodiment, the first interaction information includes an average number of viewers in the first duration and a total amount of bullet comments in the first duration; the method further includes:

[0106] When the average number of viewers is greater than the first number threshold, and the total amount of barrage is greater than the first barrage threshold, it is determined that the first interaction condition is met.

[0107] In one embodiment, the second interaction information includes a bullet screen set and each user account that has existed in the live broadcast room within the preset period, and the bullet screen set includes a plurality of first bullet screens carrying any interactive tag within the preset period and a plurality of second bullet screens not carrying any interactive tag; the target model is used to perform the following operations:

[0108] Counting the total number of user accounts;

[0109] For each of the user accounts, extract the historical behavior record corresponding to the user account; if the historical behavior record meets the preset behavior requirement, mark the user account with a first mark;

[0110] For each of the first bullet comments, if the first bullet comment includes at least two interactive tags, a second tag is added to the user account that posted the first bullet comment;

[0111] For each of the user accounts, if the barrage collection contains at least two barrages posted by the user account, a third mark is added to the user account;

[0112] The initial time is determined according to the total number of the user accounts and the marking status of each of the user accounts.

[0113] In one embodiment, the target model further performs the following operations:

[0114] For each of the interactive tags, the number of first bullet comments carrying the interactive tag is counted to obtain the number of first targets corresponding to the interactive tag;

[0115] Input each of the interactive tags into a preset extraction model to obtain a corresponding tag representation;

[0116] For each of the second bullet comments, the second bullet comments are input into the preset extraction model to obtain a corresponding bullet comment representation; a target label representation with the smallest distance from the bullet comment representation is determined among the label representations; if the semantic distance between the target label representation and the bullet comment representation is less than a preset semantic threshold, the second bullet comment is used as an associated bullet comment corresponding to the interactive label corresponding to the target label representation;

[0117] For each interactive tag, count the number of bullet comments associated with the interactive tag to obtain the number of second targets corresponding to the interactive tag;

[0118] For each of the interactive tags, the first target quantity corresponding to the interactive tag and the second target quantity corresponding to the interactive tag are added to obtain a reference quantity corresponding to the interactive tag;

[0119] Based on the reference number corresponding to each of the interactive tags and the initial time, the explanation time limit corresponding to each of the interactive tags is determined.

[0120] In one embodiment, the method further comprises:

[0121] According to the explanation time limit corresponding to each interactive tag, the explanation deadline corresponding to the interactive tag is determined; if the anchor is explaining the content corresponding to the interactive tag and the current time exceeds the deadline corresponding to the interactive tag, a corresponding reminder message is sent to the client corresponding to the anchor.

[0122] In one embodiment, determining the explanation time limit corresponding to each interactive tag based on the reference number corresponding to each interactive tag and the initial time includes:

[0123] According to the order in which the anchor explains the target object, the interactive tags are sorted to obtain an interactive tag sequence;

[0124] According to the reference number corresponding to each of the interactive tags in the interactive tag sequence, a time limit ratio sequence is obtained, wherein each element in the time limit ratio sequence indicates the proportion of the corresponding explanation time limit in the initial time;

[0125] Based on the time limit percentage sequence and the initial time, the explanation time limit corresponding to each of the interactive tags is determined.

[0126] In one embodiment, the method further comprises:

[0127] As the host explains, the interactive tags in the interactive tag sequence are lit up in sequence;

[0128] After the explanation of a certain interactive tag is completed, stop displaying the corresponding interactive tag.

[0129] In one embodiment, the method further comprises:

[0130] When the average number of viewers is less than a second number threshold, and the total amount of barrage is less than a second barrage threshold, determining that the first interaction information satisfies a second interaction condition;

[0131] When the first interaction information satisfies the second interaction condition, a target prompt is issued to the anchor, where the target prompt is used to indicate that the anchor is allowed to start the second live broadcast mode to reduce network resources occupied by the live broadcast.

[0132] Optionally, the computer-readable storage medium may include: ROM (Read Only Memory), RAM (Random Access Memory), SSD (Solid State Drives) or an optical disk, etc. Among them, the random access memory may include ReRAM (Resistance Random Access Memory) and DRAM (Dynamic Random Access Memory).

[0133] It should be understood that the "multiple" mentioned in this article refers to two or more than two. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the objects associated with each other are in an "or" relationship. In addition, the step numbers described in this article only illustrate a possible execution sequence between the steps. In some other embodiments, the above steps may not be executed in the order of the numbers, such as two steps with different numbers are executed at the same time, or two steps with different numbers are executed in the opposite order to the diagram, or a step is not executed. The embodiments of the present application do not limit this.

[0134] In addition, in the specific implementation of this application, account information and other related data are involved. When the above embodiments of this application are applied to specific products or technologies, account permission or consent is required, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards of relevant countries and regions.

[0135] The above are merely exemplary embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for recommending objects in live broadcast, characterized in that: The method comprises: Display the target object in the anchor’s corresponding live broadcast room; Acquire the target live broadcast duration in real time, where the target live broadcast duration is the duration that the live broadcast room has occupied to recommend the target object; When the target live broadcast duration reaches a first duration, first interaction information is obtained, where the first interaction information indicates an interaction status of the live broadcast room corresponding to the anchor within the first duration; When the first interaction information satisfies the first interaction condition, a first live broadcast mode is started, and in the first live broadcast mode, the following operations are performed: Displaying a plurality of interactive tags in the live broadcast room, wherein the interactive tags are explanation topics involved in the explanation process of the target object and are used to guide the audience in the live broadcast room to send corresponding bullet comments; When the target live broadcast duration reaches a second duration, obtaining second interaction information, where the second interaction information indicates the interaction status of the live broadcast room within a preset time period, where the preset time period is obtained by subtracting the second duration from the first duration; Inputting the second interaction information into the target model to obtain the initial time of the countdown effect; A countdown effect is displayed in the live broadcast room. When the time in the countdown effect reaches zero, a timeout reminder message is displayed in the live broadcast room. The timeout reminder message is used to prompt the anchor to stop recommending the target object; The first interaction information includes an average number of viewers in the first duration and a total amount of barrage in the first duration; when the average number of viewers is greater than a first number threshold, and the total amount of barrage is greater than a first barrage threshold, it is determined that the first interaction condition is met; The second interaction information includes a bullet screen set and each user account that has existed in the live broadcast room within the preset period, and the bullet screen set includes a plurality of first bullet screens carrying any interactive tag within the preset period and a plurality of second bullet screens not carrying any interactive tag; the target model is used to perform the following operations: Counting the total number of user accounts; For each of the user accounts, extract the historical behavior records corresponding to the user account; if the historical behavior records meet the preset behavior requirements, mark the user account with a first mark; if the user account has a consumption record in the live broadcast room in the last month, it is considered that it meets the preset behavior requirements; For each of the first bullet comments, if the first bullet comment includes at least two interactive tags, a second tag is added to the user account that posted the first bullet comment; For each of the user accounts, if the barrage collection contains at least two barrages posted by the user account, a third mark is added to the user account; Determining the initial time according to the total number of the user accounts and the status of each of the user accounts being marked; The target model also performs the following operations: For each of the interactive tags, the number of first bullet comments carrying the interactive tag is counted to obtain the number of first targets corresponding to the interactive tag; Input each of the interactive tags into a preset extraction model to obtain a corresponding tag representation; For each of the second bullet comments, the second bullet comments are input into the preset extraction model to obtain a corresponding bullet comment representation; a target label representation with the smallest distance from the bullet comment representation is determined among the label representations; if the semantic distance between the target label representation and the bullet comment representation is less than a preset semantic threshold, the second bullet comment is used as an associated bullet comment corresponding to the interactive label corresponding to the target label representation; For each interactive tag, count the number of bullet comments associated with the interactive tag to obtain the number of second targets corresponding to the interactive tag; For each of the interactive tags, the first target quantity corresponding to the interactive tag and the second target quantity corresponding to the interactive tag are added to obtain a reference quantity corresponding to the interactive tag; Based on the reference number corresponding to each of the interactive tags and the initial time, the explanation time limit corresponding to each of the interactive tags is determined.

2. The method according to claim 1, characterized in that The method further comprises: According to the explanation time limit corresponding to each interactive tag, the explanation deadline corresponding to the interactive tag is determined; if the anchor is explaining the content corresponding to the interactive tag and the current time exceeds the deadline corresponding to the interactive tag, a corresponding reminder message is sent to the client corresponding to the anchor.

3. The method according to claim 1 or 2, characterized in that: The determining the explanation time limit corresponding to each interactive tag based on the reference number corresponding to each interactive tag and the initial time includes: According to the order in which the anchor explains the target object, the interactive tags are sorted to obtain an interactive tag sequence; According to the reference number corresponding to each of the interactive tags in the interactive tag sequence, a time limit ratio sequence is obtained, wherein each element in the time limit ratio sequence indicates the proportion of the corresponding explanation time limit in the initial time; Based on the time limit percentage sequence and the initial time, the explanation time limit corresponding to each of the interactive tags is determined.

4. The method according to claim 1 or 2, characterized in that: The method further comprises: As the host explains, the interactive tags in the interactive tag sequence are lit up in sequence; After the explanation of a certain interactive tag is completed, stop displaying the corresponding interactive tag.

5. The method according to claim 1, characterized in that The method further comprises: When the average number of viewers is less than a second number threshold, and the total amount of barrage is less than a second barrage threshold, determining that the first interaction information satisfies a second interaction condition; When the first interaction information satisfies the second interaction condition, a target prompt is issued to the anchor, where the target prompt is used to indicate that the anchor is allowed to start the second live broadcast mode to reduce network resources occupied by the live broadcast.

6. A computer-readable storage medium, characterized in that: The storage medium stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, the at least one program, the code set or the instruction set are loaded and executed by the processor to implement the method for object recommendation in live broadcast as described in any one of claims 1 to 5.

Images