Comment recommendation method, medium, device and computing device

Abstract

The embodiments of the present disclosure provide a comment recommendation method, medium, device and computing device. The method includes: in response to a received request for recommending high-quality comments, determining a target user to whom high-quality comments need to be recommended; calculating the online activity value of the target user; calculating the interaction willingness value of the target user; calculating the interest value of the target user in the comment content of the second user based on the online activity value, the interaction willingness value and the influence value of the comment content of the second user calculated in advance on the target user, wherein the second user is a user other than the target user; based on the interest value, recommending the comment content of the second user to the target user. By determining the comments recommended to the target user based on the activity status of the target user itself and the influence of the high-quality comments of the second user to be recommended on the target user, personalized differentiation according to different users is achieved, bringing a better experience to the user.

Description

Technical Field

[0001] The embodiments of the present disclosure relate to the field of computer technology, and more specifically, relate to a review recommendation method, medium, apparatus, and computing device. Background Art

[0002] This section is intended to provide a background or context to the embodiments of the disclosure that are recited in the claims. No description herein is admitted to be prior art by inclusion in this section.

[0003] With the development of Internet technology, more and more music playback applications involve content recommendation functions, which push corresponding content according to the characteristics of user traffic and improve the user experience. Among them, high-quality comments obtained after screening based on user-generated comments have gradually become the focus of recommendation.

[0004] Currently, the existing algorithms for recommending high-quality comments are mainly based on the number of interactions of the recommended comments or the number of fans of the recommended users. However, this results in the recommended high-quality comments not being personalized for different users, leading to low user satisfaction with the recommended content. Summary of the invention

[0005] The present disclosure provides a comment recommendation method, medium, apparatus and computing device to solve the problem in the prior art that the comment recommendation function fails to realize personalized differentiation according to different users, resulting in low user satisfaction with the recommended content.

[0006] In a first aspect of the embodiments of the present disclosure, a comment recommendation method is provided, comprising:

[0007] In response to the received request for recommending high-quality reviews, determining a target user to whom high-quality reviews need to be recommended;

[0008] Calculate the online activity value of the target user;

[0009] Calculate the target user's willingness to interact;

[0010] Based on the online activity value, the interaction willingness value, and the pre-calculated influence value of the second user's comment content on the target user, the interest value of the target user in the second user's comment content is calculated, wherein the second user is a user other than the target user;

[0011] Based on the interest value, the comment content of the second user is recommended to the target user.

[0012] In one embodiment of the present disclosure, the comment recommendation method includes:

[0013] Calculate the online activity value of the target user, including:

[0014] The online activity value of the target user is determined based on the ratio of the number of time units in which the target user has interactive actions within a set time period to the total number of time units, or based on the ratio of the target user's online activity duration, online duration, or the length of time the target user uses paid content to the total duration within the set time period. Interactive actions include operations in which the user posts comments or interacts with existing comment content. The total number of time units is used to represent the number of time units of fixed length into which the set time period is divided.

[0015] In an exemplary embodiment of the present disclosure, the target user's interaction willingness value is calculated, including: determining the target user's interaction willingness value based on the ratio of the target user's interaction operations on existing comments to the total number of interaction operations on the comments within a set time period, the interaction operations on existing comments including at least one of liking, forwarding, collecting, commenting on the existing comments, or sending a message to the user who generated the existing comments.

[0016] In another embodiment of the present disclosure, the comment recommendation method includes:

[0017] Based on the online activity value, the interaction willingness value, and the pre-calculated influence value of the second user's comment content on the target user, the interest value of the target user in the second user's comment content is calculated, including:

[0018] Calculate the product of the online activity value and the interaction willingness value as the first product;

[0019] Calculate the product of the first product and the pre-calculated influence value of the comment content of the second user on the target user as the second product;

[0020] The quotient of the second product and the online activity value of the second user is calculated as the interest value of the target user in the comment content of the second user.

[0021] In an exemplary embodiment of the present disclosure, the influence value of the comment content of the second user on the target user is pre-calculated in the following manner: obtain all user historical data, and divide the historical data into training data and test data; calculate the initial value of the influence value between all users based on the training data; calculate the first online activity value and the first interaction willingness value of all users based on the test data; calculate the predicted interest value of other users in the comment content of the second user based on the first online activity value, the first interaction willingness value and the initial value of the influence value between all users, where the other users are users other than the second user; correct the initial value of the influence value between all users based on the error between the actual interest value and the predicted interest value of the comment content of the second user by other users; determine the influence value of the comment content of the second user on the target user based on the corrected influence value between all users.

[0022] In an exemplary embodiment of the present disclosure, an initial value of the influence value between all users is calculated based on the training data, including: dividing the training data into K time periods of length τ1 according to the time length; calculating the second online activity value of all users in each time period; calculating the second interaction willingness value of all users in each time period; based on the second online activity value and the second interaction willingness value, calculating the initial value of the influence value between all users according to the maximum likelihood estimation method.

[0023] In another embodiment of the present disclosure, the comment recommendation method includes:

[0024] Based on the interest value, before recommending the second user's comment content to the target user, the following steps are also included:

[0025] The interest value of the target user in the comment content of the second user is calculated based on the online activity value, the interaction willingness value of the target user, the pre-calculated influence value of the comment content of the second user on the target user, and at least one of the following coefficients:

[0026] The recommendation coefficient of the second user's comment content;

[0027] A first coincidence coefficient between the target user's preference label and the label of the second user's comment content;

[0028] A second coincidence coefficient between the hotspot tag of the activity and the tag of the comment content of the second user is set.

[0029] In a second aspect of the embodiments of the present disclosure, a computer-readable storage medium is provided, comprising:

[0030] The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the review recommendation method in the first aspect of the present disclosure.

[0031] In a third aspect of the embodiments of the present disclosure, a review recommendation device is provided, comprising:

[0032] A determination module, configured to determine a target user to whom high-quality reviews need to be recommended in response to a received request for recommending high-quality reviews;

[0033] A first calculation module, used to calculate the online activity value of the target user;

[0034] The second calculation module is used to calculate the interaction willingness value of the target user;

[0035] A processing module, configured to calculate the interest value of the target user in the comment content of the second user based on the online activity value, the interaction willingness value, and the pre-calculated influence value of the comment content of the second user on the target user, wherein the second user is a user other than the target user;

[0036] The recommendation module is used to recommend the second user's comment content to the target user based on the interest value.

[0037] In an exemplary embodiment of the present disclosure, the first calculation module is specifically used to determine the online activity value of the target user based on the ratio of the number of time units in which the target user has interactive actions within a set time period to the total number of time units, or based on the ratio of the target user's online activity duration, online duration, or the length of time the target user uses paid content within the set time period to the total duration, where interactive actions include operations in which the user posts comments or interacts with existing comment content, and the total number of time units is used to represent the number of time units of fixed length into which the set time period is divided.

[0038] In an exemplary embodiment of the present disclosure, the second calculation module is specifically used to determine the target user's willingness to interact based on the ratio of the target user's number of interactive operations on existing comments to the total number of interactive operations on the comments within a set time period, where the interactive operations on existing comments include at least one of liking, forwarding, collecting, commenting on the existing comments, or sending information to the user who generated the existing comments.

[0039] In an exemplary embodiment of the present disclosure, the processing module is specifically used to: calculate the product of the online activity value and the interaction willingness value as the first product; calculate the product of the first product and the pre-calculated influence value of the second user's comment content on the target user as the second product; calculate the quotient of the second product and the second user's online activity value as the target user's interest value in the second user's comment content.

[0040] In an exemplary embodiment of the present disclosure, the processing module is specifically used to: obtain historical data of all users, and divide the historical data into training data and test data; calculate the initial value of the influence value between all users based on the training data; calculate the first online activity value and the first interaction willingness value of all users based on the test data; calculate the predicted interest value of other users in the comment content of the second user based on the first online activity value, the first interaction willingness value and the initial value of the influence value between all users, where the other users are users other than the second user; correct the initial value of the influence value between all users based on the error between the actual interest value and the predicted interest value of the comment content of the second user by other users; and determine the influence value of the comment content of the second user on the target user based on the corrected influence value between all users.

[0041] In an exemplary embodiment of the present disclosure, the processing module is specifically used to: divide the training data into K time periods of length τ1 according to the time length; calculate the second online activity value of all users in each time period; calculate the second interaction willingness value of all users in each time period; based on the second online activity value and the second interaction willingness value, calculate the initial value of the influence value between all users according to the maximum likelihood estimation method.

[0042] In an exemplary embodiment of the present disclosure, the processing module is also used to calculate the target user's interest value in the second user's comment content based on the target user's online activity value, interaction willingness value, the pre-calculated influence value of the second user's comment content on the target user, and at least one of the following coefficients: the recommendation coefficient of the second user's comment content; the first overlap coefficient between the target user's preference tag and the tag of the second user's comment content; the second overlap coefficient between the set activity's hot spot tag and the tag of the second user's comment content.

[0043] In a fourth aspect of the embodiments of the present disclosure, there is provided a computing device, comprising: at least one processor;

[0044] and a memory communicatively coupled to the at least one processor;

[0045] The memory stores instructions that can be executed by at least one processor, and the instructions are executed by at least one processor to enable the computing device to perform the comment recommendation method in the first aspect of the present disclosure.

[0046] According to the comment recommendation method, medium, device and computing device of the embodiment of the present disclosure, the target user to whom the high-quality comments need to be recommended is determined based on the received request for recommending high-quality comments, and the online activity value and interaction willingness value of the target user are calculated. Then, based on the online activity value, interaction willingness value and the influence value of the comment content of the second user calculated in advance on the target user, the interest value of the target user in the comment content of the second user is calculated, and finally, based on the interest value, the comment content of the second user is recommended to the target user. In this way, the high-quality comments pushed to the target user can be determined conveniently and quickly, and the characteristics of the target user's own activity, interaction willingness and tendency are combined, rather than only considering the popularity of the content to be recommended itself, thereby improving the relevance of the comment content to be recommended to the target user, thereby improving the interest of the target user, and reducing the recommendation of comments that are not related to the target user, bringing a better experience to the user. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] The above and other objects, features and advantages of the exemplary embodiments of the present disclosure will become readily understood by reading the detailed description below with reference to the accompanying drawings. In the accompanying drawings, several embodiments of the present disclosure are shown in an exemplary and non-limiting manner, in which:

[0048] Figure 1 The following schematically shows an application scenario diagram according to an embodiment of the present disclosure;

[0049] Figure 2 The flowchart of the comment recommendation method according to an embodiment of the present disclosure is schematically shown;

[0050] Figure 3 The flowchart of the comment recommendation method according to another embodiment of the present disclosure is schematically shown;

[0051] Figure 4 The flowchart of the comment recommendation method according to another embodiment of the present disclosure is schematically shown;

[0052] Figure 5 The flowchart of the comment recommendation method according to another embodiment of the present disclosure is schematically shown;

[0053] Figure 6 The structure diagram of a computer-readable storage medium according to another embodiment of the present disclosure is schematically shown;

[0054] Figure 7 The structure diagram of a comment recommendation device according to another embodiment of the present disclosure is schematically shown;

[0055] Figure 8 The figure schematically shows a structural diagram of a computing device according to yet another embodiment of the present disclosure.

[0056] In the drawings, the same or corresponding reference numerals represent the same or corresponding parts. DETAILED DESCRIPTION

[0057] The principles and spirit of the present disclosure will be described below with reference to several exemplary embodiments. It should be understood that these embodiments are provided only to enable those skilled in the art to better understand and implement the present disclosure, and are not intended to limit the scope of the present disclosure in any way. On the contrary, these embodiments are provided to make the present disclosure more thorough and complete, and to fully convey the scope of the present disclosure to those skilled in the art.

[0058] Those skilled in the art will appreciate that the embodiments of the present disclosure may be implemented as a system, device, apparatus, method or computer program product. Therefore, the present disclosure may be specifically implemented in the following forms, namely: complete hardware, complete software (including firmware, resident software, microcode, etc.), or a combination of hardware and software.

[0059] According to an embodiment of the present disclosure, a review recommendation method, medium, apparatus and computing device are proposed.

[0060] In this article, it should be understood that the terms involved are only used for convenience of understanding and do not limit the meaning. In addition, the number of any elements in the drawings is for illustration rather than limitation, and any naming is only for distinction and does not have any limiting meaning; and the data involved in this disclosure may be data authorized by the user or fully authorized by all parties.

[0061] The following is a description of the terms used in this disclosure:

[0062] High-quality comments: Comments generated by users, based on the number of interactive operations such as collection, forwarding, and likes by other users, and selected from comments with high interactive operations. Such comments usually have high sharing value and can be recommended to other users.

[0063] Interactive operation: In this scheme, it is used to represent operations such as liking, forwarding, collecting, and commenting on existing comments. When a user performs an interactive operation on an existing comment, the user who generated the existing comment can receive corresponding positive and negative feedback, thereby encouraging the generating user to continue to generate high-quality comments.

[0064] Interactive operations: In this solution, in addition to the aforementioned interactive operations on existing comments, interactive operations also include generating new comment content. Therefore, the interactive operations related to users and comments are no longer just consuming existing comment content, but include the generation of comment content, that is, all operations related to users and comments.

[0065] The principle and spirit of the present disclosure are explained in detail below with reference to several representative embodiments of the present disclosure. SUMMARY OF THE INVENTION

[0067] The inventors have discovered that in current technology, when music playback applications need to recommend high-quality comments to target users, based on existing algorithms, they generally recommend comments from users with a large number of fans, or recommend comments with the highest number of interactive operations within a certain period of time. However, such recommendation methods usually do not take into account the preferences of target users, or the target users' inclinations towards comment content, and therefore easily lead to low user satisfaction with the recommended content.

[0068] In this scheme, the target user's interest value in the recommended comments is determined based on the target user's own online activity value, interaction willingness value and the influence value of the target user's second user comment content on the recommended comments, thereby ensuring that the comments recommended to the target user are content that the user is interested in, thereby effectively improving the user's satisfaction with the recommended content.

[0069] After introducing the basic principles of the present disclosure, various non-limiting embodiments of the present disclosure are described in detail below.

[0070] Application Scenario Overview

[0071] First reference Figure 1 As shown, when a music playing application needs to recommend comments to the target user's client 100, high-quality comments that can be recommended to the target user's client 100 are determined and output based on the target user's relevant interactive operation records and pre-screened recommended comments recorded by the server 110.

[0072] It should be noted that Figure 1 In the scenario shown, only one server and one client are used as an example for illustration, but the present disclosure is not limited to this, that is, the number of servers and clients can be arbitrary.

[0073] Exemplary Methods

[0074] Combine the following Figure 1 For application scenarios, refer to Figures 2 to 5 To describe the method for comment recommendation according to an exemplary embodiment of the present disclosure. It should be noted that the above application scenarios are only shown to facilitate understanding of the spirit and principle of the present disclosure, and the embodiments of the present disclosure are not limited in this regard. On the contrary, the embodiments of the present disclosure can be applied to any applicable scenario.

[0075] Figure 2 A flowchart of a review recommendation method provided by an embodiment of the present disclosure. Figure 2 As shown, the comment recommendation method provided in this embodiment includes the following steps:

[0076] Step S201: In response to a received request for recommending high-quality reviews, determine a target user to whom high-quality reviews need to be recommended.

[0077] Requests are generally preset within the app, based on set functions or set activities. When the target user triggers the corresponding preset conditions, recommended high-quality reviews can be sent to the user.

[0078] There are usually more than one high-quality reviews that need to be recommended. The number of high-quality reviews that need to be recommended is determined based on the settings of specific functions or activities.

[0079] Corresponding to different functions or activity settings, the types of high-quality reviews recommended to users may be different. For example, in response to holiday activities, the recommended high-quality reviews are high-quality reviews related to the holiday theme. In response to user preference content, the recommended high-quality reviews are not limited to specific themes, but are based on the user's interest value.

[0080] By recommending high-quality comments to target users and obtaining their interactive operations on the high-quality comments, such as likes, favorites, or forwarding, and then feeding back to the second user who generated the high-quality comments, this will bring the second user a sense of satisfaction or accomplishment, forming positive feedback, thereby encouraging the second user to continue generating high-quality comments and achieving continuous output of user content.

[0081] Step S202: Calculate the online activity value of the target user.

[0082] Specifically, the online activity value is used to indicate the daily activity of the target user in order to determine the tendency of recommending content to him / her. For example, if the user is relatively active during a specific period of time, the content theme recommended to him / her can be related to that period of time. If the user has not been active for a long time, the content recommended to him / her can be filtered based on the inactive time period. On the contrary, if the user is very active for a long period of time, the comment content recommended to him / her generally needs to ensure a strong timeliness.

[0083] Furthermore, the online activity value can be determined according to a specific set algorithm, by the user's online activity time ratio, online time ratio and other values ​​within a set time period.

[0084] Step S203: Calculate the target user's interaction willingness value.

[0085] Specifically, the interaction willingness value is used to indicate the target user's willingness to interact with the comment.

[0086] The higher the interaction willingness value, the higher the probability and frequency of target users performing interactive operations. At this time, the range of high-quality reviews that can be recommended to target users will be wider, and more feedback from target users can be obtained. On the contrary, the lower the interaction willingness value, the lower the probability and frequency of target users performing interactive operations. At this time, the high-quality reviews recommended to target users will also receive lower feedback from target users. Therefore, the range of high-quality reviews that can be recommended will be lower than that of target users with high interaction willingness values, because some high-quality reviews are more difficult to arouse the target users' willingness to perform interactive operations, and the probability of receiving unsatisfactory evaluations will also be higher.

[0087] Furthermore, the interaction willingness value can be calculated based on the interaction operation frequency of the target user in a set time period.

[0088] Step S204: Calculate the interest value of the target user in the comment content of the second user based on the online activity value, the interaction willingness value, and the pre-calculated influence value of the comment content of the second user on the target user.

[0089] The second user is a user other than the target user.

[0090] The influence value is a value determined based on the target user's previous interactive operation data. For a specific second user and target user, their influence value is generally fixed within a set time period, because each person's preference for a specific content creator usually does not change frequently.

[0091] Therefore, based on the interactive operation data within a certain period of time, the influence values ​​of the comment contents of multiple second users related to the target user on the target user can be calculated.

[0092] Furthermore, in addition to considering the influence value of the second user's comment content on the target user, the interest value also needs to consider the target user's online activity value and interaction willingness value. When these two parameters are lower, even if the influence value is fixed, the target user's interest value in the second user's comment content is generally lower, and the range of recommended comments is smaller.

[0093] Furthermore, the interest value may also take into account the online activity value of the second user. For example, if one second user generates 2 high-quality comments every day and another second user generates 2 high-quality comments every month, the target user is more likely to be interested in the latter's high-quality comments because the former's large output will dilute the target user's curiosity and willingness to interact.

[0094] Step S205: Recommend the second user's comment content to the target user based on the interest value.

[0095] Specifically, the comment content of the second user is pre-screened high-quality content that can be recommended.

[0096] Furthermore, the comment content of the second user used for recommendation is selected from a preset high-quality comment content library, and finally the comment content recommended to the target user is obtained.

[0097] In some embodiments, the review content finally recommended to the target user may be a set number of review content sorted by interest value, or may be review content that satisfies a pre-set interest value benchmark value.

[0098] Furthermore, each second user's comment content can obtain a corresponding interest value according to the different second users it corresponds to and the different target users to be recommended. If the target users corresponding to the same comment content are different, the interest value will be different.

[0099] Therefore, based on each target user's online activity value, interaction willingness value and the influence value of the corresponding second user's comment content, the high-quality comments corresponding to the specific target user's interest values ​​from high to low can be determined, and then the comments with high interest values ​​can be recommended to the target user first, thereby effectively ensuring user satisfaction.

[0100] The comment recommendation method provided in this embodiment determines the target user to whom the high-quality comments need to be recommended based on the received request for recommending high-quality comments, calculates the online activity value and interaction willingness value of the target user, and then calculates the interest value of the target user in the comment content of the second user based on the online activity value, interaction willingness value and the influence value of the comment content of the second user calculated in advance on the target user, and finally recommends the comment content of the second user to the target user based on the interest value. In this way, the high-quality comments pushed to the target user can be determined conveniently and quickly, and the characteristics of the target user's own activity, interaction willingness and tendency are combined, rather than only considering the popularity of the content to be recommended itself, thereby improving the relevance of the comment content to be recommended to the target user, thereby improving the interest of the target user, and reducing the recommendation of comments that are not related to the target user, bringing a better experience to the user.

[0101] Figure 3 A flowchart of a review recommendation method provided by an embodiment of the present disclosure. Figure 3 As shown, the comment recommendation method provided in this embodiment includes the following steps:

[0102] Step S301: In response to a received request for recommending high-quality reviews, determine a target user to whom high-quality reviews need to be recommended.

[0103] In some embodiments, the target user is a user who has used a music playing application for a time period that reaches or exceeds a set value, such as more than 5 hours. This makes it convenient to recommend corresponding high-quality reviews based on the target user's historical data (such as usage time, usage habits, etc.).

[0104] Step S302: Determine the online activity value of the target user based on the ratio of the number of time units in which the target user has interactive actions within a set time period to the total number of time units.

[0105] The interactive actions include operations of users posting comments or interacting with existing comment content, and the total number of time units is used to represent the number of time units of fixed length into which the set time period is divided.

[0106] Specifically, the online activity value is determined based on the user's activity level in a set time period. The more active the user is, the higher the online activity value is. The user's activity level can be determined by the number of interactive actions per unit time.

[0107] Furthermore, for the online activity value α of the target user i within the set time period T i , divide the time period T into S hours of time units t j (j=1,2,3……S), the length of each time unit is τ0, satisfying:

[0108]

[0109] That is, the length of the time unit is the absolute value of the ratio of the length of the set time period to the number of time units.

[0110] Then, the maximum likelihood estimation algorithm is used to calculate the activity level of the target user i in the time period T. The calculation method is:

[0111]

[0112] Among them, y k is a parameter indicating whether the target user is at t k A binary variable indicating any interactive action performed within a time unit. When the target user has performed an interactive action, y k =1, otherwise, y k =0; n{y k =1} represents the number of time units in which the user has performed interactive actions.

[0113] Therefore, the online activity value of the target user in the set time period can be calculated by the ratio of the number of time units in which the target user has performed interactive actions to the total number of time units in the set time period.

[0114] For example, in the past month, the target user has 12 interactive time units of 24 hours, and the online activity value is: 12 / 30=0.4.

[0115] Step S303: determining the online activity value of the target user based on the online activity duration of the target user within a set time period, the online duration, or the ratio of the duration of the target user's use of paid content to the total duration.

[0116] This step is an optional step parallel to step S302. In the subsequent description of this embodiment, the corresponding online activity value is calculated based on the calculation method of step S302.

[0117] In some embodiments, the general server may preferentially select step S302 for execution based on the data integrity of the memory record, and may select step S303 for execution when the data integrity of the memory record is poor.

[0118] In some embodiments, the online activity value can also be determined based on the time the user uses the music player. For example, if the user prefers to play various types of music rather than perform a large number of interactive actions, it can also be considered that the user is in an active state when playing music.

[0119] At this point, it can be determined by the length of time the target user spends online, the length of time online, or the length of time using paid content.

[0120] Furthermore, for the online activity value α of the target user i within the set time period T i , the online activity duration of target user i is t p , online time is t l , the duration of using paid content is t v , t p Indicates the duration of target users using the music playback function, t l Indicates the duration of the target user being online, t v Indicates the time after the target user starts using paid content, then:

[0121] or or

[0122] According to the pre-set calculation rules, you can arbitrarily choose a calculation method to determine the online activity value of the target user.

[0123] For example, if a target user logged in online for 180 hours in the past month, then the online activity value can be calculated as: 180 / (24*30)=0.25.

[0124] Step S304: Determine the target user's interaction willingness value based on the ratio of the target user's interaction operation times on existing comments to the total interaction operation times on comments within a set time period.

[0125] The interactive operation on the existing comments includes at least one of liking, forwarding, collecting, commenting on the existing comments, or sending information to the user who generated the existing comments.

[0126] Specifically, the difference between interactive operations and interactive actions is that interactive operations are limited to interactions with existing comments, while interactive actions also include generating new comment content.

[0127] For the same target user, the higher the frequency or number of interactive operations performed by the user, the higher the user's willingness to interact can be considered.

[0128] Furthermore, for the interaction willingness value ω of the target user i within the set time period T i , the interaction willingness value of target user i in time period T is calculated by the maximum likelihood estimation algorithm. The calculation method is:

[0129]

[0130] Among them, RT i represents the number of interactions performed by target user i on existing comments within time period T, TW iRepresents the number of interactive actions of target user i in time period T, including generating new comment content.

[0131] Therefore, the interaction willingness value of the target user i is determined according to the proportion of the target user's willingness to generate new comment content.

[0132] For example, in the past month, the target user has performed a total of 15 interactive operations including collecting, forwarding, and liking, and a total of 22 interactive actions including generating new comments, collecting, forwarding, and liking. Then his interaction willingness value is: 15 / 22=0.68.

[0133] Step S305: Calculate the interest value of the target user in the comment content of the second user based on the online activity value, the interaction willingness value, and the pre-calculated influence value of the comment content of the second user on the target user.

[0134] Specifically, since the online activity value and interaction willingness value of the target user at different times may be significantly different, such as the online activity value this month is 0.5 and the online activity value last month is 0.2, the interest value calculated may have completely different results depending on the data selected for calculating the online activity value and interaction willingness value.

[0135] Therefore, when comment recommendations are needed, the online activity value and interaction willingness value of the target user are generally calculated in real time to ensure that they match the actual activity status of the target user.

[0136] In some embodiments, the server may also periodically calculate the online activity values ​​and interaction willingness values ​​of all target users, such as calculating and updating them once a day, to ensure that the compliance requirements of the comment recommendations of the target users are met.

[0137] Furthermore, the influence value of the second user's comment content on the target user is generally a pre-calculated fixed value, because the influence of the same second user on the same target user is generally fixed. Therefore, it is only necessary to combine the pre-calculated influence value with the online activity value and the interaction willingness value to obtain the corresponding interest value.

[0138] Step S306: Based on the interest value, recommend the second user's comment content to the target user.

[0139] Specifically, since the online activity value and interaction willingness value of the target user are different in different time periods, the interest value calculated for the same comment content may also change. Therefore, when it is necessary to push the comment content of the second user to the target user, it is necessary to determine the target user's interest in the comment content of the second user based on the real-time update or the latest interest value to ensure the target user's satisfaction with the recommended comment content.

[0140] The comment recommendation method provided in this embodiment, after receiving a request to recommend high-quality comments, determines the online activity value and interaction willingness value of the target user in real time or periodically, and combines them with the pre-calculated influence value of the comment content of the second user on the target user, to jointly obtain the real-time or latest target user's interest value in the comment content of the second user, so as to ensure that the interest value can match the actual state of the target user when obtaining the recommended comments, avoid the problem that the recommended comment content cannot meet the user's needs due to changes in the user's online activity value and interaction willingness value, thereby bringing a better experience to the user.

[0141] Figure 4 A flowchart of a review recommendation method provided by an embodiment of the present disclosure. Figure 4 As shown, the comment recommendation method provided in this embodiment includes the following steps:

[0142] Step S401: In response to a received request for recommending high-quality reviews, determine a target user to whom high-quality reviews need to be recommended.

[0143] This step and Figure 2 The content of step S201 in the corresponding embodiment is the same and will not be repeated here.

[0144] Step S402: Calculate the online activity value of the target user.

[0145] The online activity value can generally be obtained through real-time calculation.

[0146] Step S403: Calculate the target user's interaction willingness value.

[0147] The interaction willingness value can generally also be obtained through real-time calculation.

[0148] Step S404: Calculate the product of the online activity value and the interaction willingness value as the first product.

[0149] In some embodiments, if the online activity value and the interaction willingness value are calculated in real time, the corresponding first product will be calculated at the same time; if the online activity value and the interaction willingness value are calculated regularly, such as once a day or once every 3 days, the corresponding first product will be calculated at the same time and the first product will be saved in the server for subsequent calculation and call.

[0150] Step S405: Obtain all user historical data, and divide the historical data into training data and test data.

[0151] Specifically, all users are all users with data in the server, and generally do not include newly registered or unregistered users, because such users do not have corresponding operation data stored in the server.

[0152] Historical data includes all login information, usage time, and interactive action data for a set period of time in the past, such as one year or other time periods. Through historical data, each user's online activity value and interaction willingness value can be calculated.

[0153] Furthermore, the ratio of training data to test data can be set to any value, such as 1:1 or 7:3, by specifically dividing the historical data.

[0154] In some embodiments, data that is farther away from the present time can be used as training data, and data that is more recent can be used as test data. For example, if the length of historical data is two months, the data of the previous month can be used as training data, and the data of the next month can be used as test data.

[0155] In some embodiments, data may be selected at equal intervals as training data and test data, respectively. For example, if the length of historical data is 100 days, the data of each odd-numbered day may be used as training data, and the data of each even-numbered day may be used as test data.

[0156] Step S406: Calculate the initial value of the influence value among all users based on the training data.

[0157] First, the initial influence value needs to be obtained through training data, so that through continuous training and test adjustments, the interest value of the target user in the second user's comment content can be finally adjusted.

[0158] Furthermore, the specific method for calculating the initial value of the influence value includes:

[0159] Step 1: Divide the training data into K time periods of length τ1 according to the time length.

[0160] The value of K or τ1 is generally a preset fixed value, such as setting the value of τ1 to 24 hours or setting the value of K to 100 in advance, so as to quickly and uniformly perform calculations.

[0161] Step 2: Calculate the second online activity value of all users in each time period.

[0162] The second online activity value is based on different data from the online activity value in step S402 and is the online activity value of each time period in the historical data.

[0163] Step 3: Calculate the second interaction intention value of all users in each time period.

[0164] The second interaction willingness value is different from the data based on the interaction willingness value in step S403, and is the interaction willingness value of each time period in the historical data.

[0165] Step 4: Based on the second online activity value and the second interaction willingness value, the initial value of the influence value among all users is calculated using the maximum likelihood estimation method.

[0166] Specifically, for a specific target user i, when he is a fan of the second user j, the probability P that i will interact with the high-quality comment m generated by j (such as forwarding or sharing) can be expressed as:

[0167]

[0168] In the formula, f ij It is used to represent the influence value of the comment content of the second user j on the target user i, and P is also the interest value of the target user i in the comment content of the second user j.

[0169] Based on this, we can get the likelihood function of the target user i interacting with the second user j’s comments:

[0170]

[0171] In the formula, F ij =1 is used to indicate that the target user i follows the second user j, M j It is the collection of all information published by the second user j, including high-quality comments and non-high-quality comments.

[0172] Combined with the time periods into which the training data is divided, after processing the above formula, we get:

[0173] log(L)=LL

[0174]

[0175] in, It is a binary variable indicating whether the target user i interacts with the high-quality comment m generated by the second user j during the T period. Indicates that the target user i has performed an interactive operation. Indicates that target user i has not performed any interactive operation.

[0176] At this point, the influence value f can be calculated by combining the above formula with the optimization algorithm ij The initial value of is as follows:

[0177]

[0178] in, and That is, the second online activity value and the second interaction willingness value of target user i, is the second online activity value of the second user j. When the second user is less active, he / she publishes less information, and the target user is usually more willing to forward the high-quality content he / she produces.

[0179] Since the relative computational complexity of this algorithm is greater than that of the online activity value and the interaction willingness value, and the influence of the comment content of the same second user on the same target user is usually stable within a certain period of time, the influence value f between each second user and the target user is pre-calculated through the training data. ij In order to further analyze and obtain the corresponding influence value.

[0180] Step S407: Calculate the first online activity value and the first interaction willingness value of all users according to the test data.

[0181] Among them, the first online activity value and the first interaction willingness value are both results obtained based on test data, and are different from the aforementioned online activity value, interaction willingness value, second online activity value, and second interaction willingness value.

[0182] Specifically, when using the test data, it is also necessary to divide the test data into L time periods of length τ1 according to the time length, and then calculate the first online activity value and the first interaction willingness value corresponding to each time period respectively.

[0183] Step S408: Calculate the predicted interest value of other users in the comment content of the second user according to the first online activity value, the first interaction willingness value, and the initial value of the influence value among all users.

[0184] Among them, other users are users other than the second user.

[0185] Substituting the initial values ​​of the first online activity value, the first interaction willingness value, and the influence value among all users into the formula for calculating the interest value P in step S406, the predicted interest value P of other users' comments on the second user can be obtained. ij , specifically expressed as:

[0186]

[0187] In the formula, is the predicted interest value of the kth period, T k is the kth time period in the test data, satisfying k=1,……,K+L; and That is, the first online activity value and the first interaction willingness value of target user i, is the first online activity value of the second user j, is the user j in each time period T k The number of published information.

[0188] Step S409: correct the initial value of the influence value among all users according to the error between the actual interest value and the predicted interest value of other users' comments on the second user.

[0189] Specifically, the actual interest value can be determined by the ratio of the actual number of interactions between the target user and the second user's comment content in the historical data to the number of the second user's comment content, which is specifically expressed as:

[0190]

[0191] Among them, P T Indicates the actual interest value, Indicates that the target user has interacted with a comment of the second user. For the same comment, actions such as "like", "repost", and "collection" may be performed at the same time. No matter which one or more actions are performed, it can be considered that an interaction action has been performed with the comment content; A j =1 indicates that the second user has generated a comment.

[0192] Furthermore, the error between the actual interest value and the predicted interest value can be calculated using the mean square error, which is expressed as:

[0193]

[0194] Wherein, MSE represents mean square error, and N represents the number of user pairs that have established a follow relationship (such as target user i follows second user j).

[0195] In some embodiments, when the error is less than a set value, the calculated influence value f ij It can be used as the final influence value of the second user's comment content on the target user.

[0196] In some embodiments, when the number of training times or the influence value f ij When the number of corrections reaches the set number, such as 200 times, the final influence value can also be used as the final influence value of the second user's comment content on the target user.

[0197] Furthermore, the specific correction method is based on the following formula:

[0198]

[0199] In the formula, f ij ′ is used to represent the corrected influence value f ij .

[0200] Step S410: Determine the influence value of the second user's comment content on the target user based on the corrected influence values ​​among all users.

[0201] Specifically, the influence value of the second user's comment content on the target user can be obtained from the influence values ​​of all final second user's comment content on the target user calculated by the server.

[0202] Step S411: Calculate the product of the first product and the influence value of the second user's comment content on the target user as the second product.

[0203] Step S412: Calculate the quotient of the second product and the online activity value of the second user as the interest value of the target user in the comment content of the second user.

[0204] Based on the second product and the online activity value of the second user, combined with the formula for calculating the interest value P in step S406, the interest value of the target user in the comment content of the second user can be obtained.

[0205] Step S413: Recommend the second user's comment content to the target user based on the interest value.

[0206] Based on the finally calculated interest value of the target user in the comment content of the second user, the comment content recommended to the target user may be determined.

[0207] The comment recommendation method provided in this embodiment, after receiving a request to recommend high-quality comments, calculates the online activity value and the interaction willingness value, and combines them with the influence value of the second user's comment content on the target user obtained based on historical data training and testing, to jointly obtain the target user's interest value in the second user's comment content, so as to ensure that the interest value is determined based on the influence value of the second user's comment content on the target user, thereby increasing the probability of the target user interacting with the recommended comment content and improving the target user's satisfaction, thereby bringing a better experience to the user.

[0208] Figure 5 A flowchart of a review recommendation method provided by an embodiment of the present disclosure. Figure 5 As shown, the comment recommendation method provided in this embodiment includes the following steps:

[0209] Step S501: In response to a received request for recommending high-quality reviews, determine a target user to whom high-quality reviews need to be recommended.

[0210] In some embodiments, in addition to determining the target user, it is also necessary to determine the target user's preference tags, such as favorite music types, consumption habits, and birthday, etc., so as to push high-quality comments on related content in a targeted manner.

[0211] For example, if the target user's preference tag includes a preference for classical music, the proportion of comments related to the classical music genre will be increased in the recommended high-quality comments.

[0212] Step S502: Calculate the online activity value of the target user.

[0213] Step S503: Calculate the target user's interaction willingness value.

[0214] The online activity value and the interaction willingness value are calculated in the same way as in the aforementioned embodiment.

[0215] Step S504: combining the online activity value, the interaction willingness value, and the pre-calculated influence value of the second user's comment content on the target user as a first calculated value.

[0216] The first calculated value can be referenced Figure 4 In the corresponding embodiment, the method for calculating the interest value is as follows: Figure 4 The first calculated value obtained by the calculation method in the corresponding embodiment is only used as an intermediate value, and is not ultimately used to determine the interest value of the high-quality review to be recommended.

[0217] Step S505: multiply the first calculated value by the recommendation coefficient of the second user's comment content to calculate the interest value of the target user in the second user's comment content.

[0218] Specifically, the recommendation coefficient is a parameter preset by the system or administrator. For example, if a certain comment content is also widely circulated in environments other than music playback applications, this cannot be reflected only through the screening mechanism for high-quality comments in the system. At this time, a recommendation coefficient can be assigned to it separately, such as recommendation coefficient = 2.0, which means that compared with other comment content, the target user is more likely to have an interest in interactive operations for this comment. Therefore, when the interest value of the second user's comment content is increased, the probability of pushing the comment content to the target user is increased.

[0219] Step S506: multiply the first calculated value by the first coincidence coefficient between the target user's preference label and the label of the second user's comment content, and calculate the interest value of the target user in the second user's comment content.

[0220] In some embodiments, the comment content also includes labels, such as corresponding music works, keywords and related activities. The overlap coefficient is calculated based on the overlap between the labels of the comment content and the preference labels of the target users. The overlap coefficient is multiplied by the first calculated value, so that the comment content that is more in line with the target user's preferences is recommended to the target user to increase the probability of interactive operations of the target user.

[0221] Furthermore, the calculation method of the first coincidence coefficient d1 can be expressed as:

[0222]

[0223] Among them, n1 represents the number of labels in the target user’s preference labels that are the same as the labels of the comment content, and n i Indicates the number of target user's preference tags, n j The number of tags representing the comment content.

[0224] Step S507: multiply the first calculated value by the second coincidence coefficient of the hotspot tag of the set activity and the tag of the comment content of the second user, and calculate the interest value of the target user in the comment content of the second user.

[0225] In some embodiments, when the operator needs to launch certain activities, such as holiday activities or specific celebrations, in order to guide users to pay attention to related activities, the proportion of comments related to the activities can be increased.

[0226] For example, when an operator launches a Chinese Valentine's Day event, the operator may increase the proportion of comments related to tags such as love, lovers, Chinese Valentine's Day, and youth that are pushed to target users.

[0227] Specifically, the method for calculating the second coincidence coefficient d2 between the hotspot tag of the setting activity and the tag of the comment content of the second user can be expressed as:

[0228]

[0229] Among them, n2 represents the number of tags in the hot tags of the set activity that are the same as the tags of the comment content, and n s Indicates the number of hotspot tags set for the activity. Hotspot tags are pre-configured information when the operator exits the activity. j The number of tags representing the comment content.

[0230] Step S508: Recommend the second user's comment content to the target user based on the interest value.

[0231] Based on the product of the parameters in steps S505 to S507 and the first calculated value, the interest value of the comment content actually pushed out to the target user can be obtained, and based on the interest value, the comment content finally recommended to the target user can be determined.

[0232] The comment recommendation method provided in this embodiment, after receiving a request to recommend high-quality comments, obtains a first calculated value by calculating the online activity value and the interaction willingness value, and combining the pre-calculated influence value of the comment content of the second user on the target user, and then multiplies the first calculated value by relevant parameters to obtain the target user's interest value in the comment content of the second user, so as to improve the comment content recommended to the target user based on the interest value to meet the target user's preferences and be relevant to the activities launched by the operator, thereby increasing the probability of the target user interacting with the recommended comment content while meeting the needs of the operator, thereby bringing a better experience to both users and operators.

[0233] Exemplary Media

[0234] After introducing the method of the exemplary embodiment of the present disclosure, next, refer to Figure 6 A storage medium according to an exemplary embodiment of the present disclosure is described.

[0235] refer to Figure 6 As shown, a program product 60 for implementing the above method according to an embodiment of the present disclosure is described, which can adopt a portable compact disk read-only memory (CD-ROM) and include program code, and can be run on a terminal device, such as a personal computer. However, the program product of the present disclosure is not limited thereto.

[0236] The program product may use any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0237] The readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, wherein the readable program code is carried. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. The readable signal medium may also be any readable medium other than a readable storage medium.

[0238] Program code for performing the disclosed operations of the present disclosure may be written in any combination of one or more programming languages, including object-oriented programming languages, such as Java, C++, etc., and conventional procedural programming languages, such as "C" or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, partially on a remote computing device, or entirely on a remote computing device or server. In the case of a remote computing device, the remote computing device may be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN).

[0239] Exemplary Devices

[0240] After introducing the medium of the exemplary embodiment of the present disclosure, next, reference is made to Figure 7 The comment recommendation device of the exemplary embodiment of the present disclosure is described, which is used to implement the method in any of the above method embodiments. Its implementation principle and technical effect are similar to those of the above method embodiments, and will not be repeated here.

[0241] The review content device 700 provided by the present disclosure includes:

[0242] A determination module 710 is used to determine a target user to whom high-quality reviews need to be recommended in response to a received request for recommending high-quality reviews;

[0243] A first calculation module 720, used to calculate the online activity value of the target user;

[0244] The second calculation module 730 is used to calculate the interaction willingness value of the target user;

[0245] The processing module 740 is used to calculate the interest value of the target user in the comment content of the second user based on the online activity value, the interaction willingness value and the pre-calculated influence value of the comment content of the second user on the target user, wherein the second user is a user other than the target user;

[0246] The recommendation module 750 is used to recommend the comment content of the second user to the target user based on the interest value.

[0247] In an exemplary embodiment of the present disclosure, the first calculation module 720 is specifically used to determine the online activity value of the target user based on the ratio of the number of time units in which the target user has interactive actions within a set time period to the total number of time units, or based on the ratio of the target user's online activity duration, online duration, or the length of time the target user uses paid content within the set time period to the total duration, where interactive actions include operations in which the user posts comments or interacts with existing comment content, and the total number of time units is used to represent the number of time units of fixed length into which the set time period is divided.

[0248] In an exemplary embodiment of the present disclosure, the second calculation module 730 is specifically used to determine the target user's willingness to interact based on the ratio of the number of interaction operations of the target user on existing comments within a set time period to the total number of interaction operations on the comments, where the interaction operations on existing comments include at least one of liking, forwarding, collecting, commenting on existing comments, or sending information to the user who generated the existing comments.

[0249] In an exemplary embodiment of the present disclosure, the processing module 740 is specifically used to: calculate the product of the online activity value and the interaction willingness value as the first product; calculate the product of the first product and the pre-calculated influence value of the second user's comment content on the target user as the second product; calculate the quotient of the second product and the online activity value of the second user as the interest value of the target user in the comment content of the second user.

[0250] In an exemplary embodiment of the present disclosure, the processing module 740 is specifically used to: obtain historical data of all users, and divide the historical data into training data and test data; calculate the initial value of the influence value between all users based on the training data; calculate the first online activity value and the first interaction willingness value of all users based on the test data; calculate the predicted interest value of other users in the comment content of the second user based on the first online activity value, the first interaction willingness value and the initial value of the influence value between all users, where the other users are users other than the second user; correct the initial value of the influence value between all users based on the error between the actual interest value and the predicted interest value of the comment content of the second user by other users; and determine the influence value of the comment content of the second user on the target user based on the corrected influence value between all users.

[0251] In an exemplary embodiment of the present disclosure, the processing module 740 is specifically used to: divide the training data into K time periods of length τ1 according to the time length; calculate the second online activity value of all users in each time period; calculate the second interaction willingness value of all users in each time period; based on the second online activity value and the second interaction willingness value, calculate the initial value of the influence value between all users according to the maximum likelihood estimation method.

[0252] In an exemplary embodiment of the present disclosure, the processing module 740 is also used to calculate the target user's interest value in the second user's comment content based on the target user's online activity value, interaction willingness value, the pre-calculated influence value of the second user's comment content on the target user, and at least one of the following coefficients: the recommendation coefficient of the second user's comment content; the first overlap coefficient between the target user's preference tag and the tag of the second user's comment content; the second overlap coefficient between the set activity's hotspot tag and the tag of the second user's comment content.

[0253] Exemplary Computing Devices

[0254] After introducing the method, medium and apparatus of the exemplary embodiments of the present disclosure, next, reference is made to Figure 8 A computing device according to an exemplary embodiment of the present disclosure is described.

[0255] Figure 8 The computing device 80 shown is merely an example and should not bring any limitation to the functionality and scope of use of the embodiments of the present disclosure.

[0256] like Figure 8 As shown, the computing device 80 is in the form of a general computing device. The components of the computing device 80 may include but are not limited to: at least one processing unit 801, at least one storage unit 802, and a bus 803 connecting different system components (including the processing unit 401 and the storage unit 802).

[0257] The bus 803 includes a data bus, a control bus, and an address bus.

[0258] The storage unit 802 may include a readable medium in the form of a volatile memory, such as a random access memory (RAM) 8021 and / or a cache memory 8022 , and may further include a readable medium in the form of a non-volatile memory, such as a read-only memory (ROM) 4023 .

[0259] The storage unit 802 may also include a program having a set (at least one) of program modules 8024, such program modules 8024 including but not limited to: an operating system, one or more application programs, other program modules and program data, each of which or some combination may include the implementation of a network environment.

[0260] The computing device 80 may also communicate with one or more external devices 804 (e.g., keyboards, pointing devices, etc.). Such communication may be performed via an input / output (I / O) interface 805. Furthermore, the computing device 80 may also communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) via a network adapter 806. Figure 8 As shown, network adapter 406 communicates with other modules of computing device 80 via bus 803. It should be understood that, although not shown in the figure, other hardware and / or software modules can be used in conjunction with computing device 80, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0261] It should be noted that although several units / modules or sub-units / modules of the review recommendation device are mentioned in the above detailed description, this division is merely exemplary and not mandatory. In fact, according to the embodiments of the present disclosure, the features and functions of two or more units / modules described above can be embodied in one unit / module. Conversely, the features and functions of one unit / module described above can be further divided into multiple units / modules for embodiment.

[0262] In addition, although the operations of the disclosed method are described in a specific order in the drawings, this does not require or imply that the operations must be performed in this specific order, or that all the operations shown must be performed to achieve the desired results. Additionally or alternatively, some steps may be omitted, multiple steps may be combined into one step, and / or one step may be decomposed into multiple steps.

[0263] Although the spirit and principle of the present disclosure have been described with reference to several specific embodiments, it should be understood that the present disclosure is not limited to the disclosed specific embodiments, and the division of various aspects does not mean that the features in these aspects cannot be combined to benefit, and such division is only for the convenience of expression. The present disclosure is intended to cover various modifications and equivalent arrangements included in the spirit and scope of the attached claims.

Claims

1. A comment recommendation method, comprising: In response to the received request for recommending high-quality reviews, determining a target user to whom high-quality reviews need to be recommended; Calculate the online activity value of the target user; Calculate the target user's willingness to interact; Calculating the product of the online activity value and the interaction willingness value as a first product; Calculate the product of the first product and the pre-calculated influence value of the comment content of the second user on the target user as the second product, wherein the second user is a user other than the target user; Calculating a quotient of the second product and the online activity value of the second user as the interest value of the target user in the comment content of the second user; Based on the interest value, recommend the comment content of the second user to the target user; The pre-calculated influence value of the second user's comment content on the target user is obtained in the following manner: Obtaining all user historical data, and dividing the historical data into training data and test data; Calculate the initial value of the influence value among all users according to the training data; Calculate the first online activity value and the first interaction willingness value of all users according to the test data; Calculate the predicted interest value of other users on the comment content of the second user according to the first online activity value, the first interaction willingness value, and the initial value of the influence value among all users, where the other users are users other than the second user; According to the error between the actual interest value of other users to the comment content of the second user and the predicted interest value, correcting the initial value of the influence value among all users; Determining the influence value of the second user's comment content on the target user according to the corrected influence values ​​among all users; The step of calculating the initial values ​​of the influence values ​​among all users based on the training data includes: The training data is divided into The length is time period; Calculate the second online activity value of all users in each time period; Calculate the second interaction intention values ​​of all users in each time period; Based on the second online activity value and the second interaction willingness value, an initial value of the influence value among all users is calculated using a likelihood function of the target user's interaction with the second user's comment content according to a maximum likelihood estimation method.

2. According to the comment recommendation method of claim 1, the step of calculating the online activity value of the target user comprises: The online activity value of the target user is determined based on the ratio of the number of time units in which the target user has interactive actions within a set time period to the total number of time units, or based on the ratio of the target user's online activity duration, online duration, or the length of time the target user uses paid content to the total duration within the set time period, the interactive actions include generating new comment content and interactive operations on existing comments, the interactive operations on existing comments include at least one of liking, forwarding, collecting, commenting on existing comments, or sending information to the user who generated the existing comments, and the total number of time units is used to represent the number of time units of fixed length into which the set time period is divided.

3. According to the comment recommendation method of claim 1, the step of calculating the target user's interaction willingness value comprises: The target user's interaction willingness value is determined based on the ratio of the number of interaction operations of the target user on existing comments within a set time period to the number of interaction actions within the set time period, wherein the interaction actions include generating new comment content and interaction operations on existing comments, and the interaction operations on existing comments include at least one of liking, forwarding, collecting, commenting on existing comments, or sending information to the user who generated the existing comments.

4. The comment recommendation method according to any one of claims 1 to 3, before recommending the comment content of the second user to the target user based on the interest value, further comprising: The interest value of the target user in the comment content of the second user is calculated based on the online activity value, the interaction willingness value of the target user, the pre-calculated influence value of the comment content of the second user on the target user, and at least one of the following coefficients: The recommendation coefficient of the comment content of the second user; a first coincidence coefficient between the target user's preference label and the label of the second user's comment content; A second coincidence coefficient between the hot tag of the activity and the tag of the comment content of the second user is set.

5. A computer-readable storage medium, comprising: The computer-readable storage medium stores computer-executable instructions, which are used to implement the comment recommendation method according to any one of claims 1 to 4 when executed by a processor.

6. A review recommendation device, comprising: A determination module, configured to determine a target user to whom high-quality reviews need to be recommended in response to a received request for recommending high-quality reviews; A first calculation module, used to calculate the online activity value of the target user; The second calculation module is used to calculate the interaction willingness value of the target user; A processing module, used for calculating the product of the online activity value and the interaction willingness value as a first product; Calculate the product of the first product and the pre-calculated influence value of the comment content of the second user on the target user as the second product; Calculating the quotient of the second product and the online activity value of the second user as the interest value of the target user in the comment content of the second user; wherein the second user is a user other than the target user; A recommendation module, configured to recommend the comment content of the second user to the target user based on the interest value; The processing module is specifically used for: Obtain all user historical data and divide the historical data into training data and test data; Based on the training data, the initial value of the influence value among all users is calculated; Based on the test data, the first online activity value and the first interaction willingness value of all users are calculated; Calculate the predicted interest value of other users in the comment content of the second user according to the first online activity value, the first interaction willingness value, and the initial value of the influence value among all users, where the other users are users other than the second user; Correcting the initial value of the influence value among all users according to the error between the actual interest value and the predicted interest value of other users to the comment content of the second user; Determine the influence value of the second user's comment content on the target user according to the corrected influence values ​​among all users; The processing module is specifically used for: The training data is divided into The length is time period; Calculate the second online activity value of all users in each time period; Calculate the second interaction intention values ​​of all users in each time period; Based on the second online activity value and the second interaction willingness value, an initial value of the influence value among all users is calculated using a likelihood function of the target user's interaction with the second user's comment content according to a maximum likelihood estimation method.

7. The comment recommendation device according to claim 6, wherein the first calculation module is specifically used for: The online activity value of the target user is determined based on the ratio of the number of time units in which the target user has interactive actions within a set time period to the total number of time units, or based on the ratio of the target user's online activity duration, online duration, or the length of time the target user uses paid content to the total duration within the set time period, wherein the interactive actions include generating new comment content and interactive operations on existing comments, wherein the interactive operations on existing comments include at least one of liking, forwarding, collecting, commenting on existing comments, or sending information to the user who generated the existing comments, and the total number of time units is used to represent the number of time units of fixed length into which the set time period is divided.

8. The comment recommendation device according to claim 6, wherein the second calculation module is specifically used for: The target user's interaction willingness value is determined based on the ratio of the number of interaction operations of the target user on existing comments within a set time period to the number of interaction actions within the set time period, wherein the interaction actions include generating new comment content and interaction operations on existing comments, and the interaction operations on existing comments include at least one of liking, forwarding, collecting, commenting on existing comments, or sending information to the user who generated the existing comments.

9. The comment recommendation device according to any one of claims 6 to 8, wherein the processing module is further used for: The interest value of the target user in the comment content of the second user is calculated based on the online activity value, the interaction willingness value of the target user, the pre-calculated influence value of the comment content of the second user on the target user, and at least one of the following coefficients: The recommendation coefficient of the second user's comment content; A first coincidence coefficient between the target user's preference label and the label of the second user's comment content; A second coincidence coefficient between the hotspot tag of the activity and the tag of the comment content of the second user is set.

10. A computing device comprising: at least one processor; and a memory communicatively coupled to the at least one processor; The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the computing device to perform the comment recommendation method as described in any one of claims 1 to 4.

Images