Information recommendation method, information recommendation model training method, device and electronic equipment

Abstract

The present disclosure provides an information recommendation method, an information recommendation model training method, an apparatus, and an electronic device, which relate to the field of artificial intelligence technology, and in particular to the field of information recommendation. A specific implementation scheme is as follows: determining the target user's operation information; determining the target user's clustering probability for each group classification based on the target user's operation information; determining the group classification to which the target user belongs based on the target user's clustering probability for each group classification and the representation vector corresponding to each group classification; and recommending information to the target user based on the group classification to which the target user belongs; wherein the representation vector is used to represent the preference of the group classification, and different group classifications have different corresponding reward values.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, and in particular to an information recommendation method, an information recommendation model training method, a device, and an electronic device in the field of information recommendation. Background Art

[0002] Information recommendation has become an integral part of online services. Specifically, clustering sub-models can be used to group users and recommend information to them based on their group classification. However, in the solutions provided by related technologies, the number of samples used to train information recommendation models for different group classifications can vary significantly, resulting in poor training results. Summary of the Invention

[0003] The present disclosure provides an information recommendation method, an information recommendation model training method, a device, and an electronic device.

[0004] According to a first aspect of the present disclosure, there is provided an information recommendation method, the method comprising:

[0005] Determine the target user's operation information;

[0006] Determining the grouping probability of the target user for each group classification based on the operation information of the target user;

[0007] Determining the group category to which the target user belongs based on the grouping probability of the target user for each group category and the representation vector corresponding to each group category;

[0008] recommending information to the target user based on the group classification to which the target user belongs;

[0009] The representation vector is used to characterize the preference of group classification, and different group classifications correspond to different reward values.

[0010] According to a second aspect of the present disclosure, a method for training an information recommendation model is provided, the method comprising:

[0011] Determining sample operation information of a training sample and determining a cluster label of the training sample;

[0012] A clustering probability submodel included in the recommendation model based on the sample operation information and the sample cluster label training information, wherein the clustering probability submodel is used to determine the clustering probability of each training sample for each group classification included in the clustering probability submodel;

[0013] Based on the clustering probability and the representation vector corresponding to each of the group classifications, training the clustering sub-model included in the information recommendation model, wherein the clustering sub-model is used to determine the group classification to which the target user belongs, and recommend information to the target user based on the group classification;

[0014] The representation vector is used to characterize the preference of group classification, and different group classifications correspond to different reward values.

[0015] According to a third aspect of the present disclosure, an information recommendation model training device is provided, the information recommendation model training device comprising:

[0016] a third determining module, configured to determine sample operation information of a training sample and determine a cluster label of the training sample;

[0017] A first training module is configured to recommend a clustering probability sub-model included in a model based on the sample operation information and the sample cluster label training information, wherein the clustering probability sub-model is configured to determine a clustering probability of each training sample for each group classification included in the clustering probability sub-model;

[0018] A second training module is configured to train a clustering sub-model included in the information recommendation model based on the clustering probability and the representation vector corresponding to each of the group classifications, wherein the clustering sub-model is configured to determine the group classification to which the target user belongs and recommend information to the target user based on the group classification;

[0019] The representation vector is used to characterize the preference of group classification, and different group classifications correspond to different reward values.

[0020] According to a fourth aspect of the present disclosure, an information recommendation device is provided, comprising:

[0021] A first determining module is used to determine the operation information of the target user;

[0022] A second determining module is configured to determine the grouping probability of the target user for each group classification based on the operation information of the target user;

[0023] a processing module, configured to determine the group category to which the target user belongs based on the grouping probability of the target user for each group category and the representation vector corresponding to each group category; and recommend information to the target user based on the group category to which the target user belongs;

[0024] The representation vector is used to characterize the preference of group classification, and different group classifications correspond to different reward values.

[0025] According to a fifth aspect of the present disclosure, there is provided an electronic device, including:

[0026] at least one processor; and a memory communicatively connected to the at least one processor; wherein,

[0027] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the above-mentioned information recommendation model training method or information recommendation method.

[0028] According to a sixth aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to enable the computer to execute the above-mentioned information recommendation model training method or information recommendation method.

[0029] According to a seventh aspect of the present disclosure, a computer program product is provided, comprising a computer program / instruction, which, when executed by a processor, implements the above-mentioned information recommendation model training method or information recommendation method.

[0030] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present disclosure, nor are they intended to limit the scope of the present disclosure. Other features of the present disclosure will become readily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] The accompanying drawings are provided to facilitate a better understanding of the present invention and do not constitute a limitation of the present disclosure.

[0032] Figure 1 This is a schematic diagram of an optional processing flow of the information recommendation method provided by an embodiment of the present disclosure;

[0033] Figure 2 This is a schematic diagram of an optional processing flow of the information recommendation model training method provided by an embodiment of the present disclosure;

[0034] Figure 3 Schematic diagram of an optional processing flow for training a clustering sub-model based on clustering probabilities and representation vectors corresponding to each group classification, provided by an embodiment of the present disclosure;

[0035] Figure 4 This is a schematic diagram of an optional structural component of the information recommendation model training device provided by an embodiment of the present disclosure;

[0036] Figure 5 This is a schematic diagram of an optional structural component of the information recommendation device provided by an embodiment of the present disclosure;

[0037] Figure 6 It is a block diagram of an electronic device used to implement the information recommendation method or information recommendation model training method of the embodiment of the present disclosure. DETAILED DESCRIPTION

[0038] The following description of exemplary embodiments of the present disclosure is made in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding. These details should be considered as merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0039] In the following description, reference is made to “some embodiments”, which describes a subset of all possible embodiments, but it will be understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0040] In the following description, the terms "first\second\third" are merely used to distinguish similar objects and do not represent a specific ordering of the objects. It is understandable that "first\second\third" can be interchanged with a specific order or sequence where permitted, so that the embodiments of the present disclosure described herein can be implemented in an order other than that illustrated or described herein.

[0041] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art in the art of the present disclosure. The terms used herein are only for the purpose of describing the embodiments of the present disclosure and are not intended to limit the present disclosure.

[0042] Before describing the embodiments of the present disclosure in detail, relevant terms involved in the present disclosure are explained.

[0043] Artificial Intelligence (AI) is the theory, methods, techniques, and application systems that use digital computers or computer-controlled systems to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to achieve optimal results. Machine Learning (ML), at the core of AI, specifically studies how computers can simulate or implement human learning behaviors to acquire new knowledge or skills and reorganize existing knowledge structures to continuously improve their performance.

[0044] Clustering sub-model: A model built based on machine learning principles, used to determine the group classification to which a user belongs.

[0045] User: The user in this disclosure may be a real user, or an application account used to represent a real user (such as an account in an instant messaging program, an account in an information application, an account in an email program), etc.

[0046] In the related art, when classifying users into groups, only one factor, user operations (such as users clicking on a certain resource), is usually considered, resulting in inaccurate user group classification. Therefore, users can be classified into groups based on multiple factors of user operations (such as the duration of user clicks on resources, the interactivity of users with other users, the diversity of resources clicked by users, etc.). Specifically, the influence of multiple factors can be learned based on evolutionary strategies (ES) to maximize the reward value corresponding to each group classification. However, there are differences in user groups when training the ES-based clustering model. For example, the number of users in the low-activity user group is large, but the behavior of individual users is sparse; the number of users and the behavior of users in the high-activity user group are relatively sufficient. This will lead to poor training results of the clustering model, insufficient representation ability of the trained clustering model, inaccurate group classification of users based on the clustering model, and inaccurate information recommendation for users based on the group classification to which the user belongs. Based on this, the present disclosure provides a clustering model training method that can improve the training effect of the clustering model, improve the representation ability of the trained clustering model, the accuracy of user group classification, and the accuracy of information recommendation for users.

[0047] The embodiment of the present disclosure also provides an information recommendation method, and an optional processing flow diagram of the user grouping method provided by the embodiment of the present disclosure is as follows: Figure 1 As shown, it can at least include the following steps:

[0048] Step S101: determining the operation information of the target user.

[0049] In some embodiments, the target user's operation information may be information generated based on the target user's operations; the target user's operations may include browsing operations, purchasing operations, liking operations, commenting operations, interactive operations, etc. on a client, application, or mini-program. The target user's operation information may include at least: the category and content of the information browsed by the target user, the duration of browsing, the frequency of browsing, the content purchased, the content liked, the content commented on, and the subject of the comment.

[0050] The target user's operation information may be operation information within a preset period. The preset period can be flexibly set according to the actual application scenario, such as a preset period of one month, three months, six months, or one year. In a specific implementation, the target user's operation information within the preset period can be pre-stored so that the target user's operation information can be quickly obtained or called, and the target users can be quickly classified into groups, thereby improving the efficiency of user group classification.

[0051] Step S102 : determining the grouping probability of the target user for each group classification based on the target user's operation information.

[0052] In some embodiments, the target user's operational information is input into a pre-trained clustering probability sub-model, which determines the target user's clustering probability for each group classification and recommends information to the target user based on the clustering probability. The group classification can be based on any hobby, such as novels, music, games, and movies.

[0053] In some embodiments, the clustering probability sub-model may include multiple sub-models, each sub-model corresponds to a group classification, the target user's operation information is input into each sub-model respectively, and each sub-model outputs the clustering probability that the target user belongs to the group classification corresponding to the sub-model.

[0054] In some embodiments, the group classification includes at least: a first active group and a second active group; wherein the activity level of users in the first active group is greater than the activity level of users in the second active group.

[0055] As an example, if the group classification output by the clustering probability sub-model includes three types, and the clustering probability sub-model corresponds to the first active group, the second active group, and the third active group, then the activity of the users in the first active group is greater than the first active threshold, the activity of the users in the third active group is less than the second active threshold, and the activity of the users in the second active group is less than the first active threshold and greater than the second active threshold; the user's activity can be determined based on user operation information such as the frequency and click duration of the user's click behavior. The above only takes two group classifications and three group classifications as examples. In actual applications, the group classification corresponding to the clustering sub-model can also include more than three group classifications, which is not limited in the embodiments of the present disclosure.

[0056] In some embodiments, the rewards for different group classifications output by the cluster probability sub-model are different. Because users with different levels of activity have significantly different operational behaviors, setting different rewards for different group classifications allows the cluster probability sub-model to be learned based on the characteristics of different users. Each group classification independently evolves its own rewards, which can better learn from its own characteristics.

[0057] In some embodiments, the operational behaviors of users with different levels of activity have significant differences, which can be manifested in that low-activity users have sparse click behaviors and shorter click durations; high-activity users have longer total browsing time, or intensive click behaviors, or longer click durations, or sparse click behaviors but longer browsing time for a single click behavior, etc.

[0058] In some embodiments, a first reward value corresponding to the first active group can be determined based on the operation information of users in the first active group, and the first reward value is used to measure the first evolution parameter corresponding to the first active group; a second reward value corresponding to the second active group can be determined based on the operation information of users in the second active group, and the second reward value is used to measure the second evolution parameter corresponding to the second active group.

[0059] In specific implementation, for the second active group corresponding to low-activity users, the second reward value corresponding to the second active group can be set to the number of consumptions, such as the number of clicks; for the first active group corresponding to high-activity users, the first reward value corresponding to the first active group can be set to the consumption duration, such as the click duration.

[0060] In some embodiments, the evolutionary parameters of the ES model corresponding to each population classification can be adjusted based on the ES algorithm, and the sub-model included in the cluster probability sub-model is the ES model. Specifically, the evolutionary parameters of the ES model corresponding to each population classification can be obtained by repeatedly iteratively adjusting a normal distribution for search, wherein the iterative normal distribution in the evolution strategy is generally written as Contains three parameters m t ,σ t and C t ; The role played by the parameters of the normal distribution is: m t is the mean, which determines the center position of the distribution, m t Determine the search area in the ES algorithm; σ t is the step size parameter, which determines the overall variance of the distribution; σ t Determines the size and intensity of the search range in the ES algorithm; C t is the covariance matrix, which determines the shape of the distribution; C t In the ES algorithm, the dependencies between variables and the relative scales between search directions are determined. t ,σ t and C t Adjusting the three parameters can speed up the convergence rate of the ES algorithm.

[0061] Step S103 : determining the group category to which the target user belongs based on the grouping probability of the target user for each group category and the representation vector corresponding to each group category.

[0062] In some embodiments, each group classification corresponds to a group embedding, which is used to characterize the fusion preference of the group classification. The interests of each group classification can be learned separately; specifically, the group embedding of each group classification can be separately characterized based on the MoE system, and each group classification corresponds to an Expert in the MoE system.

[0063] In some embodiments, a clustering sub-model can be used to multiply the clustering probability of each group classification and the representation vector corresponding to the group classification to obtain the classification value corresponding to each group classification; the classification values ​​of all the group classifications are sorted, and the group classification corresponding to the highest classification value is determined as the group classification of the target user.

[0064] Step S104: recommend information to the target user based on the group classification to which the target user belongs.

[0065] In some embodiments, if the target user belongs to a group classified as a high-activity group, information is recommended to the target user based on the attributes of the high-activity group, so that the target user can spend a longer period of time on the recommended information. For example, if the target user belongs to a high-activity group, information that has been consumed for a longer period of time in the target user's browsing history can be recommended to the target user.

[0066] In some embodiments, if the group classification to which the target user belongs is a low-activity group classification, information is recommended to the target user based on the attributes of the low-activity group classification, so that the target user can consume the recommended information (such as click or browse the recommended information); for example, if the target user belongs to a low-activity group classification, information with a shorter consumption frequency and duration in the historical browsing behavior can be recommended to the target user.

[0067] In the embodiment of the present disclosure, the information recommended to the target user may be video resources, audio resources, web resources, etc.

[0068] In the embodiment of the present disclosure, the group classification to which the user belongs is determined based on the clustering probability of each group classification and the representation vector corresponding to the group classification, and information is recommended to the user based on the group classification to which the user belongs. This can integrate the learning results of each group classification and improve the accuracy of information recommendation.

[0069] Figure 1 The information recommendation method can be implemented based on the information recommendation model; accordingly, the embodiment of the present disclosure also provides an information recommendation model training method, an optional processing flow diagram of the information recommendation model training method provided by the present disclosure, as shown in FIG. Figure 2 As shown, it can at least include the following steps:

[0070] Step S201: Determine sample operation information of a training sample and determine a cluster label of the training sample.

[0071] In some optional embodiments, the training samples may refer to historical users; the sample operations may be browsing operations, purchasing operations, liking operations, comment operations, interactive operations, etc. performed by users on the client, application, or mini-program; accordingly, the sample operation information may at least include: the information classification, content, browsing time, browsing frequency, purchased content, liked content, commented content, commented objects, etc. browsed by historical users.

[0072] In some embodiments, model training is completed based on big data, and the trained model is universal. Therefore, the training model uses multiple training samples; the more training samples used in the training model and the richer the sample operation information content of the training samples, the higher the accuracy of the trained model.

[0073] In some optional embodiments, the training samples may be clustered based on the sample operation information to obtain sample cluster labels for the training samples. For example, the cluster labels for the training samples may be determined based on a clustering algorithm, such as using a Kmeans algorithm to cluster the training samples to obtain sample cluster labels for the training samples.

[0074] Step S202 : training a clustering probability sub-model included in the information recommendation model based on the sample operation information and the sample clustering labels. The clustering probability sub-model is used to determine the clustering probability of each training sample for each group classification included in the information recommendation model.

[0075] In some embodiments, sample operation information is used as input to a clustering probability submodel, which outputs clustering probabilities for each group classification corresponding to the training sample. The clustering probability submodel can be a tree model; in specific implementations, this disclosure does not limit the specific form of the tree model. The clustering probabilities are used to characterize the relationship between the training sample and each group classification; the group classification can be based on any hobby, such as novels, music, games, and movies.

[0076] In some embodiments, the group classification to which the target user belongs output by the clustering probability sub-model is one of at least two group classifications, that is, the clustering sub-model corresponds to at least two group classifications, namely, the first active group and the second active group; wherein, the activity of users in the first active group is greater than the activity of users in the second active group. As an example, if the group classification output by the clustering probability sub-model includes three types, namely, the first active group, the second active group and the third active group, then the activity of users in the first active group is greater than the first active threshold, the activity of users in the third active group is less than the second active threshold, and the activity of users in the second active group is less than the first active threshold and greater than the second active threshold; the activity of users can be determined based on user operation information such as the frequency and click duration of user click behavior. The above only takes two group classifications and three group classifications as examples. In actual applications, the group classification corresponding to the clustering probability sub-model can also include more than three group classifications, which is not limited in the embodiments of the present disclosure.

[0077] In some embodiments, the clustering probability sub-model outputs different rewards for different group classifications. Because users with different levels of activity have significantly different operational behaviors, setting different rewards for different group classifications allows the clustering sub-model to be learned based on the characteristics of different users. Each group classification evolves its own rewards independently, effectively learning from its own characteristics.

[0078] In some embodiments, the operational behaviors of users with different levels of activity have significant differences, which can be manifested in that low-activity users have sparse click behaviors and shorter click durations; high-activity users have longer total browsing time, or intensive click behaviors, or longer click durations, or sparse click behaviors but longer browsing time for a single click behavior, etc.

[0079] In some embodiments, for the second active group corresponding to low-activity users, the second reward value corresponding to the second active group can be set to the number of click behaviors; for the first active group corresponding to high-activity users, the first reward value corresponding to the first active group can be set to the click duration.

[0080] In some embodiments, the evolution parameters of the ES model corresponding to each population classification can be adjusted based on the ES algorithm. Specifically, the evolution parameters of the ES model corresponding to each population classification can be obtained by repeatedly iteratively adjusting a normal distribution for search, where the iterative normal distribution in the evolution strategy is generally written as Contains three parameters m t ,σ t and C t ; The role played by the parameters of the normal distribution is: m t is the mean, which determines the center position of the distribution, mt Determine the search area in the ES algorithm; σ t is the step size parameter, which determines the overall variance of the distribution; σ t Determines the size and intensity of the search range in the ES algorithm; C t is the covariance matrix, which determines the shape of the distribution; C t In the ES algorithm, the dependencies between variables and the relative scales between search directions are determined. t ,σ t and C t Adjusting the three parameters can speed up the convergence rate of the ES algorithm.

[0081] Step S203: Based on the clustering probability and the representation vector corresponding to each group classification, the clustering sub-model included in the training information recommendation model is used to determine the group classification to which the target user belongs, and recommend information to the target user based on the group classification.

[0082] In some embodiments, a schematic diagram of an optional processing flow for training a clustering sub-model based on the clustering probability and the representation vector corresponding to each group classification is shown in FIG. Figure 3 As shown, at least the following steps are included:

[0083] Step S301: input the clustering probability of each group classification and the representation vector corresponding to the group classification into the clustering sub-model.

[0084] In some optional embodiments, one training sample corresponds to multiple group classifications, and the clustering probability of one training sample for each group classification is different.

[0085] In some embodiments, the representation vector (group embedding) corresponding to a group classification is used to characterize the fusion preference of the group classification. The interests of each group classification can be learned separately; specifically, the group embedding of each group classification can be separately characterized based on a Mixture of Experts (MoE) system, where each group classification corresponds to an expert in the MoE system.

[0086] Step S302 : adjusting the parameters of the clustering sub-model based on the group classification output by the clustering sub-model and the actual group classification of the training samples.

[0087] In some embodiments, after the clustering probability of each group classification and the representation vector corresponding to the group classification are input into the clustering sub-model, the clustering sub-model determines the group classification for the training sample; the parameters of the clustering sub-model are adjusted according to the group classification for the training sample output by the clustering sub-model and the actual group classification of the training sample to realize the training of the clustering sub-model; the trained clustering sub-model can determine the group classification to which the target user belongs based on the operation information of the target user, and recommend information corresponding to the group classification to the target user based on the group classification.

[0088] In some embodiments, the clustering probability of each group classification and the representation vector corresponding to the group classification can be multiplied to obtain the classification value corresponding to each group classification; the classification values ​​of all the group classifications are sorted, and the group classification corresponding to the highest classification value is determined to be the group classification determined by the clustering sub-model.

[0089] In the embodiment of the present disclosure, the output of the clustering sub-model is determined based on the clustering probability of each group classification and the representation vector corresponding to the group classification, which can integrate the learning results of each group classification and improve the training accuracy of the clustering sub-model; when using the clustering sub-model to group users, the user's clustering accuracy can be improved; when recommending information to users based on the user's group classification, the accuracy of information recommendation can be improved.

[0090] The present disclosure also provides an information recommendation model training device, the structure diagram of the information recommendation model training device is as follows: Figure 4 Shown, including:

[0091] The third determining module 401 is used to determine the sample operation information of the training sample and determine the cluster label of the training sample;

[0092] A first training module 402 is configured to train a clustering probability submodel included in the information recommendation model based on the sample operation information and the sample clustering labels, wherein the clustering probability submodel is configured to determine the clustering probability of each training sample for each group classification included in the information recommendation model;

[0093] A second training module 403 is configured to train a clustering sub-model included in the information recommendation model based on the clustering probability and the representation vector corresponding to each of the group classifications, wherein the clustering sub-model is configured to determine the group classification to which the target user belongs and recommend information to the target user based on the group classification;

[0094] The representation vector is used to characterize the preference of group classification, and different group classifications correspond to different reward values.

[0095] In some optional embodiments, the second training module 403 is used to input the clustering probability of each of the group classifications and the vector corresponding to the group classification into the clustering sub-model; and adjust the parameters of the clustering sub-model based on the group classification output by the clustering sub-model and the actual group classification of the training sample.

[0096] In some optional embodiments, the different group classifications output by the clustering sub-model include at least:

[0097] the first active group and the second active group;

[0098] The activity level of users in the first active group is greater than the activity level of users in the second active group.

[0099] In some optional embodiments, the second training module 403 is further configured to determine a first reward value corresponding to the first active group based on operation information of users in the first active group, where the first reward value is used to measure a first evolution parameter corresponding to the first active group;

[0100] And / or, a second reward value corresponding to the second active group is determined based on operation information of users in the second active group, where the second reward value is used to measure a second evolution parameter corresponding to the second active group.

[0101] In some optional embodiments, the first reward value is the consumption duration; the second reward value is the number of consumption times.

[0102] In some embodiments, the third determination module 401 is configured to cluster the training samples based on the sample operation information to obtain sample cluster labels of the training samples.

[0103] The embodiment of the present disclosure further provides an information recommendation device, an optional structural diagram of the user grouping device, as shown in FIG. Figure 5 Shown, including:

[0104] A first determining module 501 is used to determine the operation information of the target user;

[0105] A second determining module 502 is configured to determine the grouping probability of the target user for each group classification based on the operation information of the target user;

[0106] The processing module 503 is configured to determine the group category to which the target user belongs based on the grouping probability of the target user for each group category and the representation vector corresponding to each group category; and recommend information to the target user based on the group category to which the target user belongs;

[0107] The representation vector is used to characterize the preference of group classification, and different group classifications correspond to different reward values.

[0108] In some embodiments, the group classification includes at least: a first active group and a second active group; wherein the activity level of users in the first active group is greater than the activity level of users in the second active group.

[0109] In some embodiments, the processing module 503 is further configured to determine a first reward value corresponding to the first active group based on operation information of users in the first active group, where the first reward value is used to measure a first evolution parameter corresponding to the first active group;

[0110] And / or, a second reward value corresponding to the second active group is determined based on operation information of users in the second active group, where the second reward value is used to measure a second evolution parameter corresponding to the second active group.

[0111] In some embodiments, the processing module 503 is configured to perform the following operations for each group classification:

[0112] Multiplying the target user's grouping probability for the group classification by the representation vector corresponding to the group classification to obtain a classification value for the group classification;

[0113] The classification values ​​of all the group classifications are sorted, and the group classification corresponding to the highest classification value is determined as the group classification of the target user.

[0114] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0115] Figure 6 A schematic block diagram of an example electronic device 800 that can be used to implement an embodiment of the present disclosure is shown. The electronic device 800 is used to implement the information recommendation method or information recommendation model training method of an embodiment of the present disclosure. In some optional embodiments, the electronic device 800 can be a terminal device or a server. In some optional embodiments, the electronic device 800 can implement the visual question answering method provided by the embodiment of the present application by running a computer program. For example, the computer program can be a native program or software module in the operating system; it can be a local (Native) application (Application, APP), that is, a program that needs to be installed in the operating system to run; it can also be a small program, that is, a program that can be run only by downloading it to a browser environment; it can also be a small program that can be embedded in any APP. In short, the above-mentioned computer program can be an application, module or plug-in in any form.

[0116] In actual applications, the electronic device 800 can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides 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, CDN, and big data and artificial intelligence platforms. Cloud technology refers to a hosting technology that unifies a series of resources such as hardware, software, and networks within a wide area network or local area network to achieve data computing, storage, processing, and sharing. The electronic device 800 can be a smartphone, tablet computer, laptop computer, desktop computer, smart speaker, smart TV, smart watch, etc., but is not limited to these.

[0117] Electronic device is intended to refer to various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic device may also refer to various forms of mobile devices, such as personal digital assistants, cellular phones, smart phones, wearable devices, vehicle-mounted terminals, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are intended to be examples only and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0118] like Figure 6 As shown, the electronic device 800 includes a computing unit 801, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 802 or a computer program loaded from a storage unit 808 into a random access memory (RAM) 803. In the RAM 803, various programs and data required for the operation of the electronic device 800 can also be stored. The computing unit 801, the ROM 802, and the RAM 803 are connected to each other via a bus 804. An input / output (I / O) interface 805 is also connected to the bus 804.

[0119] Multiple components in the electronic device 800 are connected to the I / O interface 805, including an input unit 806, such as a keyboard, a mouse, etc.; an output unit 807, such as various types of displays, speakers, etc.; a storage unit 808, such as a magnetic disk, an optical disk, etc.; and a communication unit 809, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 809 allows the electronic device 800 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0120] The computing unit 801 can be various general-purpose and / or specialized processing components with processing and computing capabilities. Some examples of the computing unit 801 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units that run machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The computing unit 801 performs the various methods and processes described above, such as the visual question answering method. For example, in some optional embodiments, the visual question answering method can be implemented as a computer software program that is tangibly contained in a machine-readable medium, such as the storage unit 808. In some optional embodiments, part or all of the computer program can be loaded and / or installed on the electronic device 800 via the ROM 802 and / or the communication unit 809. When the computer program is loaded into the RAM 803 and executed by the computing unit 801, one or more steps of the visual question answering method described above can be performed. Alternatively, in other embodiments, the computing unit 801 can be configured as the visual question answering method by any other appropriate means (e.g., by means of firmware).

[0121] Various embodiments of the systems and techniques described herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), system-on-chip systems (SOCs), programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include being implemented in one or more computer programs that are executable and / or interpreted on a programmable system that includes at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0122] The program code used to implement the traffic indicator data processing method of the present disclosure can be written in any combination of one or more programming languages. This program code can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when executed by the processor or controller, the program code implements the functions / operations specified in the flowcharts and / or block diagrams. The program code can be executed entirely on the machine, partially on the machine, as a stand-alone software package, partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0123] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of machine-readable storage media would include an electrical connection based on one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0124] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0125] The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (LAN), a wide area network (WAN), and the Internet.

[0126] A computer system may include a client and a server. The client and server are generally remote from each other and typically interact through a communication network. The client-server relationship arises from computer programs running on the respective computers and having a client-server relationship with each other. The server may be a cloud server, a server in a distributed system, or a server integrated with a blockchain.

[0127] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this disclosure can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved. This is not a limitation herein.

[0128] The above specific embodiments do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure shall be included within the scope of protection of this disclosure.

Claims

1. An information recommendation method, comprising: Determine the target user's operation information; Determining the grouping probability of the target user for each group classification based on the operation information of the target user; The group classification includes at least: a first active group and a second active group; wherein the activity level of users in the first active group is greater than the activity level of users in the second active group; determining a first reward value corresponding to the first active group based on operation information of users in the first active group, where the first reward value is used to measure a first evolution parameter corresponding to the first active group; and / or, determining a second reward value corresponding to the second active group based on operation information of users in the second active group, wherein the second reward value is used to measure a second evolution parameter corresponding to the second active group; Determining the group category to which the target user belongs based on the grouping probability of the target user for each group category and the representation vector corresponding to each group category; recommending information to the target user based on the group classification to which the target user belongs; The representation vector is used to characterize the preference of group classification, and different group classifications correspond to different reward values.

2. The method according to claim 1, wherein The first reward value is the consumption time; The second reward value is the number of consumption times.

3. The method according to claim 1, wherein The determining the group category to which the target user belongs based on the grouping probability of the target user for each group category and the representation vector corresponding to each group category includes: For each population classification, perform the following operations: Multiplying the target user's grouping probability for the group classification by the representation vector corresponding to the group classification to obtain a classification value for the group classification; The classification values ​​of all the group classifications are sorted, and the group classification corresponding to the highest classification value is determined as the group classification of the target user.

4. A method for training an information recommendation model, comprising: Determining sample operation information of a training sample and determining a cluster label of the training sample; Training a clustering probability submodel included in the information recommendation model based on the sample operation information and the sample clustering labels, wherein the clustering probability submodel is used to determine the clustering probability of each training sample for each group classification included in the information recommendation model; Based on the clustering probability and the representation vector corresponding to each of the group classifications, training a clustering sub-model included in the information recommendation model, wherein the clustering sub-model is used to determine the group classification to which the target user belongs, and recommending information to the target user based on the group classification; The representation vector is used to represent the preference of group classification, and different group classifications correspond to different reward values; the different group classifications output by the clustering sub-model include at least: a first active group and a second active group; wherein the activity of users in the first active group is greater than the activity of users in the second active group; determining a first reward value corresponding to the first active group based on operation information of users in the first active group, where the first reward value is used to measure a first evolution parameter corresponding to the first active group; And / or, a second reward value corresponding to the second active group is determined based on operation information of users in the second active group, where the second reward value is used to measure a second evolution parameter corresponding to the second active group.

5. The method according to claim 4, wherein The training of the clustering sub-model based on the clustering probability and the vector representation corresponding to each of the group classifications includes: Inputting the clustering probability of each of the group classifications and the vector corresponding to the group classification into the clustering sub-model; The parameters of the clustering sub-model are adjusted based on the group classification output by the clustering sub-model and the actual group classification of the training samples.

6. The method according to claim 4, wherein: Determining the cluster labels of the training samples includes: The training samples are clustered based on the sample operation information to obtain sample cluster labels of the training samples.

7. An information recommendation device, comprising: A first determining module is used to determine the operation information of the target user; A second determining module is configured to determine the grouping probability of the target user for each group classification based on the operation information of the target user; The group classification includes at least: a first active group and a second active group; wherein the activity level of users in the first active group is greater than the activity level of users in the second active group; determining a first reward value corresponding to the first active group based on operation information of users in the first active group, where the first reward value is used to measure a first evolution parameter corresponding to the first active group; and / or, determining a second reward value corresponding to the second active group based on operation information of users in the second active group, wherein the second reward value is used to measure a second evolution parameter corresponding to the second active group; a processing module, configured to determine the group category to which the target user belongs based on the grouping probability of the target user for each group category and the representation vector corresponding to each group category; and recommend information to the target user based on the group category to which the target user belongs; The representation vector is used to characterize the preference of group classification, and different group classifications correspond to different reward values.

8. An information recommendation model training device, comprising: a third determining module, configured to determine sample operation information of a training sample and determine a cluster label of the training sample; A first training module is configured to train a clustering probability sub-model based on the sample operation information and the sample clustering labels, wherein the clustering probability sub-model is configured to determine a clustering probability of each training sample for each group classification included in the information recommendation; A second training module is configured to train a clustering sub-model included in the information recommendation model based on the clustering probability and the representation vector corresponding to each of the group classifications, wherein the clustering sub-model is configured to determine the group classification to which the target user belongs and recommend information to the target user based on the group classification; The representation vector is used to represent the preference of group classification, and different group classifications output by the information recommendation model correspond to different reward values; the different group classifications output by the clustering sub-model include at least: a first active group and a second active group; wherein the activity of users in the first active group is greater than the activity of users in the second active group; determining a first reward value corresponding to the first active group based on operation information of users in the first active group, where the first reward value is used to measure a first evolution parameter corresponding to the first active group; And / or, a second reward value corresponding to the second active group is determined based on operation information of users in the second active group, where the second reward value is used to measure a second evolution parameter corresponding to the second active group.

9. An electronic device comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1 to 3; or to enable the at least one processor to perform the method of any one of claims 4 to 6.

10. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to cause the computer to execute the method according to any one of claims 1 to 3 , or the computer instructions are used to cause the computer to execute the method according to any one of claims 4 to 6 .

11. A computer program product, comprising a computer program / instruction, wherein when the computer program / instruction is executed by a processor, the method according to any one of claims 1 to 3 is implemented; or when the computer program / instruction is executed by a processor, the method according to any one of claims 4 to 6 is implemented.

Images