An adaptive recommendation method and device based on user interest

Abstract

This application provides an adaptive recommendation method and apparatus based on user interests. The method includes: obtaining user interest tags and corresponding interest weight vectors based on pre-acquired user behavior data, wherein the interest weight vectors represent the degree of user interest in the user interest tags; calculating a user interest dispersion score based on the interest weight vectors, and determining a content recommendation mechanism based on the user interest dispersion score; and recommending content to the user based on the content recommendation mechanism. Therefore, this application embodiment determines a corresponding content push mechanism for users based on user interests and the dispersion of those interests, adaptively adjusting the recommended content according to the user's interest dispersion, thereby achieving efficient recommendation of content based on user interests and improving the data conversion rate in recommendation scenarios.

Description

Technical Field

[0001] This application relates to the field of intelligent recommendation technology, and in particular to an adaptive recommendation method and apparatus based on user interests. Background Technology

[0002] With the development of internet technology, the internet can provide users with a wide variety of online services. On internet platforms, users can search for the content they need. Furthermore, to facilitate content access, internet platforms can proactively recommend content to users based on their behavior.

[0003] In existing technologies, content recommendation methods generally calculate user interest tags based on user behavior and perform hierarchical processing of the tags. Interest exploration is carried out within the user's hierarchical tags. This type of exploration is based on the expansion exploration of multiple dimensions of user tags. For example, if the interest tag matched for user A is tag 'a', and tag 'a' and tag 'b' have a high correlation, then content corresponding to tag 'b' will be recommended to user A.

[0004] While existing content recommendation methods can satisfy users' needs for content recommendations based on a specific interest dimension, they fail to consider the impact of the concentration and dispersion of users' interests on content recommendations. This can easily lead to recommending monotonous content that results in low conversion rates in business scenarios. Therefore, there is an urgent need for a content adaptive recommendation method that can adjust the recommended content based on user interests, thereby achieving efficient recommendation of content that aligns with user interests and improving data conversion rates in recommendation scenarios. Summary of the Invention

[0005] In view of this, this application provides an adaptive recommendation method and apparatus based on user interests to achieve efficient recommendation of content based on user interests and improve the data conversion rate in recommendation scenarios.

[0006] The technical solution is as follows:

[0007] In a first aspect, embodiments of this application provide an adaptive recommendation method based on user interests, the method comprising:

[0008] User interest tags and corresponding interest weight vectors are obtained based on pre-acquired user behavior data. The interest weight vectors are used to characterize the degree of user interest in the user interest tags.

[0009] The user interest dispersion score is calculated based on the interest weight vector, and the content recommendation mechanism is determined based on the user interest dispersion score.

[0010] Content is recommended to users based on the aforementioned content recommendation mechanism.

[0011] Optionally, the user behavior data includes: explicit user behavior data and implicit user behavior data; wherein the explicit user behavior data includes at least one of: historical search data, historical viewing data, historical like data, historical sharing data, and user collection data, and the implicit user behavior data includes: exposure without click data.

[0012] Optionally, obtaining user interest tags and corresponding interest weight vectors based on pre-acquired user behavior data includes:

[0013] The user interest tags are determined based on the user behavior data;

[0014] Calculate the historical interest weight vector of the user interest tag based on the user behavior data and the preset user behavior weights;

[0015] The historical interest weight vector is reduced in weight using a pre-defined preprocessing model to obtain the interest weight vector.

[0016] Optionally, calculating the user interest dispersion score based on the interest weight vector includes:

[0017] The interest weight vector is normalized to obtain the first interest weight vector.

[0018] The dimension values ​​in the normalized interest weight vector that are less than the preset interest weight threshold are set to zero, while the dimension values ​​in the normalized interest weight vector that are greater than or equal to the preset interest weight threshold are maintained to obtain the second interest weight vector.

[0019] The user interest dispersion score is calculated based on a preset user interest dispersion score calculation model, which is as follows: Where F is the user interest dispersion score, A is the number of non-zero dimensions in the second interest weight vector, and B is the variance of the non-zero data in the second interest weight vector.

[0020] Optionally, the content recommendation mechanism based on the user interest dispersion score includes:

[0021] If the user interest dispersion score is greater than the preset interest dispersion confidence value, then the content recommendation mechanism is determined to be a priority content recommendation mechanism.

[0022] If the user interest dispersion score is less than or equal to the preset interest dispersion confidence value, then the content recommendation mechanism is determined to be a priority interest content recommendation mechanism.

[0023] Secondly, embodiments of this application provide an adaptive recommendation device based on user interests, the device comprising:

[0024] The acquisition module is used to acquire user interest tags and interest weight vectors corresponding to the user interest tags based on pre-acquired user behavior data. The interest weight vectors are used to characterize the degree of user interest in the user interest tags.

[0025] The calculation module is used to calculate the user interest dispersion score based on the interest weight vector, and determine the content recommendation mechanism based on the user interest dispersion score.

[0026] The recommendation module is used to recommend content to users based on the content recommendation mechanism.

[0027] Optionally, the user behavior data includes: explicit user behavior data and implicit user behavior data; wherein the explicit user behavior data includes at least one of: historical search data, historical viewing data, historical like data, historical sharing data, and user collection data, and the implicit user behavior data includes: exposure without click data.

[0028] Optionally, the acquisition module is specifically used for:

[0029] The user interest tags are determined based on the user behavior data;

[0030] Calculate the historical interest weight vector of the user interest tag based on the user behavior data and the preset user behavior weights;

[0031] The historical interest weight vector is reduced in weight using a pre-defined preprocessing model to obtain the interest weight vector.

[0032] Optionally, the calculation module includes a score calculation submodule; the score calculation submodule is specifically used for:

[0033] The interest weight vector is normalized to obtain the first interest weight vector.

[0034] The dimension values ​​in the normalized interest weight vector that are less than the preset interest weight threshold are set to zero, while the dimension values ​​in the normalized interest weight vector that are greater than or equal to the preset interest weight threshold are maintained to obtain the second interest weight vector.

[0035] The user interest dispersion score is calculated based on a preset user interest dispersion score calculation model, which is as follows: Where F is the user interest dispersion score, A is the number of non-zero dimensions in the second interest weight vector, and B is the variance of the non-zero data in the second interest weight vector.

[0036] Optionally, the calculation module includes a mechanism determination submodule; the mechanism determination submodule is specifically used for:

[0037] If the user interest dispersion score is greater than the preset interest dispersion confidence value, then the content recommendation mechanism is determined to be a priority content recommendation mechanism.

[0038] If the user interest dispersion score is less than or equal to the preset interest dispersion confidence value, then the content recommendation mechanism is determined to be a priority interest content recommendation mechanism.

[0039] The above technical solution has the following beneficial effects:

[0040] This application provides an adaptive recommendation method based on user interests. When executing the method, user interest tags and corresponding interest weight vectors are obtained based on pre-acquired user behavior data. The interest weight vectors represent the degree of user interest in the user interest tags. A user interest dispersion score is calculated based on the interest weight vectors, and a content recommendation mechanism is determined based on the user interest dispersion score. Content is recommended to the user based on the content recommendation mechanism. Therefore, this application determines a corresponding content push mechanism for users based on their interests and the dispersion of those interests, adaptively adjusting the recommended content according to the user's interest dispersion, achieving efficient recommendation of user-interested content and improving the data conversion rate in recommendation scenarios.

[0041] This application also provides an apparatus corresponding to the above method, which has the same beneficial effects as the above method. Attached Figure Description

[0042] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0043] Figure 1 A flowchart illustrating an adaptive recommendation method based on user interests provided in an embodiment of this application;

[0044] Figure 2 This is a schematic diagram of the structure of an adaptive recommendation device based on user interests, provided in an embodiment of this application. Detailed Implementation

[0045] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0046] To achieve efficient recommendation of content based on user interests and improve the data conversion rate in recommendation scenarios, this application provides an adaptive recommendation method based on user interests. Please refer to [link to relevant documentation]. Figure 1 The method may include:

[0047] Step S100: Obtain user interest tags and the interest weight vectors corresponding to the user interest tags based on the pre-acquired user behavior data.

[0048] Specifically, user interest tags and their corresponding interest weight vectors are obtained based on pre-acquired user behavior data. The interest weight vectors can represent the degree of user interest in the user's interest tags.

[0049] User behavior data refers to data generated by users when using internet products. For example, for video websites, user behavior data could include a user's viewing history, likes history, and favorites history. In this embodiment, user behavior data can be further divided into explicit user behavior data and implicit user behavior data. Explicit user behavior data refers to behavior data that directly reflects a user's interests, such as likes and dislikes. Implicit user behavior data, on the other hand, does not directly reflect a user's interests but can indirectly reflect their preferences, such as views without clicks.

[0050] As an option, explicit user behavior data in this application embodiment may include at least one of the following: historical search data, historical viewing data, historical like data, historical sharing data, and user collection data; implicit user behavior data may include: exposure without click data.

[0051] It is understood that the above user behavior data is only an example, and those skilled in the art can adjust the categories of user behavior data according to their needs, such as adding effective viewing time data, comment viewing behavior data, follow behavior data, unfollow behavior data, etc.

[0052] It should be noted that user behavior data can be obtained by referring to existing mature technologies, such as collecting relevant behavior logs through code tracking to obtain initial user behavior data. After obtaining the initial user behavior data, it can also be cleaned and filtered. For example, data with viewing time less than a preset threshold can be extracted to ensure the accuracy of the user interest tags obtained later.

[0053] In an optional implementation, step S100 may include steps S101 to S103, as follows:

[0054] Step S101: Determine user interest tags based on user behavior data.

[0055] Specifically, user interest tags can be obtained from a pre-established tag library through a preset mapping relationship between behavioral data and tags, or through data segmentation and extraction methods. There are no restrictions on the method of obtaining user interest tags.

[0056] Step S102: Calculate the historical interest weight vector of user interest tags based on user behavior data and preset user behavior weights.

[0057] Specifically, in this application embodiment, weight values ​​are predefined for each user interest tag. Since the acquired user behavior data is data before the current time node, the weight of the user interest tag calculated is also the historical interest weight of the user interest tag, that is, the corresponding historical interest weight vector.

[0058] Step S103: Use a preset preprocessing model to reduce the weight of the historical interest weight vector to obtain the interest weight vector.

[0059] Specifically, since the obtained historical interest weight vector represents the degree of interest of a user on a certain user interest tag at a historical moment, the embodiments of this application can also use a preset preprocessing model to reduce the weight of the historical interest weight vector to obtain the user's current interest weight vector. It can be understood that the user's current interest weight vector is used to represent the degree of interest of the user on a certain interest tag at the current time node.

[0060] In practical applications, the preset preprocessing model can adopt a time decay model and a hotspot weighting model. Specifically, the time decay model is y x*24 Where D is the attenuation coefficient, which is greater than zero and less than 1. As an option, D can be 0.996. x is the number of times the action has occurred since the current time, and x is a positive integer greater than or equal to zero. The hotspot weighting model is as follows: Where hot is greater than 0 and less than or equal to 1, x is the ranking of the platform resources in the preset time period, and a is the preset coefficient, which can be 0.001.

[0061] It should be noted that other existing weighting models can also be used for preprocessing, but this is limited here.

[0062] Through the above steps S101 to S103, user interest tags and the interest weight vectors corresponding to those user interest tags can be obtained.

[0063] Step S200: Calculate the user interest dispersion score based on the interest weight vector, and determine the content recommendation mechanism based on the user interest dispersion score.

[0064] Specifically, the aforementioned steps have already obtained the user's interest tags and their corresponding interest weight vectors on the platform. Based on the calculated interest weight vectors, a user interest dispersion score can be calculated. The user interest dispersion score reflects the degree of dispersion of the user's interests. A higher user interest dispersion score indicates a broader range of user interests, allowing for more exploratory content recommendations to be made for that user, thus strengthening the dispersion during the recommendation re-ranking process. A lower user interest dispersion score indicates a more concentrated user interest, allowing for the recommendation of vertical content related to the user's interest categories, appropriately reducing exploration, decreasing the dispersion during recommendation re-ranking, and improving the accuracy of recommendations.

[0065] In one optional implementation, calculating the user's interest dispersion score based on the interest weight vector may include steps S201 to S203, as follows:

[0066] Step S201: Normalize the interest weight vector to obtain the first interest weight vector.

[0067] Step S202: Set the dimension values ​​in the normalized interest weight vector that are less than the preset interest weight threshold to zero, and maintain the dimension values ​​in the normalized interest weight vector that are greater than or equal to the preset interest weight threshold to obtain the second interest weight vector.

[0068] Step S203: Calculate the user interest dispersion score based on the preset user interest dispersion score calculation model. The user interest dispersion score calculation model is as follows: Where F is the user interest dispersion score, A is the number of non-zero dimensions in the second interest weight vector, and B is the variance of the non-zero data in the second interest weight vector.

[0069] The following specific example illustrates steps S201 to S203.

[0070] Assume the user's interest weight vector is {2.5 3.5 0.5 1.5}, which are the interest weight vectors of user interest tags A, B, C, and D respectively, and the preset interest weight threshold is 0.3.

[0071] Step (1): After normalizing the interest weight vector, the first interest weight vector is obtained as {0.3125 0.4375 0.0625 0.1875}. Normalization is a mature technology and will not be elaborated here.

[0072] Step (2): The first interest weight vector is truncated by a threshold to obtain the second interest weight vector as {0.3125 0.4375 0 0}.

[0073] Step (3): Calculate the user interest dispersion score based on the user interest dispersion score calculation model.

[0074] It should be noted that the above example of calculating user interest dispersion score is only used to illustrate the calculation steps, and the data involved is for reference only and does not constitute a specific limitation.

[0075] In one optional implementation, determining the content recommendation mechanism based on the user's interest dispersion score may specifically include: if the user's interest dispersion score is greater than a preset interest dispersion confidence value, then the content recommendation mechanism is determined to be a priority exploration content recommendation mechanism; if the user's interest dispersion score is less than or equal to the preset interest dispersion confidence value, then the content recommendation mechanism is determined to be a priority interest content recommendation mechanism.

[0076] It is understandable that the preset confidence value of interest dispersion can be dynamically adjusted according to the actual business situation and the platform's promotion and new product launches, and this application does not limit its specific value.

[0077] It should be noted that a user's interest tag can be one or multiple. As a user's usage time on the platform increases, the more historical behaviors they exhibit, and the wider the range of tags they have encountered, the more accurate the subsequently calculated user interest dispersion score will be. When a user has few interest tags, such as only one, subsequent recommendations based on user interest dispersion may result in low data conversion rates, low exploration levels, and decreased user stickiness. Therefore, specific content recommendation methods can be set up for newly registered users, such as prioritizing exploration-based content recommendations. After a user has accumulated a certain amount of historical behavior and encountered a preset number of tags, the content recommendation method can be determined based on the user interest dispersion score, as described in the embodiments of this application.

[0078] Step S300: Recommend content to users based on the content recommendation mechanism.

[0079] Specifically, when the content recommendation mechanism prioritizes exploration, we can increase interest exploration in recommendations for these users, strengthen the dispersion during recommendation re-ranking, and design a wider range of content for users to explore their own interests, thereby improving the data conversion rate of the recommendation scenario. When the content recommendation mechanism is determined to prioritize interest-based content, we can recommend vertical content related to interest categories to these users, appropriately reduce exploration, reduce the dispersion of recommendation re-ranking, and improve the accuracy of recommendations.

[0080] In summary, this application provides an adaptive recommendation method based on user interests. When executing this method, user interest tags and corresponding interest weight vectors are obtained based on pre-acquired user behavior data. The interest weight vectors represent the degree of user interest in the user interest tags. A user interest dispersion score is calculated based on the interest weight vectors, and a content recommendation mechanism is determined based on the user interest dispersion score. Content is recommended to the user based on the content recommendation mechanism. Therefore, this application determines a corresponding content push mechanism for the user based on user interests and the dispersion of those interests, adaptively adjusting the recommended content according to the user's interest dispersion, achieving efficient recommendation of user-interested content and improving the data conversion rate in recommendation scenarios.

[0081] Corresponding to the above methods, embodiments of this application also provide an adaptive recommendation device based on user interests. Figure 2 A schematic diagram of the device is shown, which may include:

[0082] The acquisition module 201 is used to acquire user interest tags and interest weight vectors corresponding to user interest tags based on pre-acquired user behavior data. The interest weight vector is used to represent the degree of user interest in user interest tags.

[0083] The calculation module 202 is used to calculate the user interest dispersion score based on the interest weight vector, and to determine the content recommendation mechanism based on the user interest dispersion score.

[0084] Recommendation module 203 is used to recommend content to users based on a content recommendation mechanism.

[0085] As an optional implementation method, user behavior data includes: explicit user behavior data and implicit user behavior data; wherein explicit user behavior data includes at least one of: historical search data, historical viewing data, historical like data, historical sharing data, and user collection data, and implicit user behavior data includes: exposure without click data.

[0086] As an optional implementation, the acquisition module 201 is specifically used for:

[0087] Determine user interest tags based on user behavior data;

[0088] Calculate the historical interest weight vector of user interest tags based on user behavior data and preset user behavior weights;

[0089] The historical interest weight vector is reduced in weight using a pre-defined preprocessing model to obtain the interest weight vector.

[0090] As an optional implementation, the calculation module 202 includes a score calculation submodule; the score calculation submodule is specifically used for:

[0091] The interest weight vector is normalized to obtain the first interest weight vector.

[0092] Set the dimension values ​​in the normalized interest weight vector that are less than the preset interest weight threshold to zero, and maintain the dimension values ​​in the normalized interest weight vector that are greater than or equal to the preset interest weight threshold to obtain the second interest weight vector.

[0093] The user interest dispersion score is calculated based on a pre-defined user interest dispersion score calculation model. The user interest dispersion score calculation model is as follows: Where F is the user interest dispersion score, A is the number of non-zero dimensions in the second interest weight vector, and B is the variance of the non-zero data in the second interest weight vector.

[0094] As an optional implementation, the calculation module 202 further includes a mechanism determination submodule; the mechanism determination submodule is specifically used for:

[0095] If the user's interest dispersion score is greater than the preset interest dispersion confidence value, then the content recommendation mechanism is determined to be the priority content recommendation mechanism to be explored.

[0096] If a user's interest dispersion score is less than or equal to the preset interest dispersion confidence value, then the content recommendation mechanism is determined to be a priority interest content recommendation mechanism.

[0097] It should be noted that the steps and related technical features of each module in the adaptive recommendation device based on user interests provided in this application correspond to the method provided in the application embodiment. The description of the device part can be found in the embodiments of the aforementioned method part, and will not be repeated here.

[0098] In summary, this application provides an adaptive recommendation device based on user interests. The device includes: an acquisition module, used to acquire user interest tags and corresponding interest weight vectors based on pre-acquired user behavior data, whereby the interest weight vectors characterize the degree of user interest in the user interest tags; a calculation module, used to calculate a user interest dispersion score based on the interest weight vectors and determine a content recommendation mechanism based on the user interest dispersion score; and a recommendation module, used to recommend content to the user based on the content recommendation mechanism. Therefore, this application determines a corresponding content push mechanism for the user based on user interests and the dispersion of those interests, adaptively adjusting the recommended content according to the user's interest dispersion, thereby achieving efficient recommendation of user-interested content and improving the data conversion rate in recommendation scenarios.

[0099] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0100] Those skilled in the art will understand that the flowchart shown is merely an example in which the embodiments of this application can be implemented, and the scope of application of the embodiments of this application is not limited by any aspect of the flowchart.

[0101] In the several embodiments provided in this application, it should be understood that the disclosed methods, apparatuses, and devices can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual couplings or direct couplings or communication connections may be through some communication interfaces; indirect couplings or communication connections between devices or units may be electrical, mechanical, or other forms.

[0102] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, the functional units in the various embodiments of this application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0103] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0104] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An adaptive recommendation method based on user interests, characterized in that, The method includes: When the number of tags a user has historically encountered has not reached a preset number, content is recommended to the user based on a priority content exploration recommendation method. After a user's historical contact with tags reaches the preset number, user interest tags and corresponding interest weight vectors are obtained based on pre-acquired user behavior data. The interest weight vectors are used to characterize the degree of user interest in the user interest tags. The user interest dispersion score is calculated based on the interest weight vector, and the content recommendation mechanism is determined based on the user interest dispersion score. The step of calculating the user interest dispersion score based on the interest weight vector includes: The interest weight vector is normalized to obtain the first interest weight vector. The dimension values ​​in the normalized interest weight vector that are less than the preset interest weight threshold are set to zero, while the dimension values ​​in the normalized interest weight vector that are greater than or equal to the preset interest weight threshold are maintained to obtain the second interest weight vector. The user interest dispersion score is calculated based on a preset user interest dispersion score calculation model; The user interest dispersion score calculation model is as follows: Where F is the user interest dispersion score, A is the number of non-zero dimensions in the second interest weight vector, and B is the variance of the non-zero data in the second interest weight vector; Content is recommended to users based on the aforementioned content recommendation mechanism.

2. The method according to claim 1, characterized in that, The user behavior data includes explicit user behavior data and implicit user behavior data; wherein the explicit user behavior data includes at least one of historical search data, historical viewing data, historical like data, historical sharing data, and user collection data, and the implicit user behavior data includes exposure without click data.

3. The method according to claim 2, characterized in that, The step of obtaining user interest tags and corresponding interest weight vectors based on pre-acquired user behavior data includes: The user interest tags are determined based on the user behavior data; Calculate the historical interest weight vector of the user interest tag based on the user behavior data and the preset user behavior weights; The historical interest weight vector is reduced in weight using a pre-defined preprocessing model to obtain the interest weight vector.

4. The method according to claim 1, characterized in that, The content recommendation mechanism based on the user interest dispersion score includes: If the user interest dispersion score is greater than the preset interest dispersion confidence value, then the content recommendation mechanism is determined to be a priority content recommendation mechanism. If the user interest dispersion score is less than or equal to the preset interest dispersion confidence value, then the content recommendation mechanism is determined to be a priority interest content recommendation mechanism.

5. An adaptive recommendation device based on user interests, characterized in that, The device includes: The recommendation module is used to recommend content to users based on a priority content recommendation method when the number of tags a user has historically encountered has not reached a preset number. The acquisition module is used to acquire user interest tags and interest weight vectors corresponding to the user interest tags based on pre-acquired user behavior data after the number of tags that the user has historically contacted has reached the preset number. The interest weight vector is used to characterize the degree of interest of the user in the user interest tags. The calculation module is used to calculate the user interest dispersion score based on the interest weight vector, and determine the content recommendation mechanism based on the user interest dispersion score. The calculation module includes a score calculation submodule; the score calculation submodule is specifically used for: The interest weight vector is normalized to obtain the first interest weight vector. The dimension values ​​in the normalized interest weight vector that are less than the preset interest weight threshold are set to zero, while the dimension values ​​in the normalized interest weight vector that are greater than or equal to the preset interest weight threshold are maintained to obtain the second interest weight vector. The user interest dispersion score is calculated based on a preset user interest dispersion score calculation model; The user interest dispersion score calculation model is as follows: Where F is the user interest dispersion score, A is the number of non-zero dimensions in the second interest weight vector, and B is the variance of the non-zero data in the second interest weight vector; The recommendation module is also used to recommend content to users based on the content recommendation mechanism.

6. The apparatus according to claim 5, characterized in that, The user behavior data includes explicit user behavior data and implicit user behavior data; wherein the explicit user behavior data includes at least one of historical search data, historical viewing data, historical like data, historical sharing data, and user collection data, and the implicit user behavior data includes exposure without click data.

7. The apparatus according to claim 6, characterized in that, The acquisition module is specifically used for: The user interest tags are determined based on the user behavior data; Calculate the historical interest weight vector of the user interest tag based on the user behavior data and the preset user behavior weights; The historical interest weight vector is reduced in weight using a pre-defined preprocessing model to obtain the interest weight vector.

8. The apparatus according to claim 5, characterized in that, The calculation module includes a mechanism determination submodule; the mechanism determination submodule is specifically used for: If the user interest dispersion score is greater than the preset interest dispersion confidence value, then the content recommendation mechanism is determined to be a priority content recommendation mechanism. If the user interest dispersion score is less than or equal to the preset interest dispersion confidence value, then the content recommendation mechanism is determined to be a priority interest content recommendation mechanism.

Images