Media content recommendation methods, devices, electronic devices, and readable storage media

Abstract

This application provides a media content recommendation method, apparatus, electronic device, and readable storage medium. The method includes: acquiring user feedback information on target media content; if the feedback information does not match preset feedback information, determining a correction coefficient for the target media content based on the feedback information and the preset feedback information, wherein the preset feedback information is associated with the display information of the target media content on the client side, and the display information includes display location information and / or display time information; and correcting the recommendation parameters of the target media content based on the correction coefficient. This application can improve the flexibility of media content recommendation.

Description

Technical Field

[0001] This application relates to the field of content recommendation technology, and in particular to a media content recommendation method, apparatus, electronic device, and readable storage medium. Background Technology

[0002] In recent years, with the rapid development of computer and internet technologies, the types of media content have become increasingly diverse, and how to reasonably recommend media content to users has become an important research direction.

[0003] Currently, in order to attract users and increase user engagement, media websites typically select a batch of media content to recommend to all users based on current events or market demands, in order to increase the exposure of this content. However, this content may include material that users are not interested in, which will occupy unnecessary network resources and waste the traffic generated by certain exposure slots. Summary of the Invention

[0004] The purpose of this application is to provide a media content recommendation method, apparatus, electronic device, and readable storage medium, which solves the problem of wasted recommendation resources in the prior art.

[0005] To address the aforementioned problems, in a first aspect, embodiments of this application provide a media content recommendation method, including:

[0006] Obtain user feedback on target media content;

[0007] If the feedback information does not match the preset feedback information, a correction coefficient for the target media content is determined based on the feedback information and the preset feedback information. The preset feedback information is associated with the display information of the target media content on the client, and the display information includes display location information and / or display time information.

[0008] The recommendation parameters for the target media content are adjusted based on the correction coefficient.

[0009] Optionally, before obtaining user feedback on the target media content, the method further includes:

[0010] Target users are determined based on the content tag information of the target media content, wherein the user tag information of the target users is matched with the content tag information of the target media content.

[0011] The process of obtaining user feedback on target media content includes:

[0012] Obtain feedback information from the target user regarding the target media content.

[0013] Optionally, determining the correction coefficient for the target media content based on the feedback information and the preset feedback information when the feedback information does not match includes:

[0014] Obtain the user's first feedback information on the target media content within the first time period;

[0015] If the first feedback information does not match the first preset feedback information, a first correction coefficient for the target media content is determined based on the first feedback information and the first preset feedback information, wherein the first preset feedback information is the preset feedback information corresponding to the first time period.

[0016] The step of correcting the recommendation parameters of the target media content according to the correction coefficient includes:

[0017] Based on the first correction coefficient, the recommendation parameters of the target media content in the second time period are corrected. The second time period is continuous with the first time period and is after the first time period.

[0018] Optionally, the feedback information includes the number of clicks, and the preset feedback information includes a preset number of clicks;

[0019] In the case where the feedback information does not match the preset feedback information, determining the correction coefficient for the target media content based on the feedback information and the preset feedback information includes:

[0020] If the number of clicks does not match the preset number of clicks, a correction coefficient for the target media content is determined based on the ratio of the number of clicks to the preset number of clicks.

[0021] Optionally, the preset number of clicks is determined according to the following formula:

[0022] click j =exp(a·j b +c)

[0023] Among them, click j The preset click count is denoted as j, where j is the display order of the target media content on the client, and the values ​​of a, b, and c are determined based on the display time of the target media content on the client.

[0024] Optionally, the step of correcting the recommendation parameters of the target media content according to the correction coefficient includes:

[0025] Obtain the first recommendation score of the target media content before correction;

[0026] The second recommendation score of the target media content is determined by multiplying the correction coefficient by the first recommendation score.

[0027] Optionally, before obtaining user feedback on the target media content, the method further includes:

[0028] Determine the primary media content;

[0029] Based on a pre-defined recall strategy, the content of the first media outlet is recalled;

[0030] Based on a preset filtering strategy, the target media content is determined from the first media content.

[0031] Optionally, the primary media content is determined, including:

[0032] Identify candidate media content;

[0033] Based on the feature information of each candidate media content, the estimated traffic information of each candidate media content is determined;

[0034] Based on the estimated traffic information of each candidate media content, a first media content is determined from the candidate media content.

[0035] Optionally, recalling the first media content based on a preset recall strategy includes:

[0036] Based on the identification information of the first media content, the first media content is recalled.

[0037] Secondly, embodiments of this application provide a media content recommendation device, including:

[0038] The first acquisition module is used to acquire user feedback information on the target media content;

[0039] The first determining module is used to determine the correction coefficient of the target media content based on the feedback information and the preset feedback information when the feedback information does not match the preset feedback information. The preset feedback information is associated with the display information of the target media content on the client, and the display information includes display location information and / or display time information.

[0040] The correction module is used to correct the recommendation parameters of the target media content according to the correction coefficient.

[0041] Optionally, the device further includes:

[0042] The second determining module is used to determine the target user based on the content tag information of the target media content, wherein the user tag information of the target user is matched with the content tag information of the target media content.

[0043] The first acquisition module is used for:

[0044] Obtain feedback information from the target user regarding the target media content.

[0045] Optionally, the first determining module includes:

[0046] The first acquisition unit is used to acquire the first feedback information of the user on the target media content within a first time period;

[0047] The first determining unit is configured to determine a first correction coefficient for the target media content based on the first feedback information and the first preset feedback information when the first feedback information does not match the first preset feedback information. The first preset feedback information is the preset feedback information corresponding to the first time period.

[0048] The correction module is used for:

[0049] Based on the first correction coefficient, the recommendation parameters of the target media content in the second time period are corrected. The second time period is continuous with the first time period and is after the first time period.

[0050] Optionally, the feedback information includes the number of clicks, and the preset feedback information includes a preset number of clicks;

[0051] The first determining module is used for:

[0052] If the number of clicks does not match the preset number of clicks, a correction coefficient for the target media content is determined based on the ratio of the number of clicks to the preset number of clicks.

[0053] Optionally, the preset number of clicks is determined according to the following formula:

[0054] click j =exp(a·j b +c)

[0055] Among them, click j The preset click count is denoted as j, where j is the display order of the target media content on the client, and the values ​​of a, b, and c are determined based on the display time of the target media content on the client.

[0056] Optionally, the correction module includes:

[0057] The second acquisition unit is used to acquire the first recommendation score of the target media content before correction;

[0058] The second determining unit is used to determine the second recommended score of the target media content after correction based on the product of the correction coefficient and the first recommended score.

[0059] Optionally, the device further includes:

[0060] The third determining module is used to determine the content of the first media.

[0061] The recall module is used to recall the first media content based on a preset recall strategy;

[0062] The filtering module is used to determine the target media content in the first media content based on a preset filtering strategy.

[0063] Optionally, the third determining module includes:

[0064] The fourth determining unit is used to determine candidate media content;

[0065] The fifth determining unit is used to determine the estimated traffic information of each candidate media content based on the feature information of each candidate media content;

[0066] The sixth determining unit is used to determine the first media content among the candidate media content based on the estimated traffic information of each candidate media content.

[0067] Optionally, the recall module is used for:

[0068] Based on the identification information of the first media content, the first media content is recalled.

[0069] Thirdly, embodiments of this application provide an electronic device, including a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the media content recommendation method as described above.

[0070] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the media content recommendation method described above.

[0071] In this embodiment of the application, after recommending predetermined target media content to the client, the server can obtain user feedback information on the target media content and adjust the recommendation parameters of the target media content so that the target media content can receive sufficient exposure and promotion and have recommendation parameters that match the user's level of interest, thereby avoiding the occupation of unnecessary recommendation resources and improving the flexibility of media content recommendation. Attached Figure Description

[0072] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments of this application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0073] Figure 1 This is one of the flowcharts of the media content recommendation method provided in the embodiments of this application;

[0074] Figure 2 This is a schematic diagram illustrating the relationship between display order and click volume provided in an embodiment of this application;

[0075] Figure 3 This is the second flowchart of the media content recommendation method provided in the embodiments of this application;

[0076] Figure 4 This is a structural diagram of the media content recommendation device provided in the embodiments of this application;

[0077] Figure 5 This is a structural diagram of the electronic device provided in the embodiments of this application. Detailed Implementation

[0078] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0079] The terms "first," "second," etc., used in this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to these processes, methods, products, or apparatuses. Additionally, the use of "and / or" in this application indicates at least one of the connected objects, such as A and / or B and / or C, representing seven possibilities: including A alone, B alone, C alone, and the presence of both A and B, both B and C, both A and C, and the presence of A, B, and C.

[0080] In the embodiments of this application, the terms "exemplary" or "for example" are used to indicate that something is an example, illustration, or description. Any embodiment or design that is described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design. Specifically, the use of the terms "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0081] Please see Figure 1 , Figure 1 This is a flowchart illustrating a media content recommendation method provided in an embodiment of this application. The media content recommendation method can be executed by a server-side device, which can be a cloud-based computer, server, or other device or data platform with data processing capabilities. Here, the method is described using a server executing the media content recommendation method as an example.

[0082] like Figure 1 As shown, the media content recommendation method may include the following steps:

[0083] Step 101: Obtain user feedback on the target media content.

[0084] Media content can also be called media resources (or simply media content), media data, etc. Specifically, it can include various types such as news, advertisements, videos, audio, merchants, and products. Among them, videos can include TV series, movies, documentaries, variety shows, and other video content.

[0085] The target media content refers to pre-determined media content that has been recommended to the client to increase exposure. The server can obtain user feedback information on the target media content from the client. The aforementioned client can be a client on a mobile electronic device such as a mobile phone or tablet, or a client on a large-screen electronic device such as a television or projector. The specific client can be determined according to the actual situation, and this application embodiment does not limit it.

[0086] In practice, the target media content can be predetermined based on current events or market demands. For example, if National Day is approaching, the target media content could be video content on themes such as National Day and patriotism to express celebration; or, if actor A has passed away, the target media content could be video content corresponding to actor A's representative film and television works to express remembrance.

[0087] The server can recommend the target media content to the client. For example, the target media content can be displayed on the client's homepage, or it can be displayed on the client's search results page when a keyword associated with the target media content is triggered. The server can monitor the target media content. Optionally, the server can monitor the target media content based on its identification information, such as its name or number, and collect user feedback on the target media content from the client.

[0088] The feedback information may include positive feedback from the user regarding the target media content, that is, feedback information indicating the user's interest in the target media content. In one example, the positive feedback information may include information about the user's click operation on the target media content, such as the number of clicks, or the click frequency or click-through rate within a certain time period; it may also include information about the user's actions such as saving, liking, or donating coins to the target media content. The feedback information may also include negative feedback from the user regarding the target media content, that is, feedback information indicating the user's lack of interest in the target media content. In one example, the negative feedback information may include information about the user's actions such as showing disinterest, reducing recommendations, or disliking the target media content. In addition, the feedback information may also include other feedback information from the user regarding the target media content, such as information about comments, ratings, or sending bullet comments. It is understood that the feedback information may include any user's response information to the target media content, which can be determined according to the actual situation, and this application embodiment does not limit it.

[0089] The feedback information can reflect the user's level of interest in the target media content, and the server can determine the user traffic that the target media content can bring based on the feedback information. Afterwards, the server can continue to execute step 102.

[0090] Step 102: If the feedback information does not match the preset feedback information, determine the correction coefficient of the target media content based on the feedback information and the preset feedback information. The preset feedback information is associated with the display information of the target media content on the client. The display information includes display location information and / or display time information.

[0091] In practice, the server can determine the preset feedback information—that is, the expected user traffic that the current display information of the target media content on the client can bring—based on the target media content's current display information on the client. The display information may include the target media content's display position on the client, such as its position on the page, its display order, or the size of the display area; it may also include the display time information of the target media content on the client. Different display information provides different levels of exposure, and therefore, different expected user traffic.

[0092] The server can compare the received feedback information with preset feedback information to determine whether the actual user traffic brought by the target media content meets expectations. If the feedback information does not match the preset feedback information, it means that the user's level of interest in the target media content does not match the exposure provided by the currently displayed information. Taking the display order of the target media content on the client as an example, if the feedback information of the target media content is lower than expected, it means that the user traffic that the target media content can bring is lower than expected. If it is continuously placed in a high display order, it will waste the exposure provided by the high display order and occupy unnecessary network resources. If the feedback information of the target media content is higher than expected, it means that the user traffic that the target media content can bring is higher than expected. If it is continuously placed in a low display order, it will suppress the exposure of the target media content and lose some of the user traffic that the target media content can bring.

[0093] If the feedback information does not match the preset feedback information, the server can correct the recommendation parameters of the target media content. The recommendation parameters may include recommendation score and / or recommendation probability.

[0094] Step 103: Adjust the recommended parameters of the target media content according to the correction coefficient.

[0095] In specific implementation, the server can obtain the current recommendation score and / or recommendation probability of the target media content. Taking the recommendation score as an example, the server can determine the second recommendation score of the target media content after correction based on the correction coefficient and the first recommendation score of the target media content before correction. In an optional implementation, the server can determine the second recommendation score based on the product of the correction coefficient and the first recommendation score. When the target media content is first delivered, the first recommendation score can be determined based on the experience of the operators. Alternatively, in an optional implementation, the first recommendation score can be determined based on recommendation score calculation methods in related technologies. For example, the recommendation score can be calculated based on the Upper Confidence Bound Algorithm (UCB), as shown in the following formula:

[0096]

[0097] Among them, score i C is the first recommendation score for the target media content. i,t T represents the number of times the target media content is clicked in t experiments. i,t The first recommendation score is the number of times the target media content is recommended, or in other words, the number of times it is displayed, across t experiments. It is understood that the implementation of determining this first recommendation score is not limited thereto, and the embodiments of this application do not limit it.

[0098] After determining the revised recommendation parameters, the server can update the recommendation algorithm based on the modified parameters. This allows the recommendation algorithm to make recommendations for the target media content based on the revised parameters, thereby redetermining the display information of the target media content on the client side.

[0099] In this embodiment of the application, after recommending predetermined target media content to the client, the server can obtain user feedback information on the target media content and adjust the recommendation parameters of the target media content so that the target media content can receive sufficient exposure and promotion and have recommendation parameters that match the user's level of interest, thereby avoiding the occupation of unnecessary recommendation resources and improving the flexibility of media content recommendation.

[0100] In this embodiment of the application, the target media content can be understood as media content that has been recommended to the client from among the pre-determined media content that needs to increase exposure or needs to be promoted.

[0101] Optionally, before step 101, the method further includes:

[0102] Determine the primary media content;

[0103] Based on a pre-defined recall strategy, the content of the first media outlet is recalled;

[0104] Based on a preset filtering strategy, the target media content is determined from the first media content.

[0105] In this embodiment, the process of determining the target media content may include three stages: determination, recall, and filtering. In practical applications, the filtering stage may specifically include two steps: sorting and filtering. These will be described below:

[0106] 1) The determination phase is used to identify the primary media content. The primary media content is pre-determined media content that needs increased exposure or promotion, and can be considered as content requiring traffic support.

[0107] In one alternative implementation, determining the first media content includes:

[0108] Identify candidate media content;

[0109] Based on the feature information of each candidate media content, the estimated traffic information of each candidate media content is determined;

[0110] Based on the estimated traffic information of each candidate media content, a first media content is determined from the candidate media content.

[0111] In practice, the selection range of candidate media content can be predetermined. For example, if National Day is approaching, to express celebration, the selection range of candidate media content could be patriotic TV dramas and movies released after 2000. After determining the candidate media content, the traffic information of each candidate media content can be estimated, and this traffic information is used to characterize the ability of the candidate media content to attract users.

[0112] The characteristic information of the candidate media content may include: quality characteristic information, popularity characteristic information, content characteristic information, production characteristic information, and promotional characteristic information. The quality characteristic information is used to characterize the quality of the candidate media content. Optionally, the quality characteristic information includes a quality score, which the server can determine based on information such as playback duration, rating, and reviews. The popularity characteristic information is used to characterize the popularity of the candidate media content. Optionally, the popularity characteristic information includes a popularity score, which the server can determine based on information such as playback duration, clicks, discussion volume, and number of viewers. The content characteristic information is used to characterize the content of the candidate media content and may include characteristics such as genre content, plot content, and emotional content. The production characteristic information is used to characterize the production of the candidate media content and may include characteristics of the production personnel, such as actors, directors, and screenwriters, as well as characteristics of the production team, such as production company and post-production company. The promotional characteristic information is used to characterize the promotion of the candidate media content. It is understood that the feature information of the candidate media content is not limited to this, and can be determined according to the actual situation. This application embodiment does not limit it here.

[0113] The server can evaluate the traffic information of each candidate media content based on the above feature information.

[0114] In one optional implementation, the traffic information includes traffic levels; the higher the traffic level of the candidate media content, the stronger its ability to attract users; the lower the traffic level of the candidate media content, the weaker its ability to attract users. The server can determine the first candidate media content from the candidate media content with higher traffic levels.

[0115] In another optional implementation, the traffic information includes traffic scores. For example, the server can input the feature information of each candidate media content into a pre-trained traffic evaluation model and obtain the traffic score output by the model. The server can determine the first media content from candidate media content whose traffic scores meet preset conditions. For instance, the server can select the top N candidate media content by traffic score as the first media content.

[0116] 2) The recall phase is used to identify a portion of media content that users may be interested in from a massive media content library based on the characteristics of users and media content.

[0117] For a specific user to be recommended, since the first media content is predetermined, it may not be fully recalled if the recall strategy in related technologies is used. Therefore, a separate recall strategy needs to be determined, namely the preset recall strategy, to ensure that the first media content is fully recalled. It is understood that in this embodiment, all media content recalled during the recall phase includes the union of the first media content and the second media content, where the second media content is the media content corresponding to the user to be recommended, recalled based on the recall strategy in related technologies.

[0118] In one optional implementation, recalling the first media content based on a preset recall strategy includes: recalling the first media content according to the identification information of the first media content.

[0119] In this embodiment, after determining the first media content, the identification information of the first media content can be stored separately, for example, in a Redis system. During the recall phase, the server can open a separate recall path, retrieve the identification information of the first media content from the Redis system, and recall the first media content based on the identification information.

[0120] 3) The screening stage is used to score and rank the recalled media content, and then select the media content to be recommended to the users to be recommended.

[0121] It should be noted that the step of filtering the first media content to obtain the target media content and the step of filtering other media content in the second media content besides the first media content can be performed separately.

[0122] In practice, the server can filter the first media content based on a preset filtering strategy to obtain the target media content and determine the initial recommendation parameters for the target media content. Then, the server can recommend the target media content to the client based on the initial recommendation parameters.

[0123] In this embodiment of the application, after the server recommends the target media content to the client, it can obtain feedback information from all users on the target media content, or it can filter some users and only obtain feedback information from these users on the target media content.

[0124] Optionally, before step 101, the method further includes:

[0125] Target users are determined based on the content tag information of the target media content, wherein the user tag information of the target users is matched with the content tag information of the target media content.

[0126] Step 101 includes:

[0127] Obtain feedback information from the target user regarding the target media content.

[0128] In this embodiment, the users who receive the feedback information in step 101 are limited.

[0129] When monitoring the target media content, the server only obtains the feedback information of the target user on the target media content, and determines whether it matches the preset feedback information based on the feedback information of the target user. It can be understood that the preset feedback information can also be determined based on the target user.

[0130] The target users refer to users who match the target media content. With the premise of increasing the exposure of the target media content, the recommendation parameters of the target media content are adjusted based on the actual feedback from users who match the target media content. This not only reduces data processing volume but also determines the actual user attraction ability of the target media content from the perspective of the potential interested audience, making the subsequently determined correction coefficients more accurate and further improving the accuracy of media content recommendations.

[0131] In practice, after determining the target media content, the server can obtain the content tag information of the target media content. Taking video content as an example, the content tag information may include type tag information, theme tag information, emotion tag information, actor tag information, screenwriter tag information, director tag information, production company tag information, etc., representing the target media content. After a new user registers a client, the server can begin collecting user tag information, which may include gender tag information, age tag information, user preference tag information for video content, etc.

[0132] The server can match users based on the content tag information of the target media content. If a user's user tag information matches at least one content tag information of the target media content, that user can be identified as a target user. Matching user tag information with content tag information can include: the user tag information and content tag information being the same or similar, or the user tag information indicating that the user is interested in the content corresponding to the content tag information, such as the user marking the content tag information as "my favorite," "my favorite," or "I'll look at it later."

[0133] In this embodiment, the server can continuously acquire user feedback information on the target media content and perform statistical analysis. To further improve the flexibility of target media content recommendation, make fuller use of recommendation resources, and avoid resource waste, the server can periodically collect user feedback information on the target media content and determine the correction coefficient of the target media content for the next statistical period based on the feedback information of the target media content in the current statistical period.

[0134] Optionally, step 102 includes:

[0135] Obtain the user's first feedback information on the target media content within the first time period;

[0136] If the first feedback information does not match the first preset feedback information, a first correction coefficient for the target media content is determined based on the first feedback information and the first preset feedback information, wherein the first preset feedback information is the preset feedback information corresponding to the first time period.

[0137] The step of correcting the recommendation parameters of the target media content according to the correction coefficient includes:

[0138] Based on the first correction coefficient, the recommendation parameters of the target media content in the second time period are corrected. The second time period is continuous with the first time period and is after the first time period.

[0139] The duration of the first time period and the second time period can be the same or different. In one example, both the first time period and the second time period can be 1 hour.

[0140] In practical applications, when a user clicks on the target media content, it indicates that the target media content has attracted the user's attention. Therefore, the number of clicks best reflects the user's response to the target media content, and thus best demonstrates the target media content's ability to attract users.

[0141] Optionally, in this embodiment of the application, the feedback information includes the number of clicks, and the preset feedback information includes the preset number of clicks. The server can determine whether the user's feedback information on the target media content matches the preset feedback information based on whether the user's actual number of clicks on the target media content matches the preset number of clicks, and thus determine whether the actual user traffic brought by the target media content meets expectations.

[0142] In this embodiment, the server can determine the preset click count based on the current display information of the target media content on the client. Specifically, the preset click count is related to the display position information of the target media content on the client. Taking the size of the display area as an example, the larger the display area, the easier it is for users to notice the target media content, and the larger the preset click count; the smaller the display area, the harder it is for users to notice the target media content, and the smaller the preset click count. Alternatively, taking the display order as an example, the earlier the display order, the earlier the user notices the target media content, and the larger the preset click count; the later the display order, the later the user notices the target media content, and the smaller the preset click count.

[0143] Further, optionally, a relationship diagram between display information and actual click volume can be drawn based on the correspondence between multiple sets of display information and actual click volume. For example, such as... Figure 2 As shown, the exposure and click-through rate of the target media content will decrease as its display ranking decreases. Subsequently, the server can obtain a functional relationship between the display information and the preset click-through rate through fitting.

[0144] The preset click volume is also related to the display time information of the target media content on the client. Users typically click on the target media content between 8 AM and 10 PM more often than between 10 PM and 8 AM the following day. Therefore, the preset click volume corresponds to higher peak time periods and lower off-peak time periods. Optionally, the server can determine the corresponding preset click volume for each time period.

[0145] In one example, the preset number of clicks is determined according to the following formula:

[0146] click j =exp(a·j b +c)

[0147] Among them, click j Let j be the preset click count, j be the display order of the target media content on the client, and the values ​​of a, b, and c be determined based on the display time of the target media content on the client. The server can determine a, b, and c according to the characteristics of each time period, which corresponds to determining a function for calculating the preset click count for each time period.

[0148] The server can compare the number of clicks with a preset number of clicks to determine whether the number of clicks matches the preset number of clicks. Optionally, the server can determine a preset click volume threshold based on the preset click volume, and determine that the number of clicks on the target media content does not match the preset click volume if the number of clicks by the user on the target media content is less than or greater than the preset click volume threshold. Alternatively, the server can determine a preset click volume range based on the preset click volume, and determine that the number of clicks by the user on the target media content does not match the preset click volume if the number of clicks by the user on the target media content is less than the lower limit of the preset click volume range or greater than the upper limit of the preset click volume range.

[0149] The preset click threshold can be equal to the preset click volume; alternatively, the server can preset a threshold λ and determine the preset click threshold based on the preset click volume and the threshold λ. For example, the preset click threshold is the sum of the preset click volume and the threshold λ, and the threshold λ can be positive or negative. For ease of understanding, an example is given: assuming the preset click volume is 12000, the threshold λ can be -2000, meaning that if the number of clicks on the target media content is less than 10000, the user traffic brought by the target media content is considered to have not met expectations; the threshold λ can be 2000, meaning that if the number of clicks on the target media content is less than 14000, the user traffic brought by the target media content is considered to have not met expectations.

[0150] In one optional implementation, if the number of clicks does not match the preset number of clicks, a correction coefficient for the target media content is determined based on the ratio of the number of clicks to the preset number of clicks.

[0151] In this embodiment, when the number of clicks is less than the preset number of clicks, the correction coefficient is less than 1, and the exposure provided by the recommended parameters after correction is less than the exposure provided by the original recommended parameters. Conversely, when the number of clicks is greater than the preset number of clicks, the correction coefficient is greater than 1, and the exposure provided by the recommended parameters after correction is greater than the exposure provided by the original recommended parameters. Thus, when the actual number of clicks on the target media content is higher than expected, adjusting the recommended parameters can increase the exposure of the target media content and attract more user traffic. Conversely, when the actual number of clicks on the target media content is lower than expected, adjusting the recommended parameters can reduce the exposure of the target media content, allowing the recommended parameters with higher exposure to be reserved for media content with higher actual clicks.

[0152] For ease of understanding, a complete example of an embodiment of this application is described herein:

[0153] like Figure 3 As shown, the specific implementation process of this example is as follows:

[0154] Step 301: Determine candidate media content and estimate traffic based on the characteristic information of the candidate media content, and select the top 20 media content with the highest traffic score as the primary media content.

[0155] In this step, the server can extract feature information such as quality score, popularity score, rating, actors, director, synopsis, poster information, and production company for each candidate media content, and predict the traffic of each candidate media content using a pre-trained traffic prediction model. The quality score and popularity score can be obtained based on preset parameters and models, while the rating can be obtained by acquiring relevant information from various forum websites or social networking sites.

[0156] Based on the traffic score output by the traffic prediction model, the server can select the top 20 candidate media content based on the traffic score as the primary media content.

[0157] Step 302: During the recall phase, the first media content is forcibly recalled.

[0158] In this step, the server can use a preset recall strategy to recall the first media content using a separate strategy during the recall phase, and return this first media content along with other recall results based on the regular recall strategy. Afterwards, the server can use a preset filtering strategy to determine the target media content.

[0159] Step 303: Filter target users through the tag information of the primary media content.

[0160] In this step, the server can match the content tag information of the first media content with the user tags of all users on the current website. If the user tag information of a certain user matches at least one content tag information of at least one first media content, then the user is identified as the target user.

[0161] Step 304: Based on the target user, calculate the score of the first media content according to the UCB algorithm in the exploration and utilization algorithm, and determine the target media content.

[0162] In this step, the score of the first media content i i It can be represented as:

[0163]

[0164] Among them, C i,t T represents the number of times the primary media content is clicked in t experiments. i,tThis represents the number of times the primary media content was recommended, or in other words, the number of times it was displayed, in t experiments.

[0165] The content is sorted according to its score, and target media content is filtered based on the sorting and recommended to the client.

[0166] Step 305: Monitor the target media content in real time.

[0167] In this step, the server monitors the number of clicks on the target media content in real time and checks whether the real-time number of clicks matches the preset number of clicks every hour. If they do not match, a prompt signal is issued and step 306 is executed; if they match, step 305 is executed repeatedly.

[0168] Specifically, the server can predict click volume hourly for different display positions. Data analysis shows that higher display positions result in higher exposure and click volume. Figure 2 As shown, the server uses a mathematical formula to fit the relationship between display order and preset click volume. The specific relationship is as follows:

[0169] click j =exp(a·j b +c)

[0170] Among them, click j Let j represent the preset click count, j be the display order of the target media content on the client, and the values ​​of a, b, and c be related to the display time period of the target media content on the client. Furthermore, the server can determine the values ​​of a, b, and c based on the characteristics of each hourly time period. Then, for different hourly time periods, different relationships can be determined based on different values ​​of a, b, and c, thereby determining different preset click counts.

[0171] The server can count the real-time clicks of the target media content every hour. i .

[0172] The server can display the real-time click count of the target media content i within the current hourly time period. i With preset click count j In comparison, if click i <click j If +λ is used, it indicates whether the real-time click volume matches the preset click volume. If the actual user traffic brought by the target media content i is lower than expected, a stop-loss signal is issued.

[0173] Step 306: Reduce the ranking and suppress the target media content that issued the warning signal.

[0174] In this step, the server can calculate the weighting coefficient γ. i :

[0175]

[0176] In the next hour, the score for the target media content is calculated using click. i Then multiply by γ i The corrected score is obtained. This allows for the re-determination of recommendation parameters for the target media content in the next hourly time slot based on the corrected score. Step 305 can then be repeated.

[0177] See Figure 4 , Figure 4 This is a structural diagram of the media content recommendation device provided in the embodiments of this application.

[0178] like Figure 4 As shown, the media content recommendation device 400 includes:

[0179] The first acquisition module 401 is used to acquire user feedback information on the target media content;

[0180] The first determining module 402 is used to determine the correction coefficient of the target media content based on the feedback information and the preset feedback information when the feedback information does not match the preset feedback information. The preset feedback information is associated with the display information of the target media content on the client, and the display information includes display location information and / or display time information.

[0181] The correction module 403 is used to correct the recommendation parameters of the target media content according to the correction coefficient.

[0182] Optionally, the media content recommendation device 400 also includes:

[0183] The second determining module is used to determine the target user based on the content tag information of the target media content, wherein the user tag information of the target user is matched with the content tag information of the target media content.

[0184] The first acquisition module 401 is used for:

[0185] Obtain feedback information from the target user regarding the target media content.

[0186] Optionally, the first determining module 402 includes:

[0187] The first acquisition unit is used to acquire the first feedback information of the user on the target media content within a first time period;

[0188] The first determining unit is configured to determine a first correction coefficient for the target media content based on the first feedback information and the first preset feedback information when the first feedback information does not match the first preset feedback information. The first preset feedback information is the preset feedback information corresponding to the first time period.

[0189] The correction module 403 is used for:

[0190] Based on the first correction coefficient, the recommendation parameters of the target media content in the second time period are corrected. The second time period is continuous with the first time period and is after the first time period.

[0191] Optionally, the feedback information includes the number of clicks, and the preset feedback information includes a preset number of clicks;

[0192] The first determining module 402 is used for:

[0193] If the number of clicks does not match the preset number of clicks, a correction coefficient for the target media content is determined based on the ratio of the number of clicks to the preset number of clicks.

[0194] Optionally, the preset number of clicks is determined according to the following formula:

[0195] click j =exp(a·j b +c)

[0196] Among them, click j The preset click count is denoted as j, where j is the display order of the target media content on the client, and the values ​​of a, b, and c are determined based on the display time of the target media content on the client.

[0197] Optionally, the correction module 403 includes:

[0198] The second acquisition unit is used to acquire the first recommendation score of the target media content before correction;

[0199] The second determining unit is used to determine the second recommended score of the target media content after correction based on the product of the correction coefficient and the first recommended score.

[0200] Optionally, the media content recommendation device 400 also includes:

[0201] The third determining module is used to determine the content of the first media.

[0202] The recall module is used to recall the first media content based on a preset recall strategy;

[0203] The filtering module is used to determine the target media content in the first media content based on a preset filtering strategy.

[0204] Optionally, the third determining module includes:

[0205] The fourth determining unit is used to determine candidate media content;

[0206] The fifth determining unit is used to determine the estimated traffic information of each candidate media content based on the feature information of each candidate media content;

[0207] The sixth determining unit is used to determine the first media content among the candidate media content based on the estimated traffic information of each candidate media content.

[0208] Optionally, the recall module is used for:

[0209] Based on the identification information of the first media content, the first media content is recalled.

[0210] The media content recommendation device 400 can achieve, for example Figure 1 The various processes in the corresponding method embodiments, and the achievement of the same beneficial effects, will not be described again here to avoid repetition.

[0211] This application also provides an electronic device. Please refer to [link to relevant documentation]. Figure 5 The electronic device 500 may include a processor 501, a memory 502, and a computer program 5021 stored in the memory 502 and executable on the processor 501. When the computer program 5021 is executed by the processor 501, it can implement any of the steps in the above method embodiments and achieve the same beneficial effects, which will not be elaborated here.

[0212] Those skilled in the art will understand that all or part of the steps of the methods described in the above embodiments can be implemented by hardware related to program instructions, and the program can be stored in a readable medium. This application also provides a readable storage medium storing a computer program. When the computer program is executed by a processor, it can implement any step in the above method embodiments and achieve the same technical effect. To avoid repetition, further details are omitted here.

[0213] The storage medium may be a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0214] The above description represents the preferred embodiments of this application. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles described in this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A media content recommendation method, characterized by, The method comprises the following steps: obtaining feedback information of a target media content from a user; in a case where the feedback information does not match preset feedback information, determining a correction coefficient of the target media content according to the feedback information and the preset feedback information, the preset feedback information being associated with display information of the target media content on a client, the display information comprising display position information and / or display time information; correcting a recommendation parameter of the target media content according to the correction coefficient; the feedback information comprises a click volume, and the preset feedback information comprises a preset click volume; in a case where the click volume does not match the preset click volume, determining the correction coefficient of the target media content according to a ratio of the click volume to the preset click volume; the preset click volume is determined according to the following formula: before the step of obtaining the feedback information of the target media content from the user, the method further comprises the following steps: ； wherein, is the preset click volume, j is the display order of the target media content on the client, and the values of a, b, and c are determined according to the display time of the target media content on the client.

2. The method of claim 1, wherein, determining a target user based on content tag information of the target media content, the user tag information of the target user matching the content tag information of the target media content; the step of obtaining the feedback information of the target media content from the user comprises the following step: obtaining the feedback information of the target media content from the target user. in a case where the feedback information does not match preset feedback information, determining a correction coefficient of the target media content according to the feedback information and the preset feedback information, the preset feedback information being associated with display information of the target media content on a client, the display information comprising display position information and / or display time information; 3. The method of claim 1, wherein, obtaining first feedback information of the target media content from a user within a first time period; in a case where the first feedback information does not match first preset feedback information, determining a first correction coefficient of the target media content according to the first feedback information and the first preset feedback information, the first preset feedback information being preset feedback information corresponding to the first time period; the step of correcting the recommendation parameter of the target media content according to the correction coefficient comprises the following step: correcting a recommendation parameter of the target media content in a second time period according to the first correction coefficient, the second time period being continuous with the first time period and being after the first time period. the step of correcting the recommendation parameter of the target media content according to the correction coefficient comprises the following steps:

4. The method of claim 1, wherein, obtaining a first recommendation score of the target media content before correction; determining a second recommendation score of the target media content after correction according to a product of the correction coefficient and the first recommendation score. before the step of obtaining the feedback information of the target media content from the user, the method further comprises the following steps:

5. The method of claim 1, wherein, determining a first media content; recalling the first media content based on a preset recall strategy; determining the target media content in the first media content based on a preset screening strategy. the step of determining a first media content comprises the following steps:

6. The method of claim 5, wherein, determining a candidate media content; determining estimated traffic information of each candidate media content according to feature information of each candidate media content; ​ According to the estimated traffic information of each candidate media content, a first media content is determined from the candidate media contents.

7. The method according to claim 5 or 6, characterized in that, The first media content is recalled based on a preset recall strategy, which includes: According to the identification information of the first media content, the first media content is recalled.

8. A media content recommendation apparatus, characterized by, It includes: The first acquisition module is used for acquiring the feedback information of the target media content of the user; The first determination module is used for determining the correction coefficient of the target media content according to the feedback information and the preset feedback information in the case that the feedback information does not match the preset feedback information, and the preset feedback information is associated with the display information of the target media content on the client side, and the display information includes display position information and / or display time information; The correction module is used for correcting the recommendation parameter of the target media content according to the correction coefficient; The feedback information includes the click volume, and the preset feedback information includes the preset click volume; The first determination module is used for: In the case that the click volume does not match the preset click volume, the correction coefficient of the target media content is determined according to the ratio of the click volume and the preset click volume; The preset click volume is determined according to the following formula: ； wherein, is the preset click volume, j is the display order of the target media content on the client, and the values of a, b, and c are determined according to the display time of the target media content on the client.

9. The apparatus of claim 8, wherein, The device further includes: The second determination module is used for determining the target user based on the content label information of the target media content, and the user label information of the target user matches the content label information of the target media content; The first acquisition module is used for: Acquiring the feedback information of the target user to the target media content.

10. The apparatus of claim 8, wherein, The first determination module includes: The first acquisition unit is used for acquiring the first feedback information of the user to the target media content in the first time period; The first determination unit is used for determining the first correction coefficient of the target media content according to the first feedback information and the first preset feedback information in the case that the first feedback information does not match the first preset feedback information, and the first preset feedback information is the preset feedback information corresponding to the first time period. The correction module is used for: According to the first correction coefficient, the recommendation parameter of the target media content in the second time period is corrected, and the second time period is continuous with the first time period and after the first time period.

11. The apparatus of claim 8, wherein, The correction module includes: The second acquisition unit is used for acquiring the first recommendation score of the target media content before correction; The second determination unit is used for determining the second recommendation score of the target media content after correction according to the product of the correction coefficient and the first recommendation score.

12. The apparatus of claim 8, wherein, The device further includes: The third determination module is used for determining the first media content; The recall module is used for recalling the first media content based on a preset recall strategy; The screening module is used for determining the target media content from the first media content based on a preset screening strategy.

13. The apparatus of claim 12, wherein, The third determination module includes: The fourth determination unit is used for determining the candidate media content; The fifth determination unit is used for determining the estimated traffic information of each candidate media content according to the feature information of each candidate media content; A sixth determining unit is configured to determine a first media content from the candidate media contents according to the estimated traffic information of each candidate media content.

14. The apparatus of claim 12 or 13, wherein, The recall module is configured to: Recall the first media content according to the identification information of the first media content.

15. An electronic device, comprising: A computer program product comprising a processor, a memory, and a computer program stored on the memory and loadable on the processor, the computer program implementing the steps of the method according to any one of claims 1 to 7 when executed by the processor.

16. A computer-readable storage medium, characterized in that, A computer program product comprising a processor, a memory, and a computer program stored on the memory and loadable on the processor, the computer program implementing the steps of the method according to any one of claims 1 to 7 when executed by the processor.

Images