Model training method, information generation method, device, equipment and medium

Abstract

Embodiments of the present disclosure disclose a model training method, an information generation method, an apparatus, a device and a medium. A specific implementation of the method comprises: obtaining a set of item value presentation information and a set of associated item information corresponding to each item value presentation information; for the set of item value presentation information, performing the following training steps: inputting target item value presentation information in the set of item value presentation information and the corresponding set of associated item information into an initial multi-view click information generation model to generate presentation click information and a set of associated item click information; determining whether the initial multi-view click information generation model is trained to completion according to the presentation click information and the set of associated item click information; and in response to determining that the training is completed, determining the initial multi-view click information generation model as a multi-view click information generation model. This implementation is related to artificial intelligence, and by using the multi-view click information generation model, the presentation click information for the target item value presentation information can be accurately and efficiently determined.

Description

Technical Field

[0001] The embodiments disclosed herein relate to the field of computer technology, specifically to model training methods, information generation methods, apparatus, devices, and media. Background Technology

[0002] Currently, various industries often use information presentation (e.g., advertisements) to showcase the advantages of related products in order to increase product awareness. Popularity is typically represented by the click-through rate (CTR) of the presented information.

[0003] The typical approach to generating predicted click information corresponding to presented information is as follows: First, the presented information set is used as the training dataset, and the corresponding presented click information set is used as the label set to train an initial presented click information generation model. Then, the presented click information generation model is used to generate target presented click information corresponding to the target presented information determined by the information to be clicked.

[0004] However, the inventors discovered that the following technical problems often arise when using the above method:

[0005] The lack of application of product number information that represents user preferences in similar scenarios leads to the inability to effectively guarantee the accuracy of the trained presentation click information generation model, resulting in insufficient accuracy of the generated presentation click information.

[0006] The information disclosed in this background section is only intended to enhance the understanding of the background of the inventive concept, and therefore may contain information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0007] The summary portion of this disclosure is intended to provide a brief overview of the concepts, which will be described in detail in the detailed description portion. This summary portion is not intended to identify key or essential features of the claimed technical solutions, nor is it intended to limit the scope of the claimed technical solutions.

[0008] Some embodiments of this disclosure provide model training methods, information generation methods, apparatuses, devices, and media to address the technical problems mentioned in the background section above.

[0009] In a first aspect, some embodiments of this disclosure provide a model training method, including: acquiring an item value presentation information set and an associated item information set corresponding to each item value presentation information set; for the item value presentation information set, performing the following training steps: inputting the target item value presentation information and the corresponding associated item information set into an initial multi-view click information generation model to generate presentation click information and associated item click information sets, wherein the initial multi-view click information generation model is a model generated based on the initial presentation click information generation model and the initial associated item click information generation model; determining whether the initial multi-view click information generation model has finished training based on the presentation click information and the associated item click information sets; in response to determining that training has finished, determining the initial multi-view click information generation model as a multi-view click information generation model.

[0010] Optionally, the above method further includes: in response to determining that the training has not ended, updating the initial multi-view click information generation model based on the presentation click information and the associated item click information set to generate an updated model, and removing the target item value presentation information from the item value presentation information set to obtain the removed information set; using the updated model as the initial multi-view click information generation model and the removed information set as the item value presentation information set, and continuing to perform the above training steps.

[0011] Optionally, the aforementioned initial presentation click information generation model includes: at least one serially connected hidden layer and a presentation click information output layer; the aforementioned initial associated item click information generation model includes: at least one serially connected feature extraction layer and an associated item click information output layer; and the aforementioned inputting the target item value presentation information and the corresponding associated item information set from the item value presentation information set into the initial multi-view click information generation model to generate presentation click information and associated item click information sets includes: inputting the aforementioned target item value presentation information into a first hidden layer at a first position in the hidden layer sequence to obtain the hidden layer feature information corresponding to the aforementioned first hidden layer; inputting the aforementioned target item value presentation information and the corresponding associated item information set into a feature extraction layer corresponding to the aforementioned first hidden layer in at least one feature extraction layer to obtain first feature extraction information and a second feature extraction information set; determining the aforementioned first hidden layer as the target hidden layer; and responding to determining that the target hidden layer is not the hidden layer at the second position in the hidden layer sequence. For the hidden layer sequence, the following generation steps are performed: First feature extraction information corresponding to the target hidden layer and hidden layer feature information corresponding to the target hidden layer are fused to generate fused feature information; the fused feature information is input to the next hidden layer in the hidden layer sequence, which is the target hidden layer feature information; the first and second feature extraction information sets corresponding to the target hidden layer are input to the feature extraction layer corresponding to the next hidden layer to output the third and fourth feature extraction information sets; the third feature extraction information and the target hidden layer feature information are fused to obtain fused feature information, which is the target fused feature information; in response to determining that the next hidden layer is the hidden layer at the second position in the hidden layer sequence, the target fused feature information is input to the presentation click information output layer to generate presentation click information, and the third and fourth feature extraction information sets are input to the associated item click information output layer to generate associated item click information sets.

[0012] Optionally, the above method further includes: in response to determining that the next hidden layer is not the hidden layer at the second position in the hidden layer sequence, taking the next hidden layer as the target hidden layer, taking the target fused feature information as the hidden layer feature information corresponding to the target hidden layer, taking the third feature extraction information as the first feature extraction information corresponding to the target hidden layer, taking the fourth feature extraction information set as the second feature extraction information set corresponding to the target hidden layer, and continuing to execute the above generation steps.

[0013] Optionally, the above method further includes: in response to determining that the target hidden layer is the hidden layer at the second position in the hidden layer sequence, inputting the hidden layer feature information corresponding to the first hidden layer to the presentation click information output layer to obtain presentation click information, and inputting the first feature extraction information and the second feature extraction information set to the associated item click information output layer to generate the associated item click information set.

[0014] Optionally, the aforementioned set of item value presentation information and the associated item information set corresponding to each set of item value presentation information are stored through the following steps: For each item value presentation information in the aforementioned set of item value presentation information, the following storage steps are performed: the associated item information set corresponding to the aforementioned item value presentation information is downsampled to obtain a downsampled item information set; common feature information between the aforementioned downsampled item information set and the item value presentation information is determined; the aforementioned common feature information is removed from the aforementioned item value presentation information to obtain the removed item value presentation information; the aforementioned common feature information is removed from each downsampled item information in the downsampled item information set to generate the removed item information set; the removed item information set, the removed item value presentation information, and the aforementioned common feature information are stored accordingly.

[0015] Secondly, some embodiments of this disclosure provide a model training apparatus, including: a first acquisition unit configured to acquire an item value presentation information set and an associated item information set corresponding to each item value presentation information set; and a training unit configured to perform the following training steps for the item value presentation information set: inputting the target item value presentation information and the corresponding associated item information set into an initial multi-view click information generation model to generate presentation click information and associated item click information sets, wherein the initial multi-view click information generation model is a model generated based on the initial presentation click information generation model and the initial associated item click information generation model; determining whether the initial multi-view click information generation model has finished training based on the presentation click information and the associated item click information sets; and in response to determining that training has finished, determining the initial multi-view click information generation model as a multi-view click information generation model.

[0016] Optionally, the above apparatus further includes: in response to determining that the training has not ended, updating the initial multi-view click information generation model based on the presentation click information and the associated item click information set to generate an updated model, and removing the target item value presentation information from the item value presentation information set to obtain a removed information set; using the updated model as the initial multi-view click information generation model and the removed information set as the item value presentation information set, and continuing to execute the above training steps.

[0017] Optionally, the aforementioned initial presentation click information generation model includes: at least one serially connected hidden layer and a presentation click information output layer; the aforementioned initial associated item click information generation model includes: at least one serially connected feature extraction layer and an associated item click information output layer; and the training unit can be configured to: input the aforementioned target item value presentation information into the first hidden layer at the first position in the hidden layer sequence to obtain the hidden layer feature information corresponding to the aforementioned first hidden layer; input the aforementioned target item value presentation information and the corresponding associated item information set into the feature extraction layer corresponding to the aforementioned first hidden layer in at least one feature extraction layer to obtain the first feature extraction information and the second feature extraction information set; determine the aforementioned first hidden layer as the target hidden layer; in response to determining that the target hidden layer is not the hidden layer at the second position in the hidden layer sequence, for the hidden layer sequence, perform the following generation steps: input the first feature extraction information corresponding to the target hidden layer into the hidden layer at the first position in the hidden layer sequence to obtain the first feature extraction information and the second feature extraction information set; determine the aforementioned first hidden layer as the target hidden layer; and in response to determining that the target hidden layer is not the hidden layer at the second position in the hidden layer sequence, perform the following generation steps for the hidden layer sequence: input the first feature extraction information of the target hidden layer into the hidden layer at the first position in the hidden layer sequence to obtain the first feature extraction information and the second feature extraction information set; determine the target ... and in response to determining that the target hidden layer is The system extracts information and performs feature fusion with the hidden layer feature information corresponding to the target hidden layer to generate fused feature information. The fused feature information is then input into the next hidden layer in the hidden layer sequence, within the target hidden layer, to obtain hidden layer feature information, which serves as the target hidden layer feature information. The system also inputs the first and second feature extraction information sets corresponding to the target hidden layer into the feature extraction layer corresponding to the next hidden layer to output the third and fourth feature extraction information sets. The third feature extraction information and the target hidden layer feature information are then fused to obtain fused feature information, which serves as the target fused feature information. In response to determining that the next hidden layer is the second hidden layer in the hidden layer sequence, the target fused feature information is input into the presentation click information output layer to generate presentation click information, and the third and fourth feature extraction information sets are input into the associated item click information output layer to generate associated item click information sets.

[0018] Optionally, the training unit can be configured to: in response to determining that the next hidden layer is not the hidden layer at the second position in the hidden layer sequence, take the next hidden layer as the target hidden layer, take the target fused feature information as the hidden layer feature information corresponding to the target hidden layer, take the third feature extraction information as the first feature extraction information corresponding to the target hidden layer, take the fourth feature extraction information set as the second feature extraction information set corresponding to the target hidden layer, and continue to perform the above generation steps.

[0019] Optionally, the training unit can be configured to: in response to determining that the target hidden layer is the hidden layer at the second position in the hidden layer sequence, input the hidden layer feature information corresponding to the first hidden layer to the presentation click information output layer to obtain presentation click information, and input the first feature extraction information and the second feature extraction information set to the associated item click information output layer to generate the associated item click information set.

[0020] Optionally, the aforementioned set of item value presentation information and the associated item information set corresponding to each set of item value presentation information are stored through the following steps: For each item value presentation information in the aforementioned set of item value presentation information, the following storage steps are performed: the associated item information set corresponding to the aforementioned item value presentation information is downsampled to obtain a downsampled item information set; common feature information between the aforementioned downsampled item information set and the item value presentation information is determined; the aforementioned common feature information is removed from the aforementioned item value presentation information to obtain the removed item value presentation information; the aforementioned common feature information is removed from each downsampled item information in the downsampled item information set to generate the removed item information set; the removed item information set, the removed item value presentation information, and the aforementioned common feature information are stored accordingly.

[0021] Thirdly, some embodiments of this disclosure provide an information generation method, including: acquiring target item value presentation information; generating a presentation click information generation model based on a multi-view click information generation model, wherein the multi-view click information generation model is generated based on the model training method associated with the first aspect; and inputting the target item value presentation information into the presentation click information generation model to obtain presentation click information.

[0022] Fourthly, some embodiments of this disclosure provide an information generation apparatus, including: a second acquisition unit configured to acquire target item value presentation information; a generation unit configured to generate a presentation click information generation model based on a multi-view click information generation model, wherein the multi-view click information generation model is generated based on the model training method associated with the first aspect; and an input unit configured to input the target item value presentation information into the presentation click information generation model to obtain presentation click information.

[0023] Fifthly, some embodiments of this disclosure provide an electronic device, including: one or more processors; and a storage device having one or more programs stored thereon, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any of the implementations of the first and third aspects.

[0024] Sixthly, some embodiments of this disclosure provide a computer-readable medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the method as described in any of the implementations of the first and third aspects.

[0025] In a seventh aspect, some embodiments of this disclosure provide a computer program product, including a computer program that, when executed by a processor, implements the methods described in any of the implementations of the first and third aspects above.

[0026] The above embodiments of this disclosure have the following beneficial effects: By using the model training method of some embodiments of this disclosure, and utilizing the multi-view click information generation model, the presentation click information for the value presentation information of the target item can be accurately and efficiently determined. Specifically, the reason for the inaccuracy of the related presentation click information is that there is a lack of application of item number information representing user preferences in similar scenarios, which leads to the inability to effectively guarantee the accuracy of the trained presentation click information generation model, resulting in inaccurate generated presentation click information. Based on this, the model training method of some embodiments of this disclosure first obtains the item value presentation information set and the associated item information set corresponding to each item value presentation information, as the model training dataset for the subsequent initial multi-view click information generation model. Here, for the training of the initial multi-view click information generation model, not only is the item value presentation information used as input, but also the associated item information set that is related to the item value presentation information is considered. The associated item information set here has the same feature information as the item value presentation information, that is, the feature information that the user prefers. Therefore, by using the associated item information set corresponding to the item value presentation information as input, the initial multi-view click information generation model can learn more user-preferred feature information, making the subsequent multi-view click information generation model more accurate. Then, for the item value presentation information set, the following training steps are performed: First, the target item value presentation information and the corresponding associated item information set are input into the initial multi-view click information generation model to generate presentation click information and associated item click information sets. The initial multi-view click information generation model is a model generated based on the initial presentation click information generation model and the initial associated item click information generation model. In this way, the initial multi-view click information generation model can learn multi-faceted features that users prefer from multiple perspectives, ensuring not only accurate determination of click information for item value presentation information but also accurate determination of click information corresponding to associated item information. Here, the generated presentation click information considers not only the feature information of item value presentation information but also the feature information of associated item information, resulting in more accurate presentation click information. Furthermore, the generated associated item click information considers not only the feature information of item value presentation information but also the feature information of associated item information, resulting in more accurate associated item click information. The second step is to determine whether the initial multi-view click information generation model has finished training, based on the presented click information and the associated item click information set, thus confirming whether the initial multi-view click information generation model has been successfully trained. The third step, in response to confirming the completion of training, is to officially recognize the initial multi-view click information generation model as a multi-view click information generation model.In summary, during the training process of the initial multi-view click information generation model, it learns not only the feature information of the item's value presentation but also the feature information of the corresponding related item information. This allows the initial multi-view click information generation model to learn feature information that is more relevant to user preferences. Furthermore, since the initial multi-view click information generation model is based on the initial presentation click information generation model and the initial related item click information generation model, it can generate presentation click information and related item click information sets from multiple perspectives. Therefore, the trained multi-view click information generation model can accurately and efficiently determine the presentation click information for the target item's value presentation information. Attached Figure Description

[0027] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and elements are not necessarily drawn to scale.

[0028] Figure 1 This is a schematic diagram illustrating an application scenario of a model training method according to some embodiments of the present disclosure;

[0029] Figure 2 This is a flowchart of some embodiments of the model training method according to this disclosure;

[0030] Figure 3 These are flowcharts of other embodiments of the model training method according to this disclosure;

[0031] Figure 4 This is a flowchart of some embodiments of the information generation method according to this disclosure;

[0032] Figure 5 These are schematic diagrams illustrating the structure of some embodiments of the model training apparatus according to this disclosure;

[0033] Figure 6 These are schematic diagrams illustrating the structure of some embodiments of the information generation apparatus according to this disclosure;

[0034] Figure 7 This is a schematic diagram of the structure of an electronic device suitable for implementing some embodiments of the present disclosure. Detailed Implementation

[0035] Embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of this disclosure are shown in the drawings, it should be understood that this disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of this disclosure. It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure.

[0036] It should also be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings. Unless otherwise specified, the embodiments and features described in this disclosure can be combined with each other.

[0037] It should be noted that the concepts of "first" and "second" mentioned in this disclosure are used only to distinguish different devices, modules or units, and are not used to limit the order of functions performed by these devices, modules or units or their interdependencies.

[0038] It should be noted that the terms "a" and "a plurality of" used in this disclosure are illustrative rather than restrictive, and those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0039] The names of messages or information exchanged between multiple devices in the embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of such messages or information.

[0040] Before performing any of the operations involving the collection, storage, and use of user personal information (such as information on the value of goods and information sets of related goods) disclosed in this disclosure, the relevant organizations or individuals shall fulfill their obligations, including conducting personal information security impact assessments, informing personal information subjects, and obtaining prior authorization and consent from personal information subjects.

[0041] This disclosure will now be described in detail with reference to the accompanying drawings and embodiments.

[0042] Figure 1 This is a schematic diagram illustrating an application scenario of a model training method according to some embodiments of the present disclosure.

[0043] exist Figure 1In this application scenario, firstly, the electronic device 101 can acquire the item value presentation information set 102 and the associated item information set corresponding to each item value presentation information. Then, for the item value presentation information set, the electronic device 101 can perform the following training steps: First, input the target item value presentation information 103 and the corresponding associated item information set 104 from the item value presentation information set 102 into the initial multi-view click information generation model 105 to generate presentation click information 108 and associated item click information set 109. The initial multi-view click information generation model 105 is a model generated based on the initial presentation click information generation model 106 and the initial associated item click information generation model 107. In this application scenario, the target item value presentation information 103 could be "Never let the wind dry your skin's moisture." The associated item information set 104 can include: associated item information 1041, associated item information 1042, associated item information 1043, and associated item information 1044. Associated item information 1041 could be "Brand A, Model A toner." Associated item information 1042 could be "Brand B moisturizing toner". Associated item information 1043 could be "Brand A, Model B toner". Associated item information 1044 could be "Brand B lotion". Presented click information 108 could be "10 times". Associated item click information set 109 includes: associated item click information set 1091 corresponding to associated item information 1041, associated item click information set 1092 corresponding to associated item information 1042, associated item click information set 1093 corresponding to associated item information 1043, and associated item click information set 1094 corresponding to associated item information 1044. Associated item click information set 1091 could be "12 times". Associated item click information set 1092 could be "22 times". Associated item click information set 1093 could be "32 times". Associated item click information set 1094 could be "21 times". The second step is to determine whether the initial multi-view click information generation model 105 has finished training, based on presented click information 108 and associated item click information set 109. Third, in response to the completion of training, the initial multi-view click information generation model 105 is designated as the multi-view click information generation model 110. In this application scenario, the initial presentation click information generation model 106 is designated as the presentation click information generation model 111. The initial associated item click information generation model 107 is designated as the associated item click information generation model 112.

[0044] It should be noted that the aforementioned electronic device 101 can be either hardware or software. When the electronic device is hardware, it can be implemented as a distributed cluster consisting of multiple servers or terminal devices, or as a single server or a single terminal device. When the electronic device is software, it can be installed in the hardware devices listed above. It can be implemented as, for example, multiple software programs or software modules used to provide distributed services, or as a single software program or software module. No specific limitations are made here.

[0045] It should be understood that Figure 1 The number of electronic devices shown is merely illustrative. Any number of electronic devices can be used depending on the implementation requirements.

[0046] Continue to refer to Figure 2 The diagram illustrates a flow 200 of some embodiments of a model training method according to the present disclosure. This model training method includes the following steps:

[0047] Step 201: Obtain the set of item value presentation information and the set of associated item information corresponding to each item value presentation information.

[0048] In some embodiments, the execution entity of the above model training method (e.g. Figure 1 The electronic device 101 shown can acquire a set of item value presentation information and a set of associated item information corresponding to each item value presentation information via a wired or wireless connection. The item value presentation information can be information presenting the value of a target item and displaying item information. Specifically, the item value presentation information includes: value presentation information, clicker information, and item information. Value presentation information can be information presenting the value of a target item. Value presentation information can be in various forms. For example, value presentation information can be in image form or text form. Clicker information can be the user information of the user who clicks on the value presentation information. Item information can be the item information of the target item whose value is presented. In practice, the target item can be a target product. Value presentation information can be a product advertisement for the target product. Clicker information can be the user information of the user who clicks on the target product advertisement. The set of associated item information corresponding to the item value presentation information can be a set of associated item information that shares the same clicker characteristics as the item value presentation information. That is, the clicker information corresponding to the item value presentation information is the same as the user information corresponding to the set of associated item information. In practice, the set of associated item information can be a set of item information displayed for users' daily searches. For example, when a user enters a target word in the search bar of a target page, the target page displays a set of recommended items. The set of item information corresponding to the recommended item set is then identified as the user's associated item information set.

[0049] It should be noted that the item value presentation information also includes the corresponding click information tags. In practice, this applies to product advertisements where the value presentation information targets the product. The click information tag can be the actual click-through rate (CTR) of the product advertisement. Related item information also includes related item click information tags. These related item click information tags can characterize user clicks on related items.

[0050] In some optional implementations of certain embodiments, the above-mentioned item value presentation information set and the associated item information set corresponding to each item value presentation information set are stored through the following steps:

[0051] For each item value presentation information in the above item value presentation information set, perform the following storage steps:

[0052] Sub-step 1: The aforementioned executing entity can perform downsampling processing on the associated item information set corresponding to the aforementioned item value presentation information to obtain a downsampled item information set.

[0053] As an example, the aforementioned executing entity can perform random downsampling on the associated item information set corresponding to the aforementioned item value presentation information to obtain a downsampled item information set.

[0054] Here, the set of related item information is generally large in number. By downsampling, the information storage pressure is reduced, which alleviates the pressure on subsequent model training time.

[0055] Sub-step 2: The aforementioned executing entity can determine the common feature information between the downsampled item information set and the item value presentation information. This common feature information can be information where the corresponding feature attribute information in the downsampled item information set is identical to the item feature information corresponding to the item value presentation information. This common feature information includes: user feature information and request feature information. Request feature information can indicate that the request corresponding to the downsampled item information set and the request corresponding to the item value presentation information are the same request. In practice, the request corresponding to the downsampled item information set can be an item search request for the clicking user. Similarly, the request corresponding to the item value presentation information is also an item search request for the clicking user.

[0056] Sub-step 3: The aforementioned executing entity can remove the aforementioned common feature information from the aforementioned item value presentation information to obtain the item value presentation information after removal.

[0057] Sub-step 4: The aforementioned executing entity can remove the common feature information from each downsampled item information in the downsampled item information set to generate the removed item information, thus obtaining the removed item information set.

[0058] Sub-step 5: The aforementioned executing entity can store the aforementioned set of removed item information, the aforementioned value presentation information of removed items, and the aforementioned common feature information accordingly.

[0059] As an example, the aforementioned executing entity can store the removed item information set, the removed item value presentation information, and the aforementioned common feature information in the form of a triple for corresponding storage.

[0060] Here, by storing the removed item information set, the value presentation information of the removed items, and the aforementioned common feature information in a corresponding manner, the storage pressure on the corresponding feature information of the associated item information set can be reduced.

[0061] Step 202: For the set of information presenting the value of items, perform the following training steps:

[0062] Step 2021: Input the target item value presentation information and the corresponding associated item information set from the item value presentation information set into the initial multi-view click information generation model to generate presentation click information and associated item click information set.

[0063] In some embodiments, the aforementioned executing entity can input the target item value presentation information and the corresponding associated item information set from the item value presentation information set into an initial multi-view click information generation model to generate presentation click information and associated item click information sets. The initial multi-view click information generation model is a model generated based on the initial presentation click information generation model and the initial associated item click information generation model. The initial multi-view click information generation model is a multi-view click information generation model that has not yet finished training. The multi-view click information generation model can be a model that generates click information through multiple perspectives. Specifically, multiple perspectives can include: feature perspectives corresponding to the item value presentation information and feature perspectives corresponding to the associated item information. The target item value presentation information can be item value presentation information randomly selected from the item value presentation information set. The presentation click information can be the predicted click information for the target item value presentation information. In practice, the target item value presentation information is the advertising information for the product advertisement corresponding to the target product, and the presentation click information can be the predicted click-through rate for the advertisement corresponding to the target product. There is a one-to-one correspondence between the associated item click information in the associated item click information set and the associated item information in the associated item information set corresponding to the target item value presentation information. Related item click information can represent the click information of users on related items corresponding to the value presentation information of the target item. In practice, the value presentation information of the target item refers to the advertising information of the product advertisement corresponding to the target product, and the related item click information can be the click-through rate of users who clicked on the advertising information on the related items. The initial presentation click information generation model can be a presentation click information generation model that has not yet finished training. The presentation click information generation model can be a neural network model that generates presentation click information. In practice, the presentation click information generation model can be a multi-layered, sequentially connected convolutional neural network model. The initial related item click information generation model can be a related item click information generation model that has not yet finished training. The related item click information generation model can be a neural network model that generates related item click information. In practice, the related item click information generation model can be a multi-layered, sequentially connected convolutional neural network.

[0064] As an example, the initial multi-view click information generation model is generated through the following steps:

[0065] The first target number layer of the multi-layer serially connected convolutional neural network corresponding to the initial presentation click information generation model and the first target number layer of the multi-layer serially connected convolutional neural network corresponding to the initial associated item click information generation model are set as common networks. The remaining layers of the multi-layer serially connected convolutional neural network corresponding to the initial presentation click information generation model are used as the output layers of the initial presentation click information generation model, and the remaining layers of the multi-layer serially connected convolutional neural network corresponding to the initial associated item click information generation model are used as the output layers of the initial associated item click information generation model, so as to generate the initial multi-view click information generation model.

[0066] Step 2022: Based on the presented click information and the set of click information for related items, determine whether the initial multi-view click information generation model has finished training.

[0067] In some embodiments, the aforementioned execution entity can determine whether the initial multi-view click information generation model has finished training based on the presented click information and the associated item click information set.

[0068] As an example, firstly, the aforementioned executing entity can determine the information loss between the presented click information tags and the presented click information, as the first loss information. Then, it determines the information loss between the set of related item click information and the set of related item click information tags, as the second loss information, thus obtaining the second loss information set. Next, the first and second loss information sets are weighted and summed to obtain the weighted summation loss information. Finally, in response to determining that the weighted summation loss information is less than a target value, the initial multi-view click information generation model training is considered complete. In response to determining that the weighted summation loss information is greater than or equal to the target value, the initial multi-view click information generation model training is considered incomplete.

[0069] Step 2023: In response to determining the end of training, the initial multi-view click information generation model is determined as the multi-view click information generation model.

[0070] In some embodiments, in response to determining the end of training, the aforementioned executing entity may determine the initial multi-view click information generation model as the multi-view click information generation model. Here, the multi-view click information generation model is a trained model.

[0071] In some optional implementations of certain embodiments, after step 202, the steps further include:

[0072] The first step, in response to the determination that the training has not ended, is to update the initial multi-view click information generation model based on the presentation click information and the related item click information set, so as to generate the updated model, and to remove the target item value presentation information from the item value presentation information set to obtain the removed information set.

[0073] As an example, in response to the determination that training is not yet complete, firstly, a weighted summation loss is generated based on the presentation click information and the associated item click information set. Then, based on the weighted summation loss, the initial multi-view click information generation model is updated using backpropagation to generate an updated model.

[0074] The second step is to use the updated model as the initial multi-view click information generation model, remove the information set as the item value presentation information set, and continue to perform the above training steps.

[0075] The above embodiments of this disclosure have the following beneficial effects: By using the model training method of some embodiments of this disclosure, and utilizing the multi-view click information generation model, the presentation click information for the value presentation information of the target item can be accurately and efficiently determined. Specifically, the reason for the inaccuracy of the related presentation click information is that there is a lack of application of item number information representing user preferences in similar scenarios, which leads to the inability to effectively guarantee the accuracy of the trained presentation click information generation model, resulting in inaccurate generated presentation click information. Based on this, the model training method of some embodiments of this disclosure first obtains the item value presentation information set and the associated item information set corresponding to each item value presentation information, as the model training dataset for the subsequent initial multi-view click information generation model. Here, for the training of the initial multi-view click information generation model, not only is the item value presentation information used as input, but also the associated item information set that is related to the item value presentation information is considered. The associated item information set here has the same feature information as the item value presentation information, that is, the feature information that the user prefers. Therefore, by using the associated item information set corresponding to the item value presentation information as input, the initial multi-view click information generation model can learn more user-preferred feature information, making the subsequent multi-view click information generation model more accurate. Then, for the item value presentation information set, the following training steps are performed: First, the target item value presentation information and the corresponding associated item information set are input into the initial multi-view click information generation model to generate presentation click information and associated item click information sets. The initial multi-view click information generation model is a model generated based on the initial presentation click information generation model and the initial associated item click information generation model. In this way, the initial multi-view click information generation model can learn multi-faceted features that users prefer from multiple perspectives, ensuring not only accurate determination of click information for item value presentation information but also accurate determination of click information corresponding to associated item information. Here, the generated presentation click information considers not only the feature information of item value presentation information but also the feature information of associated item information, resulting in more accurate presentation click information. Furthermore, the generated associated item click information considers not only the feature information of item value presentation information but also the feature information of associated item information, resulting in more accurate associated item click information. The second step is to determine whether the initial multi-view click information generation model has finished training, based on the presented click information and the associated item click information set, thus confirming whether the initial multi-view click information generation model has been successfully trained. The third step, in response to confirming the completion of training, is to officially recognize the initial multi-view click information generation model as a multi-view click information generation model.In summary, during the training process of the initial multi-view click information generation model, it learns not only the feature information of the item's value presentation but also the feature information of the corresponding related item information. This allows the initial multi-view click information generation model to learn feature information that is more relevant to user preferences. Furthermore, since the initial multi-view click information generation model is based on the initial presentation click information generation model and the initial related item click information generation model, it can generate presentation click information and related item click information sets from multiple perspectives. Therefore, the trained multi-view click information generation model can accurately and efficiently determine the presentation click information for the target item's value presentation information.

[0076] Further reference Figure 3 The diagram illustrates flow 300 of some other embodiments of the model training method according to this disclosure. This model training method includes the following steps:

[0077] Step 301: Obtain the set of item value presentation information and the set of associated item information corresponding to each item value presentation information.

[0078] Step 302: For the set of information presenting the value of items, perform the following training steps:

[0079] Step 3021: Input the target item value presentation information into the first hidden layer at the first position in the hidden layer sequence to obtain the hidden layer feature information corresponding to the first hidden layer.

[0080] In some embodiments, the execution entity (e.g. Figure 1The electronic device 101 shown can input the target item value presentation information into the first hidden layer at the first position in the hidden layer sequence to obtain the hidden layer feature information corresponding to the first hidden layer. The initial presentation click information generation model includes: at least one serially connected hidden layer and a presentation click information output layer. The initial associated item click information generation model includes: at least one serially connected feature extraction layer and an associated item click information output layer. In practice, the hidden layer can be an MLP (Multilayer Perceptron). The first hidden layer at the first position can be the hidden layer at the first position in the serial sequence of hidden layers. The hidden layer feature information can be the output feature information of the first hidden layer. The multilayer perceptron includes: a fully connected layer (FC) and a non-linear activation function. The target item value presentation information can be the item value presentation information randomly selected from the set of item value presentation information. The presentation click information output layer can be the output layer that outputs the presentation click information. In practice, the presentation click information output layer can be a fully connected layer. The feature extraction layer can be a multilayer perceptron. The associated item click information output layer can be the output layer that outputs the associated item click information. In practice, the output layer for related item click information can be a fully connected layer.

[0081] It should be noted that there is a network correspondence between the hidden layers in at least one serially connected hidden layer and the feature extraction layers in at least one serially connected feature extraction layer. Therefore, the number of network layers corresponding to at least one serially connected hidden layer is the same as the number of network layers corresponding to at least one serially connected feature extraction layer.

[0082] Step 3022: Input the target item value presentation information and the corresponding associated item information set into the feature extraction layer corresponding to the first hidden layer in at least one feature extraction layer to obtain the first feature extraction information and the second feature extraction information set.

[0083] In some embodiments, the executing entity can input the target item value presentation information and the corresponding associated item information set into the feature extraction layer corresponding to the first hidden layer in at least one feature extraction layer to obtain first feature extraction information and a second feature extraction information set. The corresponding associated item information set can be the associated item information set corresponding to the target item value presentation information. The first feature extraction information can characterize the semantic content of the value presentation information corresponding to the target item value presentation information. The second feature extraction information set has a one-to-one correspondence with the corresponding associated item information set. The second feature extraction information can characterize the semantic content of the item features corresponding to the associated item information. In practice, both the first and second feature extraction information can be in vector form.

[0084] Step 3023: Determine the first hidden layer as the target hidden layer.

[0085] In some embodiments, the execution entity may determine the first hidden layer as the target hidden layer.

[0086] Step 3024: In response to determining that the target hidden layer is not the second hidden layer in the hidden layer sequence, the following generation steps are performed for the hidden layer sequence:

[0087] Step 30241: The first feature extraction information corresponding to the target hidden layer and the hidden layer feature information corresponding to the target hidden layer are fused to generate fused feature information.

[0088] In some embodiments, the execution entity may fuse the first feature extraction information corresponding to the target hidden layer and the hidden layer feature information corresponding to the target hidden layer to generate fused feature information. The hidden layer sequence consists of at least one serially connected hidden layer. The hidden layer at the second position is the last hidden layer in the hidden layer sequence.

[0089] As an example, the aforementioned execution entity can input the first feature extraction information corresponding to the target hidden layer and the hidden layer feature information corresponding to the target hidden layer into the feature fusion (Fuse) layer to generate fused feature information. The feature fusion layer can be a network layer that performs cascaded operation on multiple feature information channels.

[0090] Step 30242: Input the fused feature information into the next hidden layer in the target hidden layer of the hidden layer sequence to obtain hidden layer feature information, which is used as target hidden layer feature information.

[0091] In some embodiments, the execution entity may input the fused feature information into the next hidden layer in the target hidden layer sequence to obtain hidden layer feature information, which is then used as the target hidden layer feature information. The next hidden layer in the target hidden layer sequence may be a hidden layer serially connected to the layer following the target hidden layer.

[0092] Step 30243: Input the first feature extraction information and the second feature extraction information set corresponding to the target hidden layer into the feature extraction layer corresponding to the next hidden layer, so as to output the third feature extraction information and the fourth feature extraction information set.

[0093] In some embodiments, the execution entity may input the first feature extraction information and the second feature extraction information set corresponding to the target hidden layer to the feature extraction layer corresponding to the next hidden layer, so as to output the third feature extraction information and the fourth feature extraction information set.

[0094] Step 30244: The third feature extraction information and the target hidden layer feature information are fused to obtain fused feature information, which is used as the target fused feature information.

[0095] In some embodiments, the aforementioned execution entity may fuse the third feature extraction information and the target hidden layer feature information to obtain fused feature information, which may be used as the target fused feature information.

[0096] As an example, the aforementioned execution entity can input the third feature extraction information and the target hidden layer feature information into the feature fusion layer to generate the target fused feature information.

[0097] In step 30245, in response to determining that the next hidden layer is the second hidden layer in the hidden layer sequence, the target fusion feature information is input to the presentation click information output layer to generate presentation click information, and the third feature extraction information and the fourth feature extraction information set are input to the associated item click information output layer to generate the associated item click information set.

[0098] In some embodiments, in response to determining that the next hidden layer is the hidden layer at the second position in the hidden layer sequence, the aforementioned execution entity may input target fusion feature information to the presentation click information output layer to generate presentation click information, and input the third feature extraction information and the fourth feature extraction information set to the associated item click information output layer to generate the associated item click information set.

[0099] In some optional implementations of certain embodiments, after step 302, the steps further include:

[0100] In response to determining that the next hidden layer is not the hidden layer at the second position in the hidden layer sequence, the next hidden layer is taken as the target hidden layer, the target fused feature information is taken as the hidden layer feature information corresponding to the target hidden layer, the third feature extraction information is taken as the first feature extraction information corresponding to the target hidden layer, and the fourth feature extraction information set is taken as the second feature extraction information set corresponding to the target hidden layer, and the above generation steps are continued.

[0101] In some optional implementations of certain embodiments, after step 302, the steps further include:

[0102] In response to determining that the target hidden layer is the second hidden layer in the hidden layer sequence, the aforementioned execution entity can input the hidden layer feature information corresponding to the first hidden layer to the presentation click information output layer to obtain presentation click information, and input the first feature extraction information and the second feature extraction information set to the associated item click information output layer to generate the associated item click information set.

[0103] from Figure 3 It can be seen from this that, with Figure 2Compared to the description of some corresponding embodiments, Figure 3 In some corresponding embodiments, the model training method process 300 often extracts different types of features through hidden layers of different depths. To further enhance the network's knowledge transfer capabilities, feature fusion is achieved by extracting features from multiple hidden layers and fusing hidden layer feature information, realizing hierarchical cross-domain information fusion. This allows the model to learn more feature information and greatly improves the accuracy of the trained model.

[0104] Continue to refer to Figure 4 The flowchart 400 illustrates some embodiments of the information generation method according to this disclosure. The information generation method includes the following steps:

[0105] Step 401: Obtain the value presentation information of the target item.

[0106] In some embodiments, the entity executing the above-described information generation method (e.g., an electronic device) can acquire the target item value presentation information via wired or wireless means. The target item value presentation information may be the item value presentation information determined by the click information to be presented.

[0107] Step 402: Generate a click information generation model based on the multi-view click information generation model.

[0108] In some embodiments, the aforementioned execution entity can generate a click information generation model based on a multi-view click information generation model. This multi-view click information generation model is generated based on a model training method.

[0109] As an example, the aforementioned execution entity can decompose the presentation click information generation model from the multi-view click information generation model.

[0110] Step 403: Input the aforementioned target item value presentation information into the aforementioned presentation click information generation model to obtain presentation click information.

[0111] In some embodiments, the aforementioned executing entity may input the aforementioned target item value presentation information into the aforementioned presentation click information generation model to obtain presentation click information.

[0112] The above embodiments of this disclosure have the following beneficial effects: By using the information generation method of some embodiments of this disclosure, and utilizing the presentation click information generation model in the multi-view click information generation model, presentation click information for the value presentation information of the target item can be accurately generated.

[0113] Further reference Figure 5 As an implementation of the methods shown in the above figures, this disclosure provides some embodiments of a model training apparatus, which are similar to... Figure 2 Corresponding to the method embodiments shown, this model training device can be specifically applied to various electronic devices.

[0114] like Figure 5 As shown, a model training device 500 includes a first acquisition unit 501 and a training unit 503. The first acquisition unit 501 is configured to acquire a set of item value presentation information and a set of associated item information corresponding to each item value presentation information. The training unit 502 is configured to perform the following training steps for the item value presentation information set: inputting the target item value presentation information and the corresponding associated item information set from the item value presentation information set into an initial multi-view click information generation model to generate presentation click information and an associated item click information set, wherein the initial multi-view click information generation model is a model generated based on the initial presentation click information generation model and the initial associated item click information generation model; determining whether the initial multi-view click information generation model has finished training based on the presentation click information and the associated item click information set; and, in response to determining that training has finished, confirming the initial multi-view click information generation model as a multi-view click information generation model.

[0115] In some optional implementations of certain embodiments, the model training apparatus 500 further includes an update unit and an execution unit (not shown in the figure). The update unit can be configured to: in response to determining that training has not ended, update the initial multi-view click information generation model based on the presentation click information and the associated item click information set to generate an updated model, and remove the target item value presentation information from the item value presentation information set to obtain a removed information set. The execution unit can be configured to: use the updated model as the initial multi-view click information generation model and the removed information set as the item value presentation information set, and continue executing the above training steps.

[0116] In some optional implementations of some embodiments, the above-mentioned initial presentation click information generation model includes: at least one serially connected hidden layer and a presentation click information output layer; the above-mentioned initial associated item click information generation model includes: at least one serially connected feature extraction layer and an associated item click information output layer; and the training unit 502 can be further configured to: input the above-mentioned target item value presentation information into the first hidden layer at the first position in the hidden layer sequence to obtain the hidden layer feature information corresponding to the above-mentioned first hidden layer; input the above-mentioned target item value presentation information and the corresponding associated item information set into the feature extraction layer corresponding to the above-mentioned first hidden layer in the at least one feature extraction layer to obtain the first feature extraction information and the second feature extraction information set; determine the above-mentioned first hidden layer as the target hidden layer; in response to determining that the target hidden layer is not the hidden layer at the second position in the hidden layer sequence, for the hidden layer sequence, perform the following generation steps: input the target hidden layer into the first hidden layer at the first position in the hidden layer sequence and the first hidden layer at the first position in the hidden layer sequence to obtain the first feature extraction information and the second feature extraction information set; determine the target hidden layer as the target hidden layer; and in response to determining that the target hidden layer is not the hidden layer at the second position in the hidden layer sequence, perform the following generation steps for the hidden layer sequence: input the target hidden layer into the first hidden layer at the first position in the hidden layer sequence and the first hidden layer at the first position in the hidden layer sequence to obtain the first feature extraction information and the second feature extraction information set. The first feature extraction information corresponding to the hidden layer and the hidden layer feature information corresponding to the target hidden layer are fused to generate fused feature information. The fused feature information is input to the next hidden layer in the hidden layer sequence, which is the target hidden layer feature information. The first feature extraction information and the second feature extraction information set corresponding to the target hidden layer are input to the feature extraction layer corresponding to the next hidden layer to output the third feature extraction information and the fourth feature extraction information set. The third feature extraction information and the target hidden layer feature information are fused to obtain fused feature information, which is the target fused feature information. In response to determining that the next hidden layer is the hidden layer at the second position in the hidden layer sequence, the target fused feature information is input to the presentation click information output layer to generate presentation click information, and the third feature extraction information and the fourth feature extraction information set are input to the associated item click information output layer to generate the associated item click information set.

[0117] In some optional implementations of some embodiments, the training unit 502 may be further configured to: in response to determining that the next hidden layer is not the hidden layer at the second position in the hidden layer sequence, take the next hidden layer as the target hidden layer, take the target fused feature information as the hidden layer feature information corresponding to the target hidden layer, take the third feature extraction information as the first feature extraction information corresponding to the target hidden layer, take the fourth feature extraction information set as the second feature extraction information set corresponding to the target hidden layer, and continue to perform the above generation steps.

[0118] In some optional implementations of some embodiments, the training unit 502 may be further configured to: in response to determining that the target hidden layer is the hidden layer at the second position in the hidden layer sequence, input the hidden layer feature information corresponding to the first hidden layer to the presentation click information output layer to obtain presentation click information, and input the first feature extraction information and the second feature extraction information set to the associated item click information output layer to generate the associated item click information set.

[0119] It is understandable that the units described in the model training device 500 are related to the reference... Figure 2 The steps in the described method correspond to each other. Therefore, the operations, features, and beneficial effects described above for the method also apply to the model training device 500 and the units contained therein, and will not be repeated here.

[0120] Further reference Figure 6 As an implementation of the methods shown in the above figures, this disclosure provides some embodiments of an information generation apparatus, which are similar to... Figure 4 Corresponding to the method embodiments shown, this information generation device can be specifically applied to various electronic devices.

[0121] like Figure 6 As shown, an information generation device 600 includes: a second acquisition unit 601, a generation unit 602, and an input unit 603. The second acquisition unit 601 is configured to acquire target item value presentation information; the generation unit 602 is configured to generate a presentation click information generation model based on a multi-view click information generation model, wherein the multi-view click information generation model is generated based on a model training method; and the input unit 603 is configured to input the target item value presentation information into the presentation click information generation model to obtain presentation click information.

[0122] It is understandable that the units described in the information generation device 600 and the reference Figure 4 The steps in the described method correspond to each other. Therefore, the operations, features, and beneficial effects described above for the method also apply to the information generation device 600 and the units contained therein, and will not be repeated here.

[0123] The following is for reference. Figure 7 It illustrates electronic devices suitable for implementing some embodiments of this disclosure (e.g., Figure 1 A schematic diagram of the structure of electronic device 101) 700. Figure 7 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of this disclosure.

[0124] like Figure 7As shown, the electronic device 700 may include a processing unit (e.g., a central processing unit, a graphics processing unit, etc.) 701, which can perform various appropriate actions and processes according to a program stored in a read-only memory 702 or a program loaded from a storage device 708 into a random access memory 703. The random access memory 703 also stores various programs and data required for the operation of the electronic device 700. The processing unit 701, the read-only memory 702, and the random access memory 703 are interconnected via a bus 704. An input / output interface 705 is also connected to the bus 704.

[0125] Typically, the following devices can be connected to the input / output interface 705: input devices 706 including, for example, a touchscreen, touchpad, keyboard, mouse, camera, microphone, accelerometer, gyroscope, etc.; output devices 707 including, for example, a liquid crystal display (LCD), speaker, vibrator, etc.; storage devices 708 including, for example, magnetic tape, hard disk, etc.; and communication devices 709. Communication device 709 allows electronic device 700 to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 7 An electronic device 700 with various devices is shown; however, it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed alternatively. Figure 7 Each box shown can represent a device or multiple devices as needed.

[0126] In particular, according to some embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, some embodiments of this disclosure include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 709, or installed from a storage device 708, or installed from a read-only memory 702. When the computer program is executed by the processing device 701, it performs the functions defined in the methods of some embodiments of this disclosure.

[0127] It should be noted that, in some embodiments of this disclosure, the computer-readable medium described above may be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium may be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In some embodiments of this disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In some embodiments of this disclosure, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0128] In some implementations, clients and servers can communicate using any currently known or future-developed network protocol such as HTTP (Hypertext Transfer Protocol) and can interconnect with digital data communication (e.g., communication networks) of any form or medium. Examples of communication networks include local area networks (“LANs”), wide area networks (“WANs”), the Internet (e.g., the Internet of Things), and peer-to-peer networks (e.g., ad hoc peer-to-peer networks), as well as any currently known or future-developed networks.

[0129] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device. The aforementioned computer-readable medium carries one or more programs that, when executed by the electronic device, cause the electronic device to: acquire a set of item value presentation information and a set of associated item information corresponding to each item value presentation information; for the item value presentation information set, perform the following training steps: input the target item value presentation information and the corresponding associated item information set from the item value presentation information set into an initial multi-view click information generation model to generate presentation click information and a set of associated item click information, wherein the initial multi-view click information generation model is a model generated based on the initial presentation click information generation model and the initial associated item click information generation model; determine whether the initial multi-view click information generation model has finished training based on the presentation click information and the set of associated item click information; in response to determining that training has finished, determine the initial multi-view click information generation model as a multi-view click information generation model.

[0130] Computer program code for performing operations of some embodiments of this disclosure can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0131] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0132] The units described in some embodiments of this disclosure can be implemented in software or hardware. The described units can also be housed in a processor; for example, a processor may be described as including a first acquisition unit and a training unit. The names of these units do not necessarily limit the unit itself; for example, the acquisition unit may also be described as "a unit that acquires a set of item value presentation information and a set of associated item information corresponding to each item value presentation information."

[0133] The functions described above in this document can be performed at least in part by one or more hardware logic components. For example, exemplary types of hardware logic components that can be used, without limitation, include: field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), system-on-a-chip (SoCs), complex programmable logic devices (CPLDs), and so on.

[0134] Some embodiments of this disclosure also provide a computer program product, including a computer program that, when executed by a processor, implements any of the model training methods or information generation methods described above.

[0135] The above description is merely a selection of preferred embodiments of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in the embodiments of this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described inventive concept. For example, technical solutions formed by substituting the above-described features with (but not limited to) technical features with similar functions disclosed in the embodiments of this disclosure.

Claims

1. A model training method, comprising: Obtain an item value presentation information set and an associated item information set corresponding to each item value presentation information set, wherein the associated item information set corresponding to the item value presentation information and the item value presentation information share the same click user characteristics; For the set of information presenting the value of items, perform the following training steps: The target item value presentation information and the corresponding associated item information set are input into the initial multi-view click information generation model to generate presentation click information and associated item click information sets. The initial multi-view click information generation model is a model generated based on the initial presentation click information generation model and the initial associated item click information generation model. The initial presentation click information generation model includes at least one serially connected hidden layer, and the initial associated item click information generation model includes at least one serially connected feature extraction layer. The hidden layers in the at least one serially connected hidden layer and the feature extraction layers in the at least one serially connected feature extraction layer have a network correspondence. Based on the presented click information and the set of click information for related items, determine whether the initial multi-view click information generation model has finished training; In response to the conclusion of training, the initial multi-view click information generation model is defined as the multi-view click information generation model.

2. The method according to claim 1, wherein, The method further includes: In response to the determination that the training has not ended, the initial multi-view click information generation model is updated based on the presentation click information and the associated item click information set to generate the updated model, and the target item value presentation information is removed from the item value presentation information set to obtain the removed information set. The updated model is used as the initial multi-view click information generation model, and the information set after removal is used as the item value presentation information set. The training steps are then continued.

3. The method according to claim 1, wherein, The initial presentation click information generation model includes: at least one serially connected hidden layer and a presentation click information output layer; the initial associated item click information generation model includes: at least one serially connected feature extraction layer and an associated item click information output layer; and The step of inputting the target item value presentation information and the corresponding associated item information set from the item value presentation information set into the initial multi-view click information generation model to generate presentation click information and associated item click information set includes: The target item value presentation information is input into the first hidden layer at the first position in the hidden layer sequence to obtain the hidden layer feature information corresponding to the first hidden layer. The target item value presentation information and the corresponding associated item information set are input into the feature extraction layer corresponding to the first hidden layer in at least one feature extraction layer to obtain the first feature extraction information and the second feature extraction information set. The first hidden layer is determined as the target hidden layer; In response to determining that the target hidden layer is not the second hidden layer in the hidden layer sequence, the following generation steps are performed for the hidden layer sequence: The first feature extraction information corresponding to the target hidden layer and the hidden layer feature information corresponding to the target hidden layer are fused to generate fused feature information. The fused feature information is input into the next hidden layer in the target hidden layer of the hidden layer sequence to obtain hidden layer feature information, which is used as target hidden layer feature information. The first and second feature extraction information sets corresponding to the target hidden layer are input into the feature extraction layer corresponding to the next hidden layer to output the third and fourth feature extraction information sets. The third feature extraction information and the target hidden layer feature information are fused to obtain fused feature information, which is used as the target fused feature information. In response to determining that the next hidden layer is the second hidden layer in the hidden layer sequence, the target fusion feature information is input to the presentation click information output layer to generate presentation click information, and the third feature extraction information and the fourth feature extraction information set are input to the associated item click information output layer to generate the associated item click information set.

4. The method according to claim 3, wherein, The method further includes: In response to determining that the next hidden layer is not the hidden layer at the second position in the hidden layer sequence, the next hidden layer is taken as the target hidden layer, the target fused feature information is taken as the hidden layer feature information corresponding to the target hidden layer, the third feature extraction information is taken as the first feature extraction information corresponding to the target hidden layer, and the fourth feature extraction information set is taken as the second feature extraction information set corresponding to the target hidden layer, and the generation step is continued.

5. The method according to claim 3, wherein, The method further includes: In response to determining that the target hidden layer is the second hidden layer in the hidden layer sequence, the hidden layer feature information corresponding to the first hidden layer is input to the presentation click information output layer to obtain presentation click information, and the first feature extraction information and the second feature extraction information set are input to the associated item click information output layer to generate the associated item click information set.

6. The method according to claim 1, wherein, The item value presentation information set and the associated item information set corresponding to each item value presentation information set are stored through the following steps: For each item value presentation information in the item value presentation information set, perform the following storage steps: The associated item information set corresponding to the item value presentation information is downsampled to obtain the downsampled item information set. Determine the common feature information between the downsampled item information set and the item value presentation information; Remove the common feature information from the item value presentation information to obtain the item value presentation information after removal; The common feature information is removed from each downsampled item information in the downsampled item information set to generate the removed item information, thus obtaining the removed item information set; The information set of removed items, the value presentation information of removed items, and the common feature information are stored accordingly.

7. An information generation method, comprising: Obtain information on the value of the target item; Based on the multi-view click information generation model, a presentation click information generation model is generated, wherein the multi-view click information generation model is generated based on the method described in any one of claims 1-6; The value presentation information of the target item is input into the presentation click information generation model to obtain the presentation click information.

8. A model training device, comprising: The first acquisition unit is configured to acquire an item value presentation information set and an associated item information set corresponding to each item value presentation information set, wherein the associated item information set corresponding to the item value presentation information set and the item value presentation information set share the same click user characteristic. The training unit is configured to perform the following training steps for a set of item value presentation information: Inputting the target item value presentation information and the corresponding associated item information set from the item value presentation information set into an initial multi-view click information generation model to generate presentation click information and associated item click information sets. The initial multi-view click information generation model is a model generated based on the initial presentation click information generation model and the initial associated item click information generation model. The initial presentation click information generation model includes at least one serially connected hidden layer, and the initial associated item click information generation model includes at least one serially connected feature extraction layer. The hidden layers in the at least one serially connected hidden layer and the feature extraction layers in the at least one serially connected feature extraction layer have a network correspondence. Based on the presentation click information and the associated item click information sets, it is determined whether the initial multi-view click information generation model has finished training. In response to determining that training has finished, the initial multi-view click information generation model is determined as a multi-view click information generation model.

9. An information generation device, comprising: The second acquisition unit is configured to acquire the value presentation information of the target item; The generation unit is configured to generate a model based on multi-view click information, and to generate a presentation click information generation model, wherein the multi-view click information generation model is generated based on the method described in any one of claims 1-6; The input unit is configured to input the value presentation information of the target item into the presentation click information generation model to obtain presentation click information.

10. An electronic device, comprising: One or more processors; Storage device, on which one or more programs are stored, When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any one of claims 1-7.

11. A computer-readable medium having a computer program stored thereon, wherein, When the computer program is executed by a processor, it implements the method as described in any one of claims 1-7.

12. A computer program product comprising a computer program that, when executed by a processor, implements the method according to any one of claims 1-7.

Images