Material recommendation method, device and computer-readable storage medium

Abstract

The present invention discloses a material recommendation method, device, and computer-readable storage medium. The method comprises: receiving a material recommendation request based on a target object from a mobile terminal, obtaining static attribute data of the target object and dynamic interaction data of the target object; extracting static attribute features of the static attribute data and interaction features corresponding to the dynamic interaction data; obtaining unified material features corresponding to a plurality of candidate materials in a predetermined scenario in a unified feature space; determining a recommendation result including a plurality of target materials recommended to the target object in the predetermined scenario based on the unified material features, static attribute features, and interaction features corresponding to the plurality of candidate materials, and feeding the recommendation result back to the mobile terminal. The present invention solves the technical problem in the related art of low accuracy of recommendation results when recommending heterogeneous materials, resulting in a poor user experience.

Description

Technical Field

[0001] The present invention relates to the field of computers, and in particular to a material recommendation method, device, and computer-readable storage medium. Background Art

[0002] With the increasing application of recommendation systems, recommendation scenarios have become increasingly complex. Within a single scenario, it's often necessary to recommend a wide variety of candidate items, often referred to as heterogeneous item recommendations. For example, in shopping apps, recommendations may include products, services, videos, marketing materials, and more. To enable the same system to recommend multiple items, a common approach is to model, train, and infer different types of items separately, ultimately integrating the recommended results across multiple items through a set of rule systems. However, this rule system is inflexible, resulting in low accuracy in the resulting recommendations.

[0003] To address the above-mentioned problems, no effective solutions have been proposed so far. Summary of the Invention

[0004] Embodiments of the present invention provide a material recommendation method, device, and computer-readable storage medium to at least solve the technical problem in the related art that when recommending heterogeneous materials, the recommendation results are inaccurate, resulting in a poor user experience.

[0005] According to one aspect of an embodiment of the present invention, a material recommendation method is provided, comprising: receiving a material recommendation request based on a target object sent by a mobile terminal, obtaining static attribute data of the target object and dynamic interaction data of the target object; extracting static attribute features of the static attribute data, and interaction features corresponding to the dynamic interaction data; obtaining unified material features corresponding to a plurality of alternative materials in a unified feature space under a predetermined scenario; determining a recommendation result including a plurality of target materials recommended to the target object under the predetermined scenario based on the unified material features, the static attribute features and the interaction features corresponding to the plurality of alternative materials, and feeding back the recommendation result to the mobile terminal.

[0006] Optionally, obtaining unified material features corresponding to multiple alternative materials in a predetermined scenario in a unified feature space includes: extracting original material features of the multiple alternative materials in the corresponding feature space; mapping the original material features corresponding to the multiple alternative materials to the unified feature space based on a space mapping function, to obtain the unified material features corresponding to the multiple alternative materials in the predetermined scenario in the unified feature space.

[0007] Optionally, based on the unified material characteristics, the static attribute characteristics and the interaction characteristics corresponding to the multiple alternative materials, a recommendation result including multiple target materials recommended to the target object in the predetermined scenario is determined, and the recommendation result is fed back to the mobile terminal, including: based on the interaction characteristics, determining a first interaction characteristic between the target object and the selected material in the dynamic interaction data, and a second interaction characteristic between the selected material and the object; splicing the unified material characteristics, the static attribute characteristics and the first interaction characteristics to obtain a first target characteristic, and splicing the unified material characteristics, the static attribute characteristics and the second interaction characteristics to obtain a second target characteristic; based on the first target characteristic and the second target characteristic, determining a recommendation result including multiple target materials recommended to the target object in the predetermined scenario, and feeding back the recommendation result to the mobile terminal.

[0008] Optionally, the extracting of static attribute features of the static attribute data and interaction features corresponding to the dynamic interaction data includes: using a feature extraction module of a material recommendation module to extract static attribute features of the static attribute data and interaction features corresponding to the dynamic interaction data; determining a recommendation result including multiple target materials recommended to the target object in the predetermined scenario based on the unified material features corresponding to the multiple alternative materials, the static attribute features and the interaction features, and feeding back the recommendation result to the mobile terminal includes: inputting the unified material features, the static attribute features and the interaction features corresponding to the multiple alternative materials into the recommendation module of the material recommendation model, obtaining a recommendation result including multiple target materials recommended to the target object in the predetermined scenario, and feeding back the recommendation result to the mobile terminal; wherein, the material recommendation model is trained using multiple groups of sample data, and the multiple groups of sample data include: static attribute data of sample objects, dynamic interaction data of sample objects, and sample recommendation results recommended to the sample objects.

[0009] Optionally, the method also includes: training the multiple groups of sample data in the following manner to obtain the material recommendation model: obtaining an initial recommendation model, and constructing a loss function for model training, wherein the loss function includes a first loss function term, a second loss function term, and a third loss function term, the first loss function term is used to maintain the relative stability of materials of the same type in the unified feature space, the second loss function term is used to maintain the alignment of different types of materials in the unified feature space, and the third loss function term is used to classify whether the alternative material is the target material; based on the loss function, the multiple groups of sample data are used to train the initial recommendation model to obtain the material recommendation model.

[0010] Optionally, the method also includes: constructing the first loss function term in the following manner: obtaining the original material characteristics of the materials of the same type in the corresponding feature space, and the unified material characteristics of the materials of the same type in the unified feature space; constructing the first loss function term based on the original material characteristics corresponding to the materials of the same type, and the unified material characteristics corresponding to the materials of the same type; constructing the second loss function term in the following manner: obtaining the unified material characteristics of different types of materials in the unified feature space; constructing the second loss function based on the unified material characteristics of the different types of materials in the unified feature space.

[0011] According to one aspect of an embodiment of the present invention, a material recommendation method is provided, comprising: obtaining a material recommendation request based on a target object, and sending the material recommendation request to a distributed processing cluster; receiving a recommendation result fed back by the distributed processing cluster, wherein the recommendation result includes a plurality of target materials recommended to the target object under a predetermined scenario, and the recommendation result is determined based on unified material features corresponding to a plurality of alternative materials in a unified feature space under the predetermined scenario, static attribute features of static attribute data of the target object, and interactive features of dynamic interactive data of the target object; and displaying the recommendation result on an interactive interface.

[0012] According to one aspect of an embodiment of the present invention, a material recommendation device is provided, comprising: a first receiving module for receiving a material recommendation request based on a target object sent by a mobile terminal, and obtaining static attribute data of the target object and dynamic interaction data of the target object; an extraction module for extracting static attribute features of the static attribute data and interaction features corresponding to the dynamic interaction data; a first acquisition module for obtaining unified material features corresponding to a plurality of alternative materials in a predetermined scenario in a unified feature space; a determination module for determining a recommendation result including a plurality of target materials recommended to the target object in the predetermined scenario based on the unified material features corresponding to the plurality of alternative materials, the static attribute features and the interaction features, and feeding back the recommendation result to the mobile terminal.

[0013] According to one aspect of an embodiment of the present invention, a material recommendation device is provided, comprising: a first sending module, configured to obtain a material recommendation request based on a target object, and to send the material recommendation request to a distributed processing cluster; a second receiving module, configured to receive a recommendation result fed back by the distributed processing cluster, wherein the recommendation result includes a plurality of target materials recommended to the target object under a predetermined scenario, and the recommendation result is determined based on unified material features corresponding to a plurality of alternative materials in a unified feature space under the predetermined scenario, static attribute features of static attribute data of the target object, and interactive features of dynamic interactive data of the target object; and a display module, configured to display the recommendation result on an interactive interface.

[0014] According to one aspect of an embodiment of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium includes a stored program, wherein when the program is run, the device where the computer-readable storage medium is located is controlled to execute any one of the above-mentioned material recommendation methods.

[0015] According to one aspect of an embodiment of the present invention, a computer device is provided, comprising: a memory and a processor, wherein the memory stores a computer program; the processor is configured to execute the computer program stored in the memory, wherein when the computer program is executed, the processor executes any one of the above-described material recommendation methods.

[0016] In an embodiment of the present invention, a material recommendation request based on a target object is received from a mobile terminal, static attribute data and dynamic interaction data of the target object are obtained, static attribute features and interaction features are extracted based on the obtained data, and then unified material features corresponding to multiple candidate materials in a predetermined scenario in a unified feature space are obtained. Based on the unified material features, static attribute features, and interaction features corresponding to the multiple candidate materials, a recommendation result including multiple target materials is determined for the target object in the predetermined scenario, and the recommendation result is fed back to the mobile terminal. Because the recommendation result including multiple target materials is obtained based on the unified material features, static attribute features, and interaction features corresponding to the multiple candidate materials, the unified material features are in a unified feature space and can better learn the relationship between heterogeneous materials; the static attribute features are obtained based on the static attribute data of the target object and are more targeted; and the interaction features are obtained based on the dynamic interaction data of the target object and can obtain features that conform to the actual situation of the target object. Therefore, the recommendation result including multiple target materials is more accurate, thereby solving the technical problem in the related art of low recommendation accuracy when recommending heterogeneous materials, resulting in a poor user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0018] Figure 1 2 is an architecture diagram of a material recommendation system for implementing a material recommendation method according to an embodiment of the present invention;

[0019] Figure 2 is a flow chart of a material recommendation method 1 according to embodiment 1 of the present invention;

[0020] Figure 3 is a flow chart of a second material recommendation method according to embodiment 1 of the present invention;

[0021] Figure 4 is a flow chart of a material recommendation method provided according to an optional embodiment of the present invention;

[0022] Figure 5 This is a structural block diagram of a material recommendation device 1 provided according to Example 2 of the present invention;

[0023] Figure 6 This is a structural block diagram of a second material recommendation device provided according to embodiment 3 of the present invention;

[0024] Figure 7 is a device block diagram of a terminal according to an embodiment of the present invention. DETAILED DESCRIPTION

[0025] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0026] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0027] First, some nouns or terms that appear in the description of the embodiments of the present application are subject to the following interpretations:

[0028] Heterogeneous materials: Heterogeneous materials refer to recommendation candidates with different properties. For example, in the same scenario, you may need to recommend products, articles, videos, or music, movies, books, etc.

[0029] Original features: Features used to describe materials based on the material's own properties or related statistical values.

[0030] Feature space: Here, feature space means that for each recommended material, it can be represented by a vector related to its characteristics, and the space where the vector is located is the feature space.

[0031] CCCFNet: Cross-domain Content-boosted Collaborative Filtering neuralNetwork, a cross-domain recommendation algorithm that integrates collaborative filtering features and content-based features into a multi-view network.

[0032] DDTCDR: Deep Dual Transfer Cross Domain Recommendation, a cross-domain recommendation algorithm based on transfer learning.

[0033] CORAL loss: A commonly used domain adaptation loss function.

[0034] Example 1

[0035] According to an embodiment of the present invention, an embodiment of a material recommendation method is also provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0036] The method embodiment provided in Example 1 of the present application can be applied to various recommendation scenarios. Figure 1 is an architecture diagram of a material recommendation system for implementing a material recommendation method according to an embodiment of the present invention. Figure 1 As shown, the material recommendation system includes: a mobile terminal (on the terminal side, such as mobile terminal A and mobile terminal B in the figure) and a distributed processing cluster (on the cloud side). The mobile terminal can include various types, such as mobile phones, laptops, PADs, wearable electronic devices, etc. The distributed processing cluster can also include various types, such as computers with distributed processing capabilities, servers, etc. The mobile terminal and the distributed processing cluster communicate via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0037] In the above-mentioned material recommendation system, the mobile terminal mainly includes: a recommendation request triggering module, a static attribute information acquisition module, a dynamic behavior information extraction module, a material recommendation request sending module, and a material recommendation result display module. The recommendation request triggering module is used to trigger a recommendation request. For example, it can trigger a corresponding recommendation request when a user opens an application on the mobile terminal, or on the opened application interface (for example, by entering a corresponding search request). The static attribute information acquisition module is used to obtain the static attribute information of the user (corresponding to the user account). The static attribute information of the user can include various types, for example, it can be fixed representation information of the user, such as identity and appearance characteristics that are disclosed with the user's permission. The dynamic behavior information extraction module is used to obtain some interaction information completed by the user in the predetermined application. For example, it can be the user's historical browsing information in the predetermined application within a certain historical time, or it can be the interactive operation information of the user interacting with the predetermined application. The dynamic behavior information extraction module can extract the required dynamic interaction information from the log record file of the predetermined application. The time period corresponding to the dynamic interaction information can be flexibly set based on specific needs. It should be noted that in the material recommendation request triggered by the above-mentioned recommendation request triggering module, during the triggering process, the recommendation request may carry the user's static attribute information obtained by the above-mentioned static attribute information acquisition module and the dynamic interaction information on the predetermined application end extracted by the dynamic behavior information extraction module. After obtaining the above-mentioned material recommendation request including the static attribute information and dynamic interaction information, the above-mentioned material recommendation request sending module sends the material recommendation request to the distributed processing cluster. The distributed processing cluster processes the material recommendation request to obtain the material recommendation result. The above-mentioned material recommendation result display module can display the material recommendation result in the predetermined application after receiving the material recommendation result sent by the distributed processing cluster.

[0038] It should be noted that the above-mentioned mobile terminals may be one or more, and the mobile terminal A and the mobile terminal B shown in the figure are only examples.

[0039] In the aforementioned material recommendation system, the distributed processing cluster primarily includes a material recommendation request receiving module, a module for extracting raw features from heterogeneous materials, a module for mapping unified features for heterogeneous materials, a material recommendation model training module, a feature concatenation module, a target heterogeneous material decision module, and a module for transmitting heterogeneous material recommendation results. The material recommendation model training module includes a loss function construction unit and a training sample acquisition unit. The loss function construction unit further includes a loss term construction subunit for similar materials and a loss term construction subunit for heterogeneous materials. Each of these modules and units is briefly described below.

[0040] The material recommendation request receiving module is used to receive the material recommendation request sent by the mobile terminal, wherein the material recommendation request carries the static attribute data of the target object (corresponding to the static attribute information of the above-mentioned user) and the dynamic interaction data of the target object (corresponding to the dynamic interaction information of the above-mentioned user).

[0041] The above-mentioned original feature extraction module for heterogeneous materials is used to extract the features of multiple materials in the corresponding feature space, that is, the original material features. It should be noted that since multiple materials belong to different features, the corresponding feature spaces used are also different. In order to achieve the uniformity of multiple heterogeneous materials, it is necessary to map multiple materials into a unified feature space. The unified feature mapping module for heterogeneous materials is used to map the original material features corresponding to multiple heterogeneous materials into a unified feature space based on a predetermined space mapping function, and obtain the unified material features corresponding to the multiple heterogeneous materials.

[0042] The material recommendation model training module is used to train the material recommendation model. The material recommendation model training basic block includes the acquisition of training samples and the construction of a loss function when training the material recommendation model. The training sample acquisition unit is used to obtain sample data for model training. The acquisition of sample data can be carried out in a variety of ways, for example, samples can be obtained by manual labeling, samples can be obtained by machine labeling, or samples obtained by a machine based on manual labeling, etc. The loss function construction unit is used to construct a loss function for the material recommendation model, wherein the loss function can include a loss item for similar materials and a loss item for heterogeneous materials based on whether the materials are similar. The loss item construction subunit for similar materials is used to construct a loss item for similar materials. The loss item can be used to characterize that the difference between the uniform characteristics of materials of the same type is as small as possible. The loss item construction subunit for heterogeneous materials is used to construct a loss item for heterogeneous materials. The loss item is used to characterize that the difference between the uniform characteristics of materials of heterogeneous materials is as large as possible.

[0043] The feature splicing module is used to splice the static attribute features corresponding to the static attribute information carried in the material recommendation request, the interactive features corresponding to the dynamic interactive information, and the unified material features of multiple heterogeneous materials to obtain the corresponding spliced ​​features. In the process of splicing the above features, a variety of splicing methods can be used. For example, the static attribute features, the unified material features, and the modified interactive features of the above interactive features can be spliced ​​together. The modified interactive features of the interactive features corresponding to the above dynamic interactive data include: the interactive features between the target user (i.e., the user who sent the recommendation request) and the material in the dynamic interactive data, and the interactive features between the material in the dynamic interactive data and other users (not the user who sent the recommendation request).

[0044] The target heterogeneous material decision module determines a recommendation result, including multiple target materials, for the target user based on the original spliced ​​features obtained by the feature splicing module. The target heterogeneous material decision module may generate recommendation results based on a multi-layer neural network model. The heterogeneous material recommendation result sending module is used to send the recommendation results, including multiple target materials, to the mobile terminal.

[0045] Therefore, based on the above-mentioned material recommendation system, the distributed processing cluster pre-trains the machine based on the collected sample data (for example, the historical behavior data of users when browsing the web) to obtain a material recommendation model; the mobile terminal triggers a material recommendation request and sends the material recommendation request carrying the user's static attribute information and dynamic interaction information to the distributed processing cluster. The distributed processing cluster determines the recommendation result corresponding to the material recommendation request based on the material recommendation model and feeds back the obtained recommendation result to the mobile terminal. In the process of the distributed processing cluster determining the recommendation result based on the material recommendation model, the recommendation model can be a material recommendation model for recommending multiple heterogeneous materials; the mobile terminal displays the recommendation result on the display interface.

[0046] As an optional embodiment, in the case where the above recommendation model is a material recommendation model for heterogeneous materials, the present application provides the following Figure 2 The material recommendation method shown is used to recommend multiple heterogeneous materials. This material recommendation method is mainly applied to the above-mentioned distributed processing cluster, and is used for the distributed processing cluster to feedback the recommendation results corresponding to the material recommendation request to the user based on the material recommendation request received from the user. Figure 2 : is a flow chart of a material recommendation method 1 according to embodiment 1 of the present invention, such as Figure 2 As shown, the method includes the following steps:

[0047] Step S202: receiving a material recommendation request based on a target object sent by a mobile terminal, and obtaining static attribute data and dynamic interaction data of the target object;

[0048] Step S204, extracting static attribute features of the static attribute data and interaction features corresponding to the dynamic interaction data;

[0049] Step S206, obtaining unified material features corresponding to the plurality of candidate materials in the predetermined scenario in the unified feature space;

[0050] Step S208 , based on the unified material features, static attribute features, and interactive features corresponding to the multiple candidate materials, a recommendation result including multiple target materials recommended to the target object in a predetermined scenario is determined, and the recommendation result is fed back to the mobile terminal.

[0051] Through the above steps, the distributed processing cluster (cloud) receives a material recommendation request based on a target object from a mobile terminal, obtains static attribute data and dynamic interaction data of the target object, extracts static attribute features and interaction features based on the obtained data, and then obtains unified material features corresponding to multiple candidate materials in a unified feature space in a predetermined scenario. Based on the unified material features, static attribute features, and interaction features corresponding to the multiple candidate materials, a recommendation result including multiple target materials is determined for the target object in the predetermined scenario, and the recommendation result is fed back to the mobile terminal. Because the recommendation result including multiple target materials is obtained based on the unified material features, static attribute features, and interaction features corresponding to the multiple candidate materials, the unified material features are in a unified feature space and can better learn the relationship between heterogeneous materials; the static attribute features are obtained based on the static attribute data of the target object and are more targeted; the interaction features are obtained based on the dynamic interaction data of the target object and can obtain features that conform to the actual situation of the target object. Therefore, the recommendation result including multiple target materials is more accurate, thereby solving the technical problem in the related art of low accuracy of recommendation results when recommending heterogeneous materials, resulting in poor user experience.

[0052] As an optional embodiment, a material recommendation request based on a target object sent by a mobile terminal is received, and the static attribute data of the target object and the dynamic interaction data of the target object are obtained. Based on the material recommendation request of the target object, the static attribute data of the target object and the dynamic interaction data of the target object within a predetermined time period can be selectively obtained. When obtaining the dynamic interaction data of the target object, the data of all dynamic interaction behaviors of the target object can be obtained, or the data of dynamic interaction behaviors with multiple alternative materials in a predetermined scenario can be obtained based on the material recommendation request. By obtaining the static attribute data of the target object and the dynamic interaction data of the target object, it can be ensured that the recommendation results obtained subsequently are for the target object, and are obtained based on the static attributes and historical behaviors of the target object, so that the recommendation results obtained are targeted.

[0053] Among them, the static attribute data of the target object can be the basic information data of the target object, such as occupation, natural person relationship, education level, place of residence, etc. The dynamic interaction data of the target object can be the data of the target object's interactive behavior, such as consumption data, behavior data, etc. In the case where the interaction data of the target object obtained is the interaction data with a variety of alternative materials, the data obtained will be different according to different scenarios and different alternative materials, depending on the specific scenarios and materials. The acquisition of data on dynamic interaction behaviors with a variety of alternative materials in predetermined scenarios can be achieved through a variety of channels, for example, by receiving a material recommendation request based on the target object sent by the mobile terminal, that is, automatically acquiring the data based on the behavior log, network log and other data in the target object's mobile terminal when the target object is authorized to obtain.

[0054] As an optional embodiment, after acquiring the static attribute data and dynamic interaction data of the target object, the static attribute features of the static attribute data and the interaction features corresponding to the dynamic interaction data are extracted. This allows for accurate extraction of the static attribute features and interaction features of the target object based on the acquired data. This allows for better processing of material recommendation requests based on these features.

[0055] It should be noted that when extracting static attribute features of static attribute data and interaction features corresponding to dynamic interaction data, a feature extraction module can be used for extraction. For example, the feature extraction module of the material recommendation module can be used to extract static attribute features of static attribute data and interaction features corresponding to dynamic interaction data.

[0056] It should also be noted that the interaction features are further divided into interaction features between the target object and multiple candidate materials, and interaction features between each of the multiple candidate materials and multiple objects, that is, multiple objects related to the material in the predetermined scenario. Therefore, based on the interaction features, a first interaction feature can be determined between the target object and the selected material in the dynamic interaction data, that is, the candidate material, and a second interaction feature can be determined between the selected material and the object, that is, any one of the multiple candidate materials and multiple objects related to the material in the predetermined scenario. By obtaining the first and second interaction features, it is possible to better obtain the representation of the material and the object, more clearly represent the relationship between the candidate material and the target object, and facilitate better learning and training.

[0057] As an optional embodiment, unified material features corresponding to a plurality of alternative materials in a predetermined scenario in a unified feature space are obtained. First of all, it should be noted that the plurality of alternative materials are different depending on the predetermined scenario. For example, in the application scenario of a takeaway platform, the plurality of alternative materials may be food, beverages, medicines, and so on. In the application scenario of a shopping platform, the plurality of alternative materials may be commodities, recommended videos, comments, and so on. Based on the problem that the correlation between the plurality of materials is not tight due to the fact that the relevant technology only trains the plurality of alternative materials separately and then integrates them, in which case the unified material features corresponding to the plurality of alternative materials in a predetermined scenario in a unified feature space are obtained. That is, the features corresponding to different types of alternative materials are unified into one feature space, which solves the problem of loose correlation caused by different features or different distributions of different types of alternative materials.

[0058] As an optional embodiment, when obtaining the unified material features corresponding to a plurality of alternative materials in a predetermined scenario in a unified feature space, the original material features of the plurality of alternative materials in the corresponding feature space can be extracted, and then the original material features corresponding to the plurality of alternative materials are mapped to the unified feature space according to the space mapping function, so that they are transformed into a new feature space, and the unified material features corresponding to the plurality of alternative materials in the predetermined scenario in the unified feature space are obtained, which unifies the features corresponding to different types of alternative materials into one feature space, solves the problem that the features of different types of alternative materials are different or distributed differently, and enables the features corresponding to different types of alternative materials to be well integrated into the same feature space, thereby better capturing the relationship between each other. It should be noted that the form of the mapping function involved in the optional embodiment of the present invention is not limited, and a commonly used fully connected neural network can be used.

[0059] As an optional embodiment, based on the unified material characteristics, static attribute characteristics and interactive characteristics corresponding to the multiple alternative materials, a recommendation result including multiple target materials recommended to the target object in a predetermined scenario is determined, and the recommendation result is fed back to the mobile terminal. By combining multiple characteristics, it is possible to determine a recommendation result including multiple target materials recommended to the target object in a predetermined scenario, and feed the recommendation result back to the mobile terminal to achieve interaction between the cloud and the mobile terminal. In particular, when the first interactive feature and the second interactive feature are determined based on the interactive feature, the unified material feature, the static attribute feature and the first interactive feature are spliced ​​to obtain the first target feature, that is, the representation of the target object is obtained, and the unified material feature, the static attribute feature and the second interactive feature are spliced ​​to obtain the second target feature, that is, the representation of the material is obtained, and then based on the first target feature and the second target feature, a recommendation result including multiple target materials recommended to the target object in the predetermined scenario is determined.

[0060] As an optional embodiment, the process of determining a recommendation result including multiple target materials recommended to a target object in a predetermined scenario based on unified material features, static attribute features, and interaction features corresponding to multiple candidate materials can be obtained through a material recommendation model. Specifically, the unified material features, static attribute features, and interaction features corresponding to the multiple candidate materials can be input into a recommendation module of the material recommendation model to obtain a recommendation result including multiple target materials recommended to the target object in the predetermined scenario. The material recommendation model is trained using multiple sets of sample data, including static attribute data of the sample objects, dynamic interaction data of the sample objects, and sample recommendation results recommended to the sample objects. Because the material recommendation model is continuously trained using multiple sets of sample data, it can specifically learn based on the static attribute data and dynamic interaction data of the sample objects, effectively avoiding problems such as mismatched recommendation results, inaccurate recommendation results, and missed recommended materials. This allows the material recommendation model to better learn and learn distinctive features. As it continues to train with samples, the accuracy of the recommendation results is gradually improved, ensuring the accuracy of the recommendation results.

[0061] As an optional embodiment, when obtaining a material recommendation model, multiple groups of sample data can be trained in the following manner to obtain a material recommendation model. First, an initial recommendation model is obtained, and a loss function for model training is constructed, wherein the loss function includes a first loss function term, a second loss function term, and a third loss function term, the first loss function term is used to maintain the relative stability of materials of the same type in a unified feature space, the second loss function term is used to maintain the alignment of different types of materials in a unified feature space, and the third loss function term is used to classify whether the alternative material is the target material. After obtaining the initial recommendation model and the loss function, based on the loss function, multiple groups of sample data are used to train the initial recommendation model to obtain a material recommendation model. Based on the constructed loss function, multiple groups of sample data are used to train the initial recommendation model, which can ensure that the material recommendation model can better obtain a unified representation of heterogeneous materials and improve the accuracy of the recommendation results.

[0062] It should be noted that when constructing the first loss function term, the second loss function term, and the third loss function term, an optional embodiment of the present invention adopts the following method:

[0063] 1) Constructing the first loss function term: obtaining the original material characteristics of materials of the same type in the corresponding feature space, and the unified material characteristics of materials of the same type in the unified feature space; constructing the first loss function term based on the original material characteristics corresponding to materials of the same type, and the unified material characteristics corresponding to materials of the same type; that is, it can ensure the unification of the unified material characteristics of different types of materials in the unified feature space.

[0064] 2) Constructing a second loss function term: obtaining unified material characteristics of different types of materials in a unified feature space; constructing a second loss function based on the unified material characteristics of different types of materials in a unified feature space; that is, being able to ensure the unification of unified material characteristics between different types of materials in a unified feature space.

[0065] 3) Constructing the third loss function term: A classification loss function term is constructed based on the predicted recommendation results and the actual results to ensure the accuracy of the recommendation results.

[0066] Based on the above-mentioned feature mapping function and the corresponding first loss function term of topological structure retention, the second loss function term of feature space alignment, and the third loss function term of classification, the features of different types of materials are well integrated in the same feature space, which more effectively utilizes the various data obtained and produces a method for recommending heterogeneous materials that is superior to other related technologies.

[0067] As an optional embodiment, by receiving a target-object-based material recommendation request sent by a mobile terminal, ultimately determining a recommendation result including multiple target materials recommended to the target object in a predetermined scenario, and feeding back the recommendation result to the mobile terminal, interaction between the cloud and the mobile terminal is achieved. By sending a target-object-based material recommendation request, the mobile terminal provides static attribute data and dynamic interaction data of the target object, receives and extracts the corresponding static attribute features and interaction features in the cloud, and then is able to construct a unified material feature in the cloud, thereby being able to screen among multiple materials, determine a recommendation result including multiple target materials recommended to the target object in a predetermined scenario, and feed back the recommendation result to the mobile terminal, which displays the recommendation result, thus achieving interaction between the cloud and mobile terminal in recommending materials.

[0068] Figure 3 : is a flow chart of the second material recommendation method according to embodiment 1 of the present invention, such as Figure 3 As shown, the method includes the following steps:

[0069] Step S302: Obtain a material recommendation request based on the target object, and send the material recommendation request to the distributed processing cluster;

[0070] Step S304: receiving a recommendation result fed back by the distributed processing cluster, wherein the recommendation result includes multiple target materials recommended to the target object in a predetermined scenario, and the recommendation result is determined based on unified material features corresponding to the multiple candidate materials in the predetermined scenario in a unified feature space, static attribute features of the static attribute data of the target object, and interaction features of the dynamic interaction data of the target object;

[0071] Step S306: Display the recommendation results on the interactive interface.

[0072] Through the above steps, on the mobile side, a material recommendation request based on the target object is obtained on the interactive interface, the material recommendation request is sent to the distributed processing cluster, and the recommendation results fed back by the distributed processing cluster are received and displayed on the interactive interface. The recommendation results include multiple target materials recommended to the target object in a predetermined scenario. The recommendation results are determined based on the unified material features corresponding to the multiple candidate materials in the predetermined scenario in a unified feature space, the static attribute features of the target object's static attribute data, and the interactive features of the target object's dynamic interaction data. Because the recommendation results including multiple target materials are derived based on the unified material features, static attribute features, and interactive features corresponding to the multiple candidate materials, the unified material features are in a unified feature space and can better learn the relationships between heterogeneous materials; the static attribute features are derived based on the target object's static attribute data and are more targeted; and the interactive features are derived based on the target object's dynamic interaction data and can obtain features that match the target object's actual situation. Therefore, the recommendation results including multiple target materials are more accurate, thereby solving the technical problem in related arts of low recommendation accuracy for heterogeneous materials, which leads to a poor user experience.

[0073] As an optional embodiment, interaction between the cloud and the mobile terminal is achieved by obtaining a material recommendation request based on the target object and displaying the recommendation results fed back by the distributed processing cluster (cloud) on the display interface. By sending a material recommendation request based on the target object, the mobile terminal provides the static attribute data and dynamic interaction data of the target object, and the cloud receives and extracts the corresponding static attribute features and interaction features, it can then construct a unified material feature in the cloud, thereby screening among multiple materials, determining the recommendation results including multiple target materials recommended to the target object in a predetermined scenario, and feeding back the recommendation results to the mobile terminal, which displays the recommendation results, thus achieving interaction between the cloud and mobile terminal in recommending materials.

[0074] Based on the above embodiment and optional embodiment, an optional implementation manner is provided, which is described in detail below.

[0075] In related technologies, the approach to handling multiple candidate materials is to model, train, and infer different types of candidate materials separately, and then integrate the recommendation results of these multiple materials through a set of rules. However, simply integrating the multiple materials does not closely relate to each other, resulting in low recommendation accuracy.

[0076] Based on this, in an optional embodiment of the present invention, a material recommendation method is provided. Figure 4 is a flow chart of a material recommendation method according to an optional embodiment of the present invention. Figure 4 As shown, by obtaining unified material features corresponding to multiple candidate materials in a predetermined scenario within a unified feature space, the features corresponding to different types of candidate materials are unified into a single feature space. This solves the problem of loose correlations and low recommendation accuracy caused by different or distributed features of different types of candidate materials. The following describes in detail an optional embodiment of the present invention.

[0077] S1, extracting the original material features of multiple candidate materials in the corresponding feature space, and mapping the original material features corresponding to the multiple candidate materials to a unified feature space, taking the multiple candidate materials as movies, music, and books as an example.

[0078] For multiple alternative materials, we first obtain the original material features R1, R2, and R3 of movies, music, and books in the corresponding feature space, and then pass them into a mapping function f1, f2, and f3 respectively, so that they are transformed into a feature space x.

[0079] It should be noted that the form of the mapping function involved in the optional implementation manner of the present invention is not limited, and a commonly used fully connected neural network can be used.

[0080] S2, obtaining the user's static attribute data and the user's dynamic interaction data, extracting the static attribute features of the static attribute data, and the first interaction features and the second interaction features corresponding to the dynamic interaction data.

[0081] By obtaining user portrait data (the same as the static attribute data mentioned above) and the collected user / material interaction data (the same as the dynamic interaction data mentioned above), the user's static attribute characteristics are generated based on the above user portrait data, and the first interaction characteristics between the user and the interaction data with multiple alternative materials in the dynamic interaction data, as well as the second interaction characteristics between any one of the multiple alternative materials and multiple users related to the material in the scene are determined based on the above user / material interaction data.

[0082] S3, determining a recommendation result including multiple target materials to be recommended to the user in a predetermined scenario based on the unified material characteristics, static attribute characteristics, first interaction characteristics, and second interaction characteristics corresponding to the multiple candidate materials.

[0083] The unified material characteristics, static attribute characteristics and first interactive characteristics are spliced ​​together to obtain the first target characteristics, that is, the user representation is obtained, and the unified material characteristics, static attribute characteristics and second interactive characteristics are spliced ​​together to obtain the second target characteristics, that is, the material representation is obtained; based on the first target characteristics and the second target characteristics, for example, the first target characteristics and the second target characteristics can be passed into the reading layer to determine the recommendation results including multiple target materials recommended to the user in a predetermined scenario.

[0084] It should be noted that the method provided in the optional embodiment of the present invention constructs a loss function including three loss function terms, which can achieve their respective beneficial effects, as described in detail below:

[0085] 1) In order to ensure that the topological structure of the unified material features corresponding to the various alternative materials in the unified feature space remains relatively stable in the new feature space, i.e., the unified feature space, and that the original similar features are still close in distance in the new feature space, a loss function term for retaining the topological structure is constructed in the method provided in the optional embodiment of the present invention. The loss function term is obtained based on the original material features corresponding to the same type of materials and the unified material features corresponding to the same type of materials. Figure 4 It is represented by T in the figure, and T can be implemented in many different ways, such as intra-class distance, CORAL loss, etc.

[0086] 2) In order to ensure that the unified material features corresponding to the various candidate materials in the unified feature space can be represented uniformly in the new feature space, that is, the unified feature space, a loss function term for aligning features is constructed in the method provided in the optional embodiment of the present invention. The loss function term is obtained based on the unified material features of different types of materials in the unified feature space. Figure 4 Indicated by A.

[0087] 3) In order to ensure the accuracy of the recommendation results, in an optional embodiment of the present invention, a classification loss function term is constructed based on the predicted recommendation results and the actual results. Figure 4 Indicated by C.

[0088] The loss function composed of the above-mentioned multiple loss function terms is optimized based on the loss function using various optimization algorithms, such as the gradient descent algorithm, so that the final recommendation result is more accurate.

[0089] Through the above-mentioned optional implementation methods, based on the feature mapping function and the loss function term of the topological structure retention coordinated therewith, the loss function term of the feature space alignment, and the loss function term of the classification, the features of different types of materials are well integrated in the same feature space, and the acquired multiple data are more effectively utilized, achieving the effect of better unifying the features between heterogeneous materials, making the obtained recommendation results more accurate.

[0090] It should be noted that for the aforementioned method embodiments, for simplicity of description, they are all expressed as a series of action combinations. However, those skilled in the art should be aware that the present invention is not limited by the order of the actions described, because according to the present invention, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in this specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present invention.

[0091] Through the description of the above implementation methods, those skilled in the art can clearly understand that the material recommendation method according to the above embodiment can be implemented by means of software plus the necessary general hardware platform. Of course, it can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product. The computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for enabling a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods of various embodiments of the present invention.

[0092] Example 2

[0093] According to an embodiment of the present invention, a device for implementing the above-mentioned material recommendation method is also provided. Figure 5 : is a structural block diagram of a material recommendation device 1 provided according to embodiment 2 of the present invention, such as Figure 5 As shown, the device includes: a first receiving module 502, an extraction module 504, a first acquisition module 506 and a determination module 508. The device is described below.

[0094] The first receiving module 502 is used to receive a material recommendation request based on the target object sent by the mobile terminal, and obtain the static attribute data of the target object and the dynamic interaction data of the target object; the extraction module 504 is connected to the above-mentioned first receiving module 502, and is used to extract the static attribute features of the static attribute data and the interaction features corresponding to the dynamic interaction data; the first acquisition module 506 is connected to the above-mentioned extraction module 504, and is used to obtain the unified material features corresponding to multiple alternative materials in a predetermined scenario in a unified feature space; the determination module 508 is connected to the above-mentioned first acquisition module 506, and is used to determine the recommendation results including multiple target materials recommended to the target object in the predetermined scenario based on the unified material features, static attribute features and interaction features corresponding to the multiple alternative materials, and feed the recommendation results back to the mobile terminal.

[0095] It should be noted here that the above-mentioned first receiving module 502, extraction module 504, first acquisition module 506 and determination module 508 correspond to steps S202 to S208 in Example 1. The instances and application scenarios implemented by multiple modules and corresponding steps are the same, but are not limited to the contents disclosed in the above-mentioned Example 1.

[0096] Example 3

[0097] According to an embodiment of the present invention, a second device for implementing the above-mentioned material recommendation method is also provided. Figure 6 : is a structural block diagram of a second material recommendation device according to embodiment 3 of the present invention. Figure 6 As shown, the device includes: a first sending module 602, a second receiving module 604 and a display module 606. The device is described below.

[0098] The first sending module 602 is used to obtain a material recommendation request based on the target object and send the material recommendation request to the distributed processing cluster; the second receiving module 604 is connected to the above-mentioned first sending module 602, and is used to receive the recommendation results fed back by the distributed processing cluster, wherein the recommendation results include multiple target materials recommended to the target object under a predetermined scenario, and the recommendation results are determined based on the unified material features corresponding to the multiple alternative materials in the predetermined scenario in a unified feature space, the static attribute features of the static attribute data of the target object, and the interactive features of the dynamic interactive data of the target object; the display module 606 is connected to the above-mentioned second receiving module 604, and is used to display the recommendation results on the interactive interface.

[0099] It should be noted here that the above-mentioned first sending module 602, second receiving module 604 and display module 606 correspond to steps S302 to S306 in Example 1. The instances and application scenarios implemented by multiple modules and corresponding steps are the same, but are not limited to the contents disclosed in the above-mentioned Example 1.

[0100] Example 4

[0101] The embodiment of the present invention can provide a computer terminal, which can be any computer terminal device in a computer terminal group. Optionally, in this embodiment, the computer terminal can also be replaced by a terminal device such as a mobile terminal.

[0102] Optionally, in this embodiment, the computer terminal may be located in at least one network device among a plurality of network devices of a computer network.

[0103] In this embodiment, the above-mentioned computer terminal can execute the program code of the following steps in the material recommendation method of the application: receiving a material recommendation request based on the target object sent by the mobile terminal, obtaining the static attribute data of the target object and the dynamic interaction data of the target object; extracting the static attribute features of the static attribute data, and the interaction features corresponding to the dynamic interaction data; obtaining the unified material features corresponding to multiple alternative materials in a unified feature space under a predetermined scenario; based on the unified material features, static attribute features and interaction features corresponding to the multiple alternative materials, determining the recommendation results including multiple target materials recommended to the target object under the predetermined scenario, and feeding back the recommendation results to the mobile terminal.

[0104] Optionally, Figure 7 1 is a block diagram of a computer terminal according to an embodiment of the present invention. Figure 7 As shown, the computer terminal may include: one or more (only one is shown in the figure) processors 71, a memory 72, etc.

[0105] Among them, the memory can be used to store software programs and modules, such as the program instructions / modules corresponding to the material recommendation method and device in the embodiments of the present invention. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, that is, realizing the above-mentioned material recommendation method. The memory may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include a memory remotely located relative to the processor, and these remote memories may be connected to the terminal via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0106] The processor can call the information and application stored in the memory through the transmission device to perform the following steps: receive a material recommendation request based on the target object sent by the mobile terminal, obtain the static attribute data of the target object and the dynamic interaction data of the target object; extract the static attribute features of the static attribute data, and the interaction features corresponding to the dynamic interaction data; obtain the unified material features corresponding to multiple alternative materials in the unified feature space under the predetermined scenario; based on the unified material features, static attribute features and interaction features corresponding to the multiple alternative materials, determine the recommendation results including multiple target materials recommended to the target object under the predetermined scenario, and feed back the recommendation results to the mobile terminal.

[0107] Optionally, the processor may also execute the program code of the following steps: obtaining unified material features corresponding to a plurality of alternative materials in a predetermined scenario in a unified feature space, including: extracting original material features of a plurality of alternative materials in a corresponding feature space; mapping the original material features corresponding to a plurality of alternative materials to a unified feature space according to a spatial mapping function, and obtaining unified material features corresponding to a plurality of alternative materials in a predetermined scenario in a unified feature space.

[0108] Optionally, the processor may also execute the program code of the following steps: determining the recommendation results including multiple target materials recommended to the target object in a predetermined scenario based on the unified material characteristics, static attribute characteristics and interaction characteristics corresponding to the multiple alternative materials, and feeding back the recommendation results to the mobile terminal, including: determining the first interaction characteristics between the target object and the selected material in the dynamic interaction data, and the second interaction characteristics between the selected material and the object based on the interaction characteristics; splicing the unified material characteristics, static attribute characteristics and the first interaction characteristics to obtain the first target characteristics, and splicing the unified material characteristics, static attribute characteristics and the second interaction characteristics to obtain the second target characteristics; determining the recommendation results including multiple target materials recommended to the target object in a predetermined scenario based on the first target characteristics and the second target characteristics, and feeding back the recommendation results to the mobile terminal.

[0109] Optionally, the processor may also execute the program code of the following steps: extracting static attribute features of static attribute data and interaction features corresponding to dynamic interaction data, including: using the feature extraction module of the material recommendation module to extract static attribute features of static attribute data and interaction features corresponding to dynamic interaction data; determining the recommendation results including multiple target materials recommended to the target object in a predetermined scenario based on the unified material features, static attribute features and interaction features corresponding to a plurality of alternative materials, and feeding back the recommendation results to the mobile terminal, including: inputting the unified material features, static attribute features and interaction features corresponding to a plurality of alternative materials into the recommendation module of the material recommendation model, obtaining the recommendation results including multiple target materials recommended to the target object in a predetermined scenario, and feeding back the recommendation results to the mobile terminal; wherein, the material recommendation model is trained using multiple sets of sample data, and the multiple sets of sample data include: static attribute data of the sample object, dynamic interaction data of the sample object, and sample recommendation results recommended to the sample object.

[0110] Optionally, the processor may also execute the program code of the following steps: the method also includes: training multiple groups of sample data in the following manner to obtain a material recommendation model: obtaining an initial recommendation model, and constructing a loss function for model training, wherein the loss function includes a first loss function term, a second loss function term, and a third loss function term, the first loss function term is used to maintain the relative stability of materials of the same type in a unified feature space, the second loss function term is used to maintain the alignment of materials of different types in a unified feature space, and the third loss function term is used to classify whether the alternative material is the target material; based on the loss function, multiple groups of sample data are used to train the initial recommendation model to obtain a material recommendation model.

[0111] Optionally, the processor may also execute the program code of the following steps: the method also includes: constructing a first loss function term in the following manner: obtaining the original material characteristics of materials of the same type in the corresponding feature space, and the unified material characteristics of materials of the same type in the unified feature space; constructing a first loss function term based on the original material characteristics corresponding to materials of the same type, and the unified material characteristics corresponding to materials of the same type; constructing a second loss function term in the following manner: obtaining the unified material characteristics of different types of materials in the unified feature space; constructing a second loss function based on the unified material characteristics of different types of materials in the unified feature space.

[0112] The processor can call the information and application stored in the memory through the transmission device to perform the following steps: obtain a material recommendation request based on the target object, and send the material recommendation request to the distributed processing cluster; receive the recommendation results fed back by the distributed processing cluster, wherein the recommendation results include multiple target materials recommended to the target object under the predetermined scenario, and the recommendation results are determined based on the unified material characteristics corresponding to the multiple alternative materials in the predetermined scenario in the unified feature space, the static attribute characteristics of the static attribute data of the target object, and the interactive characteristics of the dynamic interactive data of the target object; and display the recommendation results on the interactive interface.

[0113] It can be understood by those skilled in the art that Figure 7 The structure shown is for illustration only, and the computer terminal may also be a smart phone (such as an Android phone, an iOS phone, etc.), a tablet computer, a handheld computer, a mobile Internet device (MID), a PAD, or other terminal devices. Figure 7 It does not limit the structure of the above electronic device. For example, the computer terminal 7 may also include Figure 7 More or fewer components (such as network interfaces, display devices, etc.) shown in, or with Figure 7 Different configurations shown.

[0114] A person skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing the hardware related to the terminal device through a program, and the program can be stored in a computer-readable storage medium, which may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0115] Example 4

[0116] The embodiment of the present invention further provides a computer-readable storage medium. Optionally, in this embodiment, the computer-readable storage medium can be used to store the program code executed by the material recommendation method provided in the embodiment 1 above.

[0117] Optionally, in this embodiment, the computer-readable storage medium may be located in any computer terminal in a computer terminal group in a computer network, or in any mobile terminal in a mobile terminal group.

[0118] Optionally, in this embodiment, the computer-readable storage medium is configured to store program code for executing the following steps: receiving a material recommendation request based on a target object sent by a mobile terminal, obtaining static attribute data of the target object and dynamic interaction data of the target object; extracting static attribute features of the static attribute data, and interaction features corresponding to the dynamic interaction data; obtaining unified material features corresponding to a plurality of alternative materials in a unified feature space under a predetermined scenario; determining a recommendation result including a plurality of target materials recommended to the target object under a predetermined scenario based on the unified material features, static attribute features and interaction features corresponding to the plurality of alternative materials, and feeding back the recommendation result to the mobile terminal.

[0119] Optionally, in this embodiment, the computer-readable storage medium is configured to store program code for executing the following steps: obtaining unified material characteristics corresponding to multiple alternative materials in a predetermined scenario in a unified feature space, including: extracting original material characteristics of multiple alternative materials in the corresponding feature space; mapping the original material characteristics corresponding to the multiple alternative materials to the unified feature space based on the space mapping function, and obtaining unified material characteristics corresponding to the multiple alternative materials in the predetermined scenario in the unified feature space.

[0120] Optionally, in this embodiment, the computer-readable storage medium is configured to store program code for executing the following steps: determining a recommendation result including multiple target materials recommended to the target object in a predetermined scenario based on the unified material characteristics, static attribute characteristics and interaction characteristics corresponding to the multiple alternative materials, and feeding back the recommendation result to the mobile terminal, including: based on the interaction characteristics, determining a first interaction feature between the target object and the selected material in the dynamic interaction data, and a second interaction feature between the selected material and the object; splicing the unified material characteristics, the static attribute characteristics and the first interaction characteristics to obtain a first target feature, and splicing the unified material characteristics, the static attribute characteristics and the second interaction characteristics to obtain a second target feature; based on the first target feature and the second target feature, determining a recommendation result including multiple target materials recommended to the target object in a predetermined scenario, and feeding back the recommendation result to the mobile terminal.

[0121] Optionally, in this embodiment, the computer-readable storage medium is configured to store program code for executing the following steps: extracting static attribute features of static attribute data and interaction features corresponding to dynamic interaction data, including: using the feature extraction module of the material recommendation module to extract static attribute features of static attribute data and interaction features corresponding to dynamic interaction data; determining a recommendation result including multiple target materials recommended to the target object in a predetermined scenario based on the unified material features, static attribute features and interaction features corresponding to a plurality of alternative materials, and feeding back the recommendation result to the mobile terminal, including: inputting the unified material features, static attribute features and interaction features corresponding to a plurality of alternative materials into the recommendation module of the material recommendation model, obtaining a recommendation result including multiple target materials recommended to the target object in a predetermined scenario, and feeding back the recommendation result to the mobile terminal; wherein, the material recommendation model is trained using multiple groups of sample data, and the multiple groups of sample data include: static attribute data of the sample object, dynamic interaction data of the sample object, and sample recommendation results recommended to the sample object.

[0122] Optionally, in this embodiment, the computer-readable storage medium is configured to store program code for executing the following steps: training multiple groups of sample data in the following manner to obtain a material recommendation model: obtaining an initial recommendation model, and constructing a loss function for model training, wherein the loss function includes a first loss function term, a second loss function term, and a third loss function term, the first loss function term is used to maintain the relative stability of materials of the same type in a unified feature space, the second loss function term is used to maintain the alignment of materials of different types in a unified feature space, and the third loss function term is used to classify whether the alternative material is the target material; based on the loss function, the initial recommendation model is trained using multiple groups of sample data to obtain a material recommendation model.

[0123] Optionally, in this embodiment, the computer-readable storage medium is configured to store program code for executing the following steps: constructing a first loss function term by obtaining the original material characteristics of materials of the same type in the corresponding feature space, and the unified material characteristics of materials of the same type in the unified feature space; constructing a first loss function term based on the original material characteristics corresponding to materials of the same type, and the unified material characteristics corresponding to materials of the same type; constructing a second loss function term by obtaining the unified material characteristics of different types of materials in the unified feature space; constructing a second loss function based on the unified material characteristics of different types of materials in the unified feature space.

[0124] Optionally, in this embodiment, the computer-readable storage medium is configured to store program code for executing the following steps: obtaining a material recommendation request based on the target object, and sending the material recommendation request to a distributed processing cluster; receiving the recommendation results fed back by the distributed processing cluster, wherein the recommendation results include a plurality of target materials recommended to the target object under a predetermined scenario, and the recommendation results are determined based on the unified material features corresponding to the plurality of alternative materials in the predetermined scenario in a unified feature space, the static attribute features of the static attribute data of the target object, and the interactive features of the dynamic interactive data of the target object; and displaying the recommendation results on an interactive interface.

[0125] The serial numbers of the above embodiments of the present invention are for description only and do not represent the advantages or disadvantages of the embodiments.

[0126] In the above embodiments of the present invention, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0127] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0128] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0129] In addition, the functional units in the various embodiments of the present invention may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0130] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a computer-readable storage medium, including a number of instructions for enabling a computer device (which can be a personal computer, server or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned computer-readable storage medium includes: various media that can store program codes, such as a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk.

[0131] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. A material recommendation method, characterized in that: include: Receive a material recommendation request based on a target object sent by a mobile terminal, and obtain static attribute data of the target object and dynamic interaction data of the target object; Extracting static attribute features of the static attribute data and interaction features corresponding to the dynamic interaction data; Obtaining unified material features corresponding to multiple candidate materials in a predetermined scenario in a unified feature space, wherein the unified material feature is a feature obtained by unifying the features corresponding to the multiple candidate materials into one feature space; Determining a recommendation result including multiple target materials recommended to the target object in the predetermined scenario based on the unified material characteristics corresponding to the multiple candidate materials, the static attribute characteristics, and the interactive characteristics, and feeding back the recommendation result to the mobile terminal; Among them, obtaining the unified material characteristics corresponding to multiple alternative materials in a predetermined scenario in a unified feature space includes: extracting the original material characteristics of the multiple alternative materials in the corresponding feature space; mapping the original material characteristics corresponding to the multiple alternative materials to the unified feature space according to a space mapping function, and obtaining the unified material characteristics corresponding to the multiple alternative materials in the predetermined scenario in the unified feature space.

2. The method according to claim 1, characterized in that Determining, based on the unified material features, the static attribute features, and the interactive features respectively corresponding to the multiple candidate materials, a recommendation result including multiple target materials recommended to the target object in the predetermined scenario, and feeding the recommendation result back to the mobile terminal, including: Based on the interaction characteristics, determining a first interaction characteristic between the target object and a selected material in the dynamic interaction data, and a second interaction characteristic between the selected material and the object; splicing the unified material feature, the static attribute feature, and the first interactive feature to obtain a first target feature, and splicing the unified material feature, the static attribute feature, and the second interactive feature to obtain a second target feature; Based on the first target feature and the second target feature, a recommendation result including multiple target materials recommended to the target object in the predetermined scenario is determined, and the recommendation result is fed back to the mobile terminal.

3. The method according to claim 1, characterized in that The extracting of the static attribute features of the static attribute data and the interaction features corresponding to the dynamic interaction data comprises: extracting the static attribute features of the static attribute data and the interaction features corresponding to the dynamic interaction data using a feature extraction module of a material recommendation module; The determining, based on the unified material characteristics, the static attribute characteristics, and the interactive characteristics respectively corresponding to the multiple candidate materials, a recommendation result including multiple target materials recommended to the target object in the predetermined scenario, and feeding the recommendation result back to the mobile terminal, includes: inputting the unified material characteristics, the static attribute characteristics, and the interactive characteristics respectively corresponding to the multiple candidate materials into a recommendation module of a material recommendation model, obtaining a recommendation result including multiple target materials recommended to the target object in the predetermined scenario, and feeding the recommendation result back to the mobile terminal; The material recommendation model is trained using multiple sets of sample data, and the multiple sets of sample data include: static attribute data of sample objects, dynamic interaction data of the sample objects, and sample recommendation results recommended to the sample objects.

4. The method according to claim 3, characterized in that The method further includes: training the multiple groups of sample data in the following manner to obtain the material recommendation model: Obtaining an initial recommendation model and constructing a loss function for model training, wherein the loss function includes a first loss function term, a second loss function term, and a third loss function term, wherein the first loss function term is used to maintain the relative stability of materials of the same type in the unified feature space, the second loss function term is used to maintain the alignment of materials of different types in the unified feature space, and the third loss function term is used to classify whether an alternative material is a target material; Based on the loss function, the initial recommendation model is trained using the multiple groups of sample data to obtain the material recommendation model.

5. The method according to claim 4, characterized in that The method further comprises: The first loss function term is constructed by: obtaining original material features of the materials of the same type in corresponding feature spaces, and unified material features of the materials of the same type in the unified feature space; constructing the first loss function term based on the original material features corresponding to the materials of the same type, and the unified material features corresponding to the materials of the same type; The second loss function term is constructed in the following manner: obtaining unified material characteristics of different types of materials in the unified feature space; and constructing the second loss function based on the unified material characteristics of the different types of materials in the unified feature space.

6. A material recommendation method, characterized in that: include: Obtaining a material recommendation request based on a target object, and sending the material recommendation request to a distributed processing cluster; Receive a recommendation result fed back by the distributed processing cluster, wherein the recommendation result includes multiple target materials recommended to the target object in a predetermined scenario, the recommendation result is determined based on unified material features corresponding to the multiple candidate materials in a unified feature space in the predetermined scenario, static attribute features of the static attribute data of the target object, and interaction features of the dynamic interaction data of the target object, the unified material features are features obtained by unifying features corresponding to the multiple candidate materials into a feature space, the unified material features are obtained by mapping original material features corresponding to the multiple candidate materials to the unified feature space based on a space mapping function, and the original material features are extracted features of the multiple candidate materials in the corresponding feature space; The recommendation results are displayed on the interactive interface.

7. A material recommendation device, characterized in that: include: A first receiving module is configured to receive a material recommendation request based on a target object sent by a mobile terminal, and obtain static attribute data and dynamic interaction data of the target object; An extraction module, configured to extract static attribute features of the static attribute data and interaction features corresponding to the dynamic interaction data; A first acquisition module is configured to acquire unified material features corresponding to a plurality of candidate materials in a predetermined scenario in a unified feature space, wherein the unified material features are features obtained by unifying features corresponding to the plurality of candidate materials into one feature space; a determination module, configured to determine a recommendation result including multiple target materials to be recommended to the target object in the predetermined scenario based on the unified material characteristics corresponding to the multiple candidate materials, the static attribute characteristics, and the interaction characteristics, and to feed back the recommendation result to the mobile terminal; Among them, the first acquisition module is also used to extract the original material features of the multiple alternative materials in the corresponding feature space; according to the space mapping function, the original material features corresponding to the multiple alternative materials are mapped to the unified feature space to obtain the unified material features corresponding to the multiple alternative materials in the predetermined scenario in the unified feature space.

8. A material recommendation device, characterized in that: include: A first sending module is configured to obtain a material recommendation request based on a target object and send the material recommendation request to a distributed processing cluster; a second receiving module, configured to receive a recommendation result fed back by the distributed processing cluster, wherein the recommendation result includes multiple target materials recommended to the target object in a predetermined scenario, the recommendation result is determined based on unified material features corresponding to the multiple candidate materials in a unified feature space in the predetermined scenario, static attribute features of the static attribute data of the target object, and interaction features of the dynamic interaction data of the target object, the unified material features being features obtained by unifying features corresponding to the multiple candidate materials into a feature space, the unified material features being obtained by mapping original material features corresponding to the multiple candidate materials into the unified feature space based on a space mapping function, and the original material features being features extracted from the multiple candidate materials in the corresponding feature space; The display module is used to display the recommendation results on the interactive interface.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute the material recommendation method according to any one of claims 1 to 6.

10. A computer device, characterized in that: include: memory and processor, The memory stores a computer program; The processor is configured to execute a computer program stored in the memory, and when the computer program is run, the processor is enabled to execute the material recommendation method according to any one of claims 1 to 6.

Images