CN113869951A

CN113869951A - Recommendation content determining method, recommendation device, recommendation equipment and storage medium

Info

Publication number: CN113869951A
Application number: CN202111144821.9A
Authority: CN
Inventors: 王广帅
Original assignee: Beijing Dajia Internet Information Technology Co Ltd
Current assignee: Beijing Dajia Internet Information Technology Co Ltd
Priority date: 2021-09-28
Filing date: 2021-09-28
Publication date: 2021-12-31

Abstract

The method comprises the steps of obtaining a recommended content set which comprises a plurality of categories and meets the preset conditions according to the recommended content; determining a target category and first recommended content belonging to the target category in multiple industry categories according to evaluation parameter information corresponding to various categories determined by a user aiming at historical record information of historical recommended content of multiple categories; and acquiring second recommended content recalled by the target user, and determining the target recommended content based on the first recommended content and the second recommended content. Therefore, the target recommended content recommended to the target user is determined by combining the recommended content corresponding to the target category and meeting the preset condition, so that the exposure of the target recommended content can be increased under the condition of reducing the user experience loss as much as possible, and the recommendation effect and the user satisfaction are improved.

Description

Recommendation content determining method, recommendation device, recommendation equipment and storage medium

Technical Field

The present disclosure relates to the field of computer technologies, and in particular, to a method, an apparatus, a device, and a storage medium for determining recommended content.

Background

Recommendation systems generally recommend content of interest to a user based on the user's historical behavior data. However, always making recommendations based on the known interests of the user may give the user the perception that repeated content is always recommended, greatly reducing freshness and surprise.

The above problem can be categorized as an Exploration and Exploitation (EE) problem in a recommender system, which is also one of the two most common problems in computing advertising and recommender systems. However, the existing solutions for the problems of exploration and utilization still have the problems of poor exploration recommendation effect, low user satisfaction, and the like, and further improvement is needed.

Disclosure of Invention

The present disclosure provides a recommendation content determining method, a recommendation device, a recommendation apparatus, and a storage medium, so as to solve at least one of the problems of poor recommendation effect, low user satisfaction, and the like in the related art. The technical scheme of the disclosure is as follows:

according to a first aspect of the embodiments of the present disclosure, there is provided a recommended content determining method, including:

acquiring a recommended content set, wherein the recommended content set comprises recommended contents which are of various categories and meet preset conditions;

determining evaluation parameter information corresponding to various categories according to the historical record information of the historical recommendation contents of the various categories of the user; the historical record information represents content information of various categories and corresponding operation information;

determining a target category in the multiple categories according to the evaluation parameter information corresponding to the various categories;

determining a first recommended content belonging to the target category from the set of recommended content;

and acquiring second recommended content recalled by the target user, and determining the target recommended content based on the first recommended content and the second recommended content.

As an optional implementation, the obtaining the recommended content set includes:

acquiring feedback information of recommended contents corresponding to various categories in a historical preset time period;

analyzing the feedback information to obtain content data of each piece of recommended content in at least three dimensions;

determining recommended content meeting a preset condition according to the content data and a preset threshold;

and establishing the recommended content set according to the determined recommended content.

As an optional implementation manner, the determining, according to the history information of the history recommended content for the multiple categories by the user, the evaluation parameter information corresponding to the various categories includes:

acquiring historical record information of the user aiming at the historical recommended contents of the various categories; the historical record information comprises content characteristic information and historical operation times corresponding to each category;

obtaining historical average evaluation parameter information corresponding to different categories based on the content characteristic information;

determining uncertain evaluation parameters corresponding to different categories based on the historical operation times corresponding to each category and the total historical operation times of each category;

and obtaining evaluation parameter information corresponding to various categories according to the historical average evaluation parameter information and the evaluation uncertain parameters.

As an optional implementation manner, the determining, according to the evaluation parameter information corresponding to the various categories, a target category in the various categories includes:

determining evaluation parameter information which meets preset evaluation conditions in the evaluation parameter information corresponding to each category as target evaluation parameter information;

and determining the category corresponding to the target evaluation parameter information as a target category.

As an optional implementation manner, the determining the target recommended content based on the first recommended content and the second recommended content includes:

determining candidate recommended content meeting a preset orientation condition based on the first recommended content and the second recommended content;

and processing the candidate recommended content based on a preset recommendation strategy to obtain target recommended content for recommending to the target user.

As an optional implementation manner, the determining, based on the first recommended content and the second recommended content, a candidate recommended content that meets a preset targeting condition includes:

correspondingly configuring the first recommended content and the second recommended content to a first channel and at least one second channel respectively, wherein each channel comprises a content tag used for representing the attribute of the recommended content;

merging the recommended contents in the first channel and the at least one second channel to obtain a merged recommended content set;

determining an expected recommended content set according to the user characteristic information corresponding to the target user;

according to the preset directional condition of the merged recommended content, performing intersection processing on the merged recommended content set and the expected recommended content set to obtain a candidate recommended content set;

and taking the candidate recommended content set as candidate recommended content meeting preset orientation conditions.

As an optional implementation manner, the preset recommendation policy includes a first recommendation policy and a second recommendation policy, and the content tag includes a first content tag representing first recommended content; the processing the candidate recommended content based on the preset recommendation strategy to obtain the target recommended content for recommending to the target user comprises:

processing the candidate recommended contents except the candidate recommended contents carrying the first content label in the candidate recommended contents based on the first recommendation strategy to obtain a first subset to be recommended;

processing the first subset to be recommended and the candidate recommended content carrying the first content tag based on the second recommendation strategy to obtain a second subset to be recommended;

under the condition that the second subset to be recommended is a non-empty set, taking the second subset to be recommended as target recommended content for recommending to the target user;

and under the condition that the second subset to be recommended is an empty set, taking the candidate recommended content carrying the first content tag as target recommended content for recommending to the target user.

As an optional implementation manner, in the case that the second subset to be recommended is a non-empty set, taking the second subset to be recommended as the target recommendation content for recommending to the target user includes:

under the condition that the second subset to be recommended is a non-empty set, performing diversity processing on the second subset to be recommended to obtain a processed second subset to be recommended;

taking the processed second subset to be recommended as target recommended content for recommending to the target user; alternatively, the first and second electrodes may be,

the taking the candidate recommended content carrying the first content tag as the target recommended content for recommending to the target user under the condition that the second subset to be recommended is an empty set includes:

under the condition that the second subset to be recommended is an empty set, performing diversity processing on the candidate recommended content carrying the first content tag to obtain a processed third subset to be recommended;

and taking the processed third subset to be recommended as target recommended content for recommending to the target user.

As an optional implementation manner, after the step of filtering and sorting the candidate recommended content based on the preset recommendation policy to obtain the target recommended content for recommendation to the target user, the method further includes:

obtaining posterior behavior data of the target user on the target recommended content;

analyzing the extended interest tag of the target user according to the posterior behavior data;

and filtering the target recommended content based on the extended interest tag to obtain updated target recommended content.

According to a second aspect of the embodiments of the present disclosure, there is provided a recommendation method including:

acquiring a target user and target recommended content recommended to the target user, wherein the target recommended content is obtained according to any recommended content determining method;

recommending the target recommended content to the target user.

As an optional implementation manner, the target user satisfies at least one of the following conditions: the user with the user activity degree greater than or equal to the preset activity degree threshold, the user with the application program use time length greater than or equal to the preset time length threshold, the user with the application program cold start times greater than or equal to the preset time threshold, and the user with the media content filling rate less than or equal to the preset filling rate threshold.

According to a third aspect of the embodiments of the present disclosure, there is provided a recommended content determining apparatus including:

the device comprises a first acquisition module, a second acquisition module and a third acquisition module, wherein the first acquisition module is configured to execute acquisition of a recommended content set, and the recommended content set comprises recommended contents of various categories and meets a preset condition;

the first determination module is configured to execute determination of evaluation parameter information corresponding to various categories according to historical record information of historical recommended contents of the various categories by a user; the historical record information represents content information of various categories and corresponding operation information;

a second determining module configured to determine a target category among the plurality of categories according to evaluation parameter information corresponding to the various categories;

a third determination module configured to perform determining a first recommended content belonging to the target category from the set of recommended contents;

and the target content determining module is configured to execute acquisition of second recommended content recalled for the target user and determine the target recommended content based on the first recommended content and the second recommended content.

As an optional implementation manner, the first obtaining module includes:

the first obtaining submodule is configured to obtain feedback information of recommended contents corresponding to various categories in a historical preset time period;

the second obtaining submodule is configured to analyze the feedback information and obtain content data of each piece of recommended content in at least three dimensions;

the determining submodule is configured to determine recommended content meeting a preset condition according to the content data and a preset threshold;

an establishing sub-module configured to establish the recommended content set according to the determined recommended content.

As an optional implementation, the first determining module includes:

a second acquisition module configured to perform acquisition of history information of the user with respect to the plurality of categories of history recommended content; the historical record information comprises content characteristic information and historical operation times corresponding to each category;

a first determination module configured to perform obtaining historical average evaluation parameter information corresponding to different categories based on the content feature information;

the second determining module is configured to execute the step of determining the uncertain evaluation parameters corresponding to different categories based on the historical operation times corresponding to each category and the total historical operation times of each category;

and the score determining module is configured to execute obtaining of evaluation parameter information corresponding to each category according to the historical average evaluation parameter information and the evaluation uncertain parameter.

As an optional implementation manner, the second determining module includes:

a target score determination submodule configured to perform determination of evaluation parameter information satisfying a preset evaluation condition among evaluation parameter information corresponding to each category as target evaluation parameter information;

and the category determining submodule is configured to determine the category corresponding to the target evaluation parameter information as a target category.

As an optional implementation manner, the third determining module includes:

a candidate determination submodule configured to determine candidate recommended content satisfying a preset orientation condition based on the first recommended content and the second recommended content;

and the target determination submodule is configured to process the candidate recommended content based on a preset recommendation strategy to obtain target recommended content for recommending to the target user.

As an optional implementation, the candidate content determination sub-module includes:

the channel configuration unit is configured to correspondingly configure the first recommended content and the second recommended content to a first channel and at least one second channel respectively, wherein each channel comprises a content tag used for representing the attribute of the recommended content;

a merging unit configured to perform merging of recommended contents in the first channel and the at least one second channel to obtain a merged recommended content set;

the expected content determining unit is configured to determine an expected recommended content set according to the user characteristic information corresponding to the target user;

the intersection unit is configured to perform intersection processing on the merged recommended content set and the expected recommended content set according to a preset orientation condition of the merged recommended content to obtain a candidate recommended content set;

a candidate content determination unit configured to perform the set of candidate recommended contents as candidate recommended contents satisfying a preset orientation condition.

As an optional implementation manner, the preset recommendation policy includes a first recommendation policy and a second recommendation policy, and the content tag includes a first content tag representing first recommended content; the target determination sub-module includes:

the first processing unit is configured to perform filtering and sorting on candidate recommended contents except for the candidate recommended contents carrying the first content tag based on the first recommendation strategy to obtain a first subset to be recommended;

the second processing unit is configured to perform filtering and sorting on the first subset to be recommended and the candidate recommended content carrying the first content tag based on the second recommendation strategy to obtain a second subset to be recommended;

a first target content determination unit configured to perform, in a case where the second subset to be recommended is a non-empty set, regarding the second subset to be recommended as target recommended content for recommendation to the target user;

and the second target content determining unit is configured to execute, in the case that the second subset to be recommended is an empty set, taking the candidate recommended content carrying the first content tag as a target recommended content for recommending to the target user.

As an optional implementation, the first target content determining unit includes:

the first processing subunit is configured to perform diversity processing on the second subset to be recommended under the condition that the second subset to be recommended is a non-empty set, and obtain a processed second subset to be recommended;

a first target content determination subunit configured to perform the processing of the second subset to be recommended as target recommended content for recommendation to the target user; alternatively, the first and second electrodes may be,

the second target content determination unit includes:

the second processing subunit is configured to perform diversity processing on the candidate recommended content carrying the first content tag under the condition that the second subset to be recommended is an empty set, so as to obtain a processed third subset to be recommended;

a second target content determining subunit, configured to execute the third processed subset to be recommended as the target recommended content for recommendation to the target user.

As an optional implementation, the apparatus further comprises:

the second acquisition module is configured to execute the acquisition of posterior behavior data of the target user on the target recommended content;

the analysis module is configured to analyze the extended interest tag of the target user according to the posterior behavior data;

and the filtering module is configured to filter the target recommended content based on the extended interest tag to obtain updated target recommended content.

According to a fourth aspect of the embodiments of the present disclosure, there is provided a recommendation apparatus including:

the acquisition module is configured to execute acquisition of a target user and target recommended content recommended to the target user, wherein the target recommended content is obtained by any one of the recommended content determination devices;

a recommending module configured to perform recommending the target recommended content to the target user.

According to a fifth aspect of the embodiments of the present disclosure, there is provided a computer-readable storage medium, wherein instructions of the computer-readable storage medium, when executed by a processor of an electronic device, enable the electronic device to perform the recommended content determining method or the recommending method according to any of the above embodiments.

According to a sixth aspect of embodiments of the present disclosure, there is provided an electronic apparatus including:

a processor;

a memory for storing the processor-executable instructions;

wherein the processor is configured to execute the instructions to implement the recommended content determining method or the recommending method according to any of the above embodiments.

According to a seventh aspect of embodiments of the present disclosure, there is provided a computer program product comprising a computer program that, when executed by a processor, implements the recommended content determining method or the recommending method provided in any of the above-mentioned embodiments.

The technical scheme provided by the embodiment of the disclosure at least brings the following beneficial effects:

the method comprises the steps of obtaining a recommended content set, wherein the recommended content set comprises recommended contents which are in various categories and meet preset conditions; determining evaluation parameter information corresponding to various categories according to the historical record information of the historical recommendation contents of the various categories of the user; the historical record information represents content information of various categories and corresponding operation information; determining a target category in the multiple categories according to the evaluation parameter information corresponding to the various categories; determining a first recommended content belonging to the target category from the set of recommended content; and acquiring second recommended content recalled by the target user, and determining the target recommended content based on the first recommended content and the second recommended content. Therefore, the target recommended content recommended to the target user is determined by combining the recommended content corresponding to the target category and meeting the preset conditions, the exposure of the target recommended content can be increased under the condition of not reducing the user experience loss, more potential interest tags of the target user are fully mined, the recommendation effect and the user satisfaction are improved, and the user viscosity is further improved.

It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this disclosure, illustrate embodiments consistent with the disclosure and together with the description, serve to explain the principles of the disclosure and are not to be construed as limiting the disclosure.

Fig. 1 is an architecture diagram illustrating a system to which a recommended content determining method or a recommending method is applied according to an exemplary embodiment.

Fig. 2 is a flowchart illustrating a recommended content determining method according to an exemplary embodiment.

Fig. 3 is a partial flow diagram illustrating another method of determining recommended content according to an example embodiment.

FIG. 4 is a flowchart illustrating a step of determining candidate recommended content that satisfies a preset targeting condition according to an exemplary embodiment.

FIG. 5 is a flowchart illustrating a step of obtaining target recommended content for recommendation to the target user in accordance with an exemplary embodiment.

Fig. 6 is a partial flowchart illustrating yet another recommended content determining method according to an exemplary embodiment.

FIG. 7 is a flow chart illustrating a recommendation method according to an example embodiment.

Fig. 8 is a block diagram illustrating a recommended content determining apparatus according to an exemplary embodiment.

FIG. 9 is a block diagram illustrating another recommendation device, according to an example embodiment.

FIG. 10 is a block diagram illustrating an electronic device in accordance with an example embodiment.

Detailed Description

In order to make the technical solutions of the present disclosure better understood by those of ordinary skill in the art, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below with reference to the accompanying drawings.

It should be noted that the terms "first," "second," and the like in the description and claims of the present disclosure and in the above-described drawings are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the disclosure described herein are capable of operation in sequences other than those illustrated or otherwise described herein. The implementations described in the exemplary embodiments below are not intended to represent all implementations consistent with the present disclosure. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present disclosure, as detailed in the appended claims.

It should be noted that, the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for presentation, analyzed data, etc.) referred to in the present disclosure are information and data authorized by the user or sufficiently authorized by each party.

Fig. 1 is an architecture diagram illustrating a system applying a recommended content determining method or a recommending method according to an exemplary embodiment, and referring to fig. 1, the architecture diagram may include a first terminal 10, a second terminal 20, and a server 30.

The first terminal 10 and the second terminal 20 may be, but are not limited to, one or more of an entity device such as a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart wearable device, a digital assistant, an augmented reality device, a virtual reality device, and an application program and an applet running in the entity device.

The server 30 may be a server corresponding to an application program on the first terminal 10 and the second terminal 20, and for example only, the server 30 may be, but is not limited to, an independent server, a server cluster or a distributed system formed by a plurality of physical servers, and one or more cloud servers providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDNs, and big data and artificial intelligence platforms.

For example only, the applications installed on the first terminal 10 and the second terminal 20 may be served by the server 30, the first terminal 10 may be a terminal corresponding to the target user, the second terminal 20 may be a terminal corresponding to the non-target user, and the server 30 determines the recommended content according to the first terminal 10 and the second terminal 20 and recommends the determined recommended content to the target user corresponding to the first terminal 10. The first terminal 10 and the server 30, and the second terminal 20 and the server 30 may be directly or indirectly connected through wired or wireless communication, and the embodiments of the present disclosure are not limited herein.

The recommended content determining method or the recommending method provided by the embodiment of the disclosure may be executed by a recommended content determining apparatus or a recommending apparatus, and the recommended content determining apparatus or the recommending apparatus may be integrated in an electronic device such as a client or a terminal in a hardware form or a software form.

Fig. 2 is a flowchart illustrating a recommended content determining method according to an exemplary embodiment, and as shown in fig. 2, the recommended content determining method may be applied to an electronic device, which is described by taking the electronic device as an example of a server in the above implementation environment schematic diagram, and includes the following steps.

In step S201, a recommended content set is obtained, where the recommended content set includes recommended contents of multiple categories and meets a preset condition.

Wherein the category may include, but is not limited to, a content category, an industry category, and the like. The content category refers to a category to which the recommended content set belongs, such as a primary content category of gourmet, movie and television, life and the like, or a secondary content category of a face point, a swordsmen drama, a square dance and the like. The industry category refers to a category of industry to which the recommended content belongs. By way of example only, the industry category may include a primary industry or a secondary industry, which is a fine-grained division of the primary industry. For example, a primary industry may include game classes, and a corresponding secondary industry may include entertainment, role playing, and the like. As another example, a primary industry may include shopping categories, and its corresponding secondary industry may include commodities, apparel, cosmetics, and the like.

It should be understood that multiple categories herein may refer to the same level of categories, such as multiple primary content categories, multiple secondary content categories, multiple primary industry categories, multiple secondary industry categories, and so forth. Of course, the class level is not limited thereto, and may include finer-grained class division.

The preset condition may be determined by a preset threshold value, which may be measured by content data for measuring the degree of goodness of the recommended content. Illustratively, the content data may include at least one dimension of a 3-second play rate, a content score, a negative rating, an end play rate, a negative feedback rate, a conversion rate, a 5-second play rate, and the like. Taking the recommended content as the video material as an example, the 3-second playing rate is the ratio of the number of users watching the video material for more than 3 seconds to the total number of users exposed to the video material. The content score refers to the content quality of the video material. The negative evaluation rate refers to a proportion of negative evaluation in the evaluation of the video material by the viewing user.

The preset thresholds correspond to the dimensions of the content data one to one, for example, the preset thresholds may include a preset 3-second play rate threshold k1, a preset content score threshold k2, a preset negative rating threshold k3, and the like. And determining whether the recommended content meets a preset condition according to the content data and the size of the corresponding preset threshold.

In an alternative embodiment, the obtaining of the recommended content set may include:

in step S2011, feedback information of recommended content corresponding to each category within a historical preset time period is acquired;

in step S2012, the feedback information is analyzed to obtain content data of each piece of recommended content in at least three dimensions;

in step S2013, determining recommended content meeting a preset condition according to the content data and a preset threshold;

in step S2014, the recommended content set is established according to the determined recommended content.

The feedback information is behavior data used for representing the recommended content of the user in a historical preset time period, for example, video data of user browsing, on-demand, praise, comment and the like. The historical preset time period may be set within the last month, within the last week, etc., and the present disclosure does not specifically limit this. The feedback information of the recommended contents of all industry categories in the historical preset time period is analyzed, so that the content data corresponding to each piece of recommended content is obtained, wherein the content data comprises at least one dimension of 3-second playing rate, content score, negative rating rate, playing completion rate, negative feedback rate, conversion rate, 5-second playing rate and the like.

In this embodiment, the feedback information may be analyzed to obtain content data in at least three dimensions, and then according to a size relationship between the content data and a preset threshold, recommended content in which the content data in the at least three dimensions simultaneously satisfy the preset threshold is determined as recommended content satisfying a preset condition; then, a recommended content set is constructed from all the recommended contents determined.

Preferably, taking the content data in three dimensions as an example, the three dimensions may include a 3-second play rate, a content score and a negative score. At this time, the correspondingly acquired content data includes a 3-second play rate, a content score, and a negative rating. When the 3-second play rate of the recommended content M is greater than a preset 3-second play rate threshold k1, the content score is less than a preset content score threshold k2, and the negative score is less than a preset negative score threshold k3, it is determined that the content quality data of the recommended content M satisfies a preset condition, that is, the recommended content M is a high-quality recommended content. And if any dimension of the playing rate, the content score and the negative rating rate in 3 seconds does not meet the corresponding preset condition, determining that the content data of the released content M does not meet the preset condition, namely that the recommended content M is non-high-quality recommended content. For example only, the preset 3-second play rate threshold k1 may be set to 0.5, the preset content score threshold k2 may be set to 0.35, and the preset negative rating threshold k3 may be set to 0.3.

In the embodiment, the content data of each piece of recommended content in at least three dimensions is obtained by analyzing the feedback information of the recommended content corresponding to each category in the historical preset time period; and then, establishing a recommended content set according to the content data and the recommended content which is determined by a preset threshold and meets a preset condition. Therefore, the constructed recommended content set is high-quality content, so that the loss of the user experience in the recommended content determining process is reduced, and the recommendation effect and the user satisfaction are improved.

In step S202, evaluation parameter information corresponding to each category is determined based on the history information of the history recommended content for the plurality of categories by the user.

The historical record information represents content information of various categories and corresponding operation information. The evaluation parameter information represents the estimated recommendation contribution degree of the recommendation contents in different categories.

Alternatively, the evaluation parameter information corresponding to various categories may be determined by a Bandit algorithm according to the history information of the history recommended content for various categories by the user.

In an example embodiment, the Bandit algorithm includes at least one of the following algorithms: a naive Bandit algorithm, a UCB (Upper Confidence Bound) algorithm, and variations thereof.

In an optional embodiment, as shown in fig. 3, the determining, according to the history information of the history recommended content for the multiple categories by the user, the evaluation parameter information corresponding to the various categories includes:

in step S301, history information of a user with respect to a plurality of categories of history recommended content is acquired; the historical record information comprises content characteristic information and historical operation times corresponding to each category;

in step S302, based on the content feature information, historical average evaluation parameter information corresponding to different categories is obtained;

in step S303, based on the historical operation frequency corresponding to each category and the total historical operation frequency of each category, an uncertain evaluation parameter corresponding to different categories is determined;

in step S304, evaluation parameter information corresponding to each category is obtained according to the historical average evaluation parameter information and the evaluation uncertainty parameter.

The content feature information may include, but is not limited to, an exposure amount, a click number, and a conversion number of the user viewing the released content of each industry category historically. According to the recommended contribution degree brought by the exposure, the number of clicks and the number of conversions, historical average evaluation parameter information under different categories can be measured. And then, determining the evaluation parameter information corresponding to the current user in various categories for the current access amount by combining the historical average evaluation parameter information and the evaluation uncertainty.

Optionally, the evaluation parameter information corresponding to various industries herein may include an evaluation parameter value, and correspondingly, the historical average evaluation parameter information may include a historical average evaluation parameter value. For example only, the evaluation parameter value may be determined by using a UCB (upper Confidence bound) algorithm, which is a selection strategy algorithm based on the current contribution degree and the number of times of exploration, and using the Confidence upper bound of the current interest point (for example, the second-level industry of the advertisement in the information flow scene) as the return pre-evaluation value. Specifically, the evaluation parameter value a resulting from the operation of each category is calculated separately from the history information of each category by trying one pass for each category. The iterative formula for evaluating parameter value a may be:

wherein, the first term Q (a) represents the historical average evaluation parameter value of the current category, and the second term

Indicating the evaluation uncertainty corresponding to the current category. a is the operation type, N represents the total historical operation times of each type, N represents the historical operation times of the current type, and A represents the estimated evaluation parameter value when the operation type is a. Operations herein include, but are not limited to, exploring for interest.

As can be seen from the above iterative formula, for the selected category, obtaining more exploration opportunities results in the second term in the formula

Becomes smaller and eventually smaller than the other non-selected categories. In addition, for non-selected categories, the confidence is high

Will increase as the number of rounds increases and will eventually be greater than the scores of the other selected categories. The UCB algorithm can be regarded as an optimization process, and multiple iterations are required to find a better scheme. In the decision making process, new knowledge is continuously added by gradually considering all uncertainties. Finally, in the time dimension, under uncertain interference, optimization can still be carried outAnd evaluating the objective function corresponding to the parameter value A.

In the embodiment, the historical average evaluation parameter information corresponding to different categories is determined according to the content characteristic information in the historical record information, and then, the evaluation uncertain parameters corresponding to different categories are determined according to the historical operation times in the historical record information corresponding to each category and the total historical operation times of each category; and obtaining evaluation parameter information corresponding to various categories according to the historical average evaluation parameter information and the evaluation uncertain parameters. Because the evaluation parameter information corresponding to each category is related to the historical average evaluation parameter information and the evaluation uncertain parameter, the historical average evaluation parameter information is in direct proportion to the content characteristic information, and the evaluation uncertain parameter is in inverse proportion to the historical operation times, the recommendation contribution degree and the operated times of the historical recommendation contents of different categories can be balanced, the determined evaluation parameter information can reflect the recommendation contribution degree brought by each category better, and the accuracy of determining the subsequent recommendation contents is improved.

In step S203, a target category is determined among the plurality of categories according to the evaluation parameter information corresponding to the respective categories.

Alternatively, when the evaluation parameter information corresponding to each category is determined, the target category may be determined among the plurality of categories according to the evaluation parameter information corresponding to each category (for example, the size of the evaluation parameter value a). The number of target categories determined herein may be one or more, and the present disclosure is not particularly limited thereto.

In an optional embodiment, the determining, according to the evaluation parameter information corresponding to each category, a target category in the categories may include:

in step S2021, determining evaluation parameter information satisfying a preset evaluation condition among the evaluation parameter information corresponding to each category as target evaluation parameter information;

in step S2022, the category corresponding to the target evaluation parameter information is determined as a target category.

Taking the example that the evaluation parameter information includes the evaluation parameter value, the preset evaluation condition may include, but is not limited to, at least one of the evaluation parameter value being greater than or equal to a preset evaluation parameter threshold value, TOP N evaluation parameter values (N is a positive integer), and the like. The target category may be a category corresponding to the evaluation parameter value in the evaluation parameter information reaching a preset evaluation parameter value, or may be at least one category with the evaluation parameter value sorted in a front order, which is not specifically limited by the present disclosure.

For example only, when determining the evaluation parameter information corresponding to each category, the maximum evaluation parameter value may be screened from the evaluation parameter information as the target evaluation parameter information, and then the category corresponding to the target evaluation parameter information is used as the target category, so as to facilitate further interest exploration for recommended content in the category with the maximum evaluation parameter value a.

In the embodiment, the category corresponding to the target evaluation parameter information which is made by the evaluation parameter information meeting the preset evaluation condition is determined as the target category, and the determined target category is the category with more recommendation contribution degrees, so that the determination quality of the recommendation content can be ensured, and the recommendation effect and the user satisfaction are further improved.

In step S204, a first recommended content belonging to the target category is determined from the recommended content set.

Optionally, when the target category is determined, the recommended content corresponding to the target category may be screened from the recommended content set, and used as the first recommended content. And subsequently, recommendation and interest exploration are carried out only aiming at the first recommended content without processing other recommended contents in the recommended content set, so that the workload can be greatly reduced, and the recommendation quality and effect can be improved.

In step S205, a second recommended content recalled for the target user is acquired, and the target recommended content is determined based on the first recommended content and the second recommended content.

Alternatively, the target user may be determined before the second recommended content is determined. The target user refers to a user who needs to make a recommendation or an interest exploration. For example only, the target user may satisfy at least one of the following conditions: the user with the user activity degree greater than or equal to the preset activity degree threshold, the user with the application program use time length greater than or equal to the preset time length threshold, the user with the application program cold start times greater than or equal to the preset time threshold, and the user with the media content filling rate less than or equal to the preset filling rate threshold. The media content may include, but is not limited to, advertisements, among others. It should be understood that the conditions satisfied by the target user are not limited thereto.

After the target user is determined, the corresponding recommended content may be recalled for the target user as the second recommended content according to the user characteristic data of the target user. The user characteristic data may include behavior data of the user, user interests, gender and the like. Then, target recommended content for recommending to the target user is determined according to the first recommended content and the second recommended content obtained from the recommended content set.

For example only, the target recommended content may include, but is not limited to, a union or an intersection of the first recommended content and the second recommended content, or recommended content that is included in a union of the first recommended content and the second recommended content and satisfies a preset recommendation condition. The preset recommendation condition may include, but is not limited to, a user portrait condition, a recommendation demand condition, a recommended content quality condition, and the like.

In an optional embodiment, the determining the target recommended content based on the first recommended content and the second recommended content may include:

in step S2051, based on the first recommended content and the second recommended content, candidate recommended content that satisfies a preset targeting condition is determined.

Optionally, before determining the target recommended content, candidate recommended content meeting a preset targeting condition may be determined according to the first recommended content and the second recommended content obtained from the recommended content set, and then the target recommended content may be determined according to the candidate recommended content. Taking the recommended content as the advertisement material, the preset targeting condition may refer to a targeting user condition for the advertiser to place the advertisement, such as user gender, age, and the like.

In an optional embodiment, as shown in fig. 4, the determining, based on the first recommended content and the second recommended content, a candidate recommended content that satisfies a preset targeting condition includes:

in step S401, correspondingly configuring the first recommended content and the second recommended content to a first channel and at least one second channel, respectively, where each channel includes a content tag for representing an attribute of the recommended content;

in step S402, merging the recommended contents in the first channel and the at least one second channel to obtain a merged recommended content set;

in step S403, determining an expected recommended content set according to the user characteristic information corresponding to the target user;

in step S404, according to a preset orientation condition of the merged recommended content, performing intersection processing on the merged recommended content set and the expected recommended content set to obtain a candidate recommended content set;

in step S405, the set of candidate recommended contents is taken as candidate recommended contents satisfying a preset targeting condition.

Optionally, the first recommended content is configured to a first channel of the plurality of channels on the line via Redis (a cross-platform non-relational database), and the second recommended content is configured to at least one second channel of the plurality of channels on the line. Each channel includes a content tag for characterizing the attributes of the recommended content to mark which channel the recommended content belongs to.

For example, the first channel includes a first content tag characterizing the first recommended content, i.e., a premium content tag. The second channel includes at least one second content tag characterizing the second recommended content, which may be, for example, a hotspot content tag, a personalized content tag, and so forth. And then, combining the recommended contents in the first channel and the at least one second channel to obtain a combined recommended content set. The merged recommended content set comprises first recommended content carrying high-quality content labels and second recommended content carrying other content labels. Then, according to the user characteristic information corresponding to the target user, a set of desired recommended content is determined, for example, the set of desired recommended content may include { women, age 24, hobby sports }. Taking the recommended content as an advertisement material as an example, each merged recommended content is preset with a targeting condition, for example, for the recommended content c1 in the merged recommended content, the preset targeting condition is { male, older than 30 years old, hobby for food }; for the recommended content c2 in the merged recommended content, the preset orientation condition is { women, under the age of 25, travel hobbies } and the like. And performing intersection processing on the merged recommended content set and the expected recommended content set according to the preset orientation condition of the merged recommended content, namely screening the merged recommended content to obtain the merged recommended content matched with the target user to form a candidate recommended content set, and taking the candidate recommended content set as the candidate recommended content meeting the preset orientation condition.

According to the embodiment, the first recommended content and the second recommended content are respectively configured on different channels, and different channels carry different content tags, so that the source of each recommended content can be quickly determined, differential processing can be performed on different recommended contents, and the calculation amount is reduced. Meanwhile, the candidate recommended content is directionally determined by intersection processing according to the combined recommended content of the multiple channels and the expected recommended content set determined by the user characteristic information, so that more potential interest tags of the target user are fully mined, the recommended quality of the candidate recommended content is ensured, the quality of the target recommended content is improved, and the recommendation effect and the user satisfaction are further improved.

In step S2052, based on a preset recommendation policy, the candidate recommended content is processed to obtain a target recommended content for recommending to the target user.

Optionally, after determining the candidate recommended content, at least one of a filtering process, a coarse ranking and a fine ranking may be performed on the candidate recommended content, so as to filter and sort the candidate recommended content, and obtain the target recommended content for recommendation to the target user. Among these, screening processes may include, but are not limited to: at least one processing mode of diversity filtering, frequency control, reward strategy and the like.

In an optional embodiment, the determined candidate recommended content is subjected to screening processing and then sorted, and the candidate recommended content with the preset number M (M is a positive integer) before the sorting result is subjected to rough sorting and fine sorting in sequence, so as to obtain the target recommended content for recommending to the target user.

In another optional embodiment, as shown in fig. 5, the preset recommendation policy includes a first recommendation policy and a second recommendation policy, and the content tag includes a first content tag representing first recommended content; the processing the candidate recommended content based on the preset recommendation strategy to obtain the target recommended content for recommending to the target user comprises:

in step S501, based on the first recommendation policy, filtering and sorting candidate recommended contents except for the candidate recommended contents carrying the first content tag to obtain a first subset to be recommended;

in step S502, based on the second recommendation policy, filtering and sorting the first subset to be recommended and the candidate recommended content carrying the first content tag to obtain a second subset to be recommended;

in step S503, in the case that the second subset to be recommended is a non-empty set, taking the second subset to be recommended as target recommended content for recommending to the target user;

in step S504, when the second subset to be recommended is an empty set, the candidate recommended content carrying the first content tag is used as a target recommended content for recommending to the target user.

Optionally, for the candidate recommended content set obtained by the directional intersection processing, if the candidate recommended content set contains recommended content carrying the first content tag, the recommended content carrying the first content tag is directly preserved to the refined ranking, that is, only the second recommendation policy needs to be executed on the recommended content carrying the first content tag, and the first recommendation policy does not need to be executed on the recommended content carrying the first content tag. The first recommendation strategy can be a rough ranking strategy, and the second recommendation strategy can be a fine ranking strategy. When the second recommendation strategy is executed, filtering and sorting can be performed according to UEQ (user score), CPM and other index scores, and a second subset to be recommended is obtained.

And if the second subset to be recommended is a non-empty set, taking the second subset to be recommended as target recommended content for recommending to the target user, so that exposure can be performed according to the arrangement sequence of the second subset to be recommended, and the target user is recommended. If the second subset to be recommended is an empty set, it indicates that all candidate recommended contents are excluded by fine exclusion, and at this time, the candidate recommended contents carrying the first content tag may be used as target recommended contents for recommending to the target user, so that at least one candidate recommended content carrying the first content tag may be randomly selected for exposure and recommended to the target user.

In an optional embodiment, the, in the case that the second subset to be recommended is a non-empty set, using the second subset to be recommended as the target recommendation content for recommendation to the target user may include:

and taking the processed second subset to be recommended as target recommended content for recommending to the target user.

Optionally, after the second recommendation subset is obtained, diversity processing may be performed on the second recommendation subset to reduce repeated recommended content, and then the diversity-processed second to-be-recommended subset is used as the target recommended content for recommending to the target user, so that the quality of the target recommended content may be improved, and the recommendation effect and the user satisfaction may be further improved.

In an optional embodiment, when the second subset to be recommended is an empty set, taking the candidate recommended content carrying the first content tag as the target recommended content for recommendation to the target user may include:

Optionally, under the condition that the second recommendation subset is an empty set, diversity processing may be performed on candidate recommendation contents carrying the first content tag, repeated recommendation contents are reduced, and then the diversity-processed third subset to be recommended is used as target recommendation contents for recommending to the target user, so that the quality of the target recommendation contents can be improved, and the recommendation effect and the user satisfaction degree are further improved.

In the embodiment, the candidate recommended contents of different content labels are differentiated through the first recommendation strategy and the second recommendation strategy, and different recommended contents are used as target recommended contents according to the condition that the second subset to be recommended is a non-empty set or an empty set, so that the quality of the recommended contents is improved, the quality of the target recommended contents is improved, and the recommendation effect and the user satisfaction are further improved.

In an optional embodiment, as shown in fig. 6, after the step of filtering and sorting the candidate recommended content based on a preset recommendation policy to obtain a target recommended content for recommendation to the target user, the method further includes:

in step S601, posterior behavior data of the target user for the target recommended content is obtained;

in step S602, analyzing the extended interest tag of the target user according to the posterior behavior data;

in step S603, the target recommended content is filtered based on the extended interest tag, and updated target recommended content is obtained.

The posterior behavior data is used for reflecting the interest condition of the target user in the target recommendation content. The a posteriori behavior data may include, but is not limited to, viewing duration, clicks, likes, masks, shares, etc. data. And analyzing the posterior behavior data of the target user to obtain the extended interest tag of the target user. The extended interest tags reflect the mined user potential interest points. And then, filtering the target recommended content by using the extended interest tag to obtain updated target recommended content which is more in line with the interest of the target user.

According to the embodiment, the posterior behavior data of the target user on the target recommended content is obtained; and analyzing according to the posterior behavior data to obtain an extended interest tag of the target user, and then filtering the target recommended content based on the extended interest tag to obtain updated target recommended content. Therefore, the determined target recommendation content is obtained by filtering and updating according to the posterior behavior data of the target user, the requirement or interest point of the target user is better met, the recommendation effect and the user satisfaction are further improved, and the user viscosity is improved.

Fig. 7 is a flowchart illustrating a recommendation method according to an exemplary embodiment, and as shown in fig. 7, the recommendation method may be applied to an electronic device, which is described as an example of a server in the above implementation environment schematic diagram, and includes the following steps.

In step S701, a target user and a target recommended content for recommending to the target user are obtained, where the target recommended content is obtained according to any one of the recommended content determination methods.

In an alternative embodiment, the target user satisfies at least one of the following conditions: the user with the user activity degree greater than or equal to the preset activity degree threshold, the user with the application program use time length greater than or equal to the preset time length threshold, the user with the application program cold start times greater than or equal to the preset time threshold, and the user with the media content filling rate less than or equal to the preset filling rate threshold.

In step S702, the target recommended content is recommended to the target user.

Optionally, after the target user and the target recommended content are determined, the target recommended content may be recommended to the corresponding target user.

Fig. 8 is a block diagram illustrating a recommended content determining apparatus according to an exemplary embodiment. Referring to fig. 8, the apparatus is applied to an electronic device, and includes:

a first obtaining module 810 configured to perform obtaining of a recommended content set, where the recommended content set includes recommended contents of multiple categories and satisfies a preset condition;

a first determining module 820 configured to determine evaluation parameter information corresponding to various categories according to the history information of the history recommended content of the various categories by the user; the historical record information represents content information of various categories and corresponding operation information;

a second determining module 830 configured to determine a target category among the multiple categories according to the evaluation parameter information corresponding to the various categories;

a third determining module 840 configured to perform determining a first recommended content belonging to the target category from the set of recommended contents;

and the target content determining module 850 is configured to execute acquiring second recommended content recalled for the target user and determine the target recommended content based on the first recommended content and the second recommended content.

As an optional implementation manner, the first obtaining module includes:

As an optional implementation, the first determining module includes:

As an optional implementation manner, the second determining module includes:

As an optional implementation manner, the third determining module includes:

As an optional implementation manner, the preset recommendation policy includes a first recommendation policy and a second recommendation policy, and the content tag includes a first content tag representing first recommended content; the target content determination sub-module includes:

the second target content determination unit includes:

As an optional implementation, the apparatus further comprises:

FIG. 9 is a block diagram illustrating a recommendation device according to an example embodiment. Referring to fig. 9, the apparatus is applied to an electronic device, and includes:

an obtaining module 910, configured to perform obtaining of a target user and target recommended content recommended to the target user, where the target recommended content is obtained according to any one of the recommended content determining apparatuses;

a recommending module 920 configured to perform recommending the target recommended content to the target user.

With regard to the apparatus in the above-described embodiment, the specific manner in which each module performs the operation has been described in detail in the embodiment related to the method, and will not be elaborated here.

In an exemplary embodiment, there is also provided an electronic device, comprising a processor; a memory for storing processor-executable instructions; wherein the processor is configured to implement the steps of any of the recommended content determining methods or the recommending methods in the above embodiments when executing the instructions stored on the memory.

The electronic device may be a terminal, a server, or a similar computing device, taking the electronic device as a server as an example, fig. 10 is a block diagram of an electronic device for determining or recommending recommended content according to an exemplary embodiment, where the electronic device 1000 may have a relatively large difference due to different configurations or performances, and may include one or more Central Processing Units (CPUs) 1010 (the processors 1010 may include but are not limited to Processing devices such as a microprocessor MCU or a programmable logic device FPGA), a memory 1030 for storing data, and one or more storage media 1020 (e.g., one or more mass storage devices) for storing application programs 1023 or data 1022. Memory 1030 and storage media 1020 may be, among other things, transient or persistent storage. The program stored in the storage medium 1020 may include one or more modules, each of which may include a sequence of instructions operating on an electronic device. Still further, the central processor 1010 may be configured to communicate with the storage medium 1020 to execute a series of instruction operations in the storage medium 1020 on the electronic device 1000.

The electronic device 1000 may also include one or more power supplies 1060, one or more wired or wireless network interfaces 1050, one or more input-output interfaces 1040, and/or one or more operating systems 1021, such as Windows Server, Mac OS X, Unix, Linux, FreeBSD, and so forth.

Input-output interface 1040 may be used to receive or transmit data via a network. Specific examples of the network described above may include a wireless network provided by a communication provider of the electronic device 1000. In one example, i/o Interface 1040 includes a Network adapter (NIC) that may be coupled to other Network devices via a base station to communicate with the internet. In an exemplary embodiment, the input/output interface 1040 may be a Radio Frequency (RF) module for communicating with the internet in a wireless manner.

It will be understood by those skilled in the art that the structure shown in fig. 10 is merely an illustration and is not intended to limit the structure of the electronic device. For example, the electronic device 1000 may also include more or fewer components than shown in FIG. 10, or have a different configuration than shown in FIG. 10.

In an exemplary embodiment, a computer-readable storage medium comprising instructions, such as a memory comprising instructions, executable by a processor of the electronic device 1000 to perform the above-described method is also provided. Alternatively, the computer readable storage medium may be a ROM, a Random Access Memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, and the like.

In an exemplary embodiment, a computer program product is also provided, which comprises a computer program or instructions, which when executed by a processor, implement the method provided in any of the above embodiments. Optionally, the computer program or instructions are stored in a computer readable storage medium. The processor of the electronic device reads the computer program or instructions from the computer-readable storage medium, and the processor executes the computer program or instructions, so that the electronic device executes the method provided in any one of the above embodiments.

Other embodiments of the disclosure will be apparent to those skilled in the art from consideration of the specification and practice of the disclosure disclosed herein. This application is intended to cover any variations, uses, or adaptations of the disclosure following, in general, the principles of the disclosure and including such departures from the present disclosure as come within known or customary practice within the art to which the disclosure pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the disclosure being indicated by the following claims.

It will be understood that the present disclosure is not limited to the precise arrangements described above and shown in the drawings and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. A recommended content determining method, comprising:

2. The method according to claim 1, wherein the obtaining of the recommended content set comprises:

3. The method for determining recommended content according to claim 1, wherein the determining, based on the historical record information of the user regarding the multiple categories of historical recommended content, the evaluation parameter information corresponding to each category includes:

4. The method for determining recommended content according to any one of claims 1 to 3, wherein the determining a target category among the plurality of categories based on the evaluation parameter information corresponding to the respective categories includes:

5. A recommendation method, comprising:

acquiring a target user and target recommended content for recommending to the target user, wherein the target recommended content is obtained according to the recommended content determination method of any one of claims 1-4;

recommending the target recommended content to the target user.

6. A recommended content determining apparatus, comprising:

7. A recommendation device, comprising:

an obtaining module configured to perform obtaining a target user and target recommended content for recommending to the target user, the target recommended content being obtained by the recommended content determining apparatus according to claim 6;

8. An electronic device, comprising:

a processor;

a memory for storing the processor-executable instructions;

wherein the processor is configured to execute the instructions to implement the recommended content determination method of any one of claims 1 to 4 or the recommendation method of claim 5.

9. A computer-readable storage medium in which instructions, when executed by a processor of an electronic device, enable the electronic device to perform the recommended content determination method of any one of claims 1 to 4 or the recommendation method of claim 5.

10. A computer program product comprising a computer program, wherein the computer program, when executed by a processor, implements the recommended content determination method of any one of claims 1 to 4 or the recommendation method of claim 5.