Article recommendation method, computing device, electronic device and storage medium

Abstract

The invention discloses an article recommendation method, computing equipment, electronic equipment and a storage medium, and relates to the field of artificial intelligence technology and data processing. The method comprises the following steps: in response to a received article recommendation request, performing feature extraction on object information of a request object to obtain object features, and performing feature extraction on at least one to-be-recommended article to obtain initial features of the at least one to-be-recommended article; performing feature enhancement on the initial feature of any one to-be-recommended article based on the independent feature of any one to-be-recommended article to obtain a target feature of any one to-be-recommended article; determining a target article from the at least one to-be-recommended article based on the object feature and a target feature of the at least one to-be-recommended article; and outputting the target article. The technical problem that articles recommended to users cannot meet user preferences in related technologies is solved.

Description

Technical Field

[0001] The present application relates to artificial intelligence technology and data processing fields, and specifically, to an item recommendation method, a computing device, an electronic device and a storage medium. Background Art

[0002] With the rapid development of the Internet and e-commerce, personalized item recommendation systems have become a key technology to improve user experience and commercial value. At present, item recommendation systems can usually recommend products or content that can meet user needs by analyzing user historical behaviors, recommendation requests and other content. However, in the actual item process, the items viewed by users can usually only meet the user needs contained in the recommendation request, but cannot meet the user's interest in different items.

[0003] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0004] The embodiments of the present application provide an item recommendation method, a computing device, an electronic device, and a storage medium to at least solve the technical problem in the related art that the items recommended to the user cannot meet the user's preferences.

[0005] According to one aspect of an embodiment of the present application, a method for recommending an item is provided, comprising: in response to receiving an item recommendation request, performing feature extraction on object information of a request object to obtain object features, and performing feature extraction on at least one item to be recommended to obtain initial features of at least one item to be recommended, wherein the request object is used to characterize the object that sends the item recommendation request; performing feature enhancement on the initial features of any item to be recommended based on the independent features of any item to be recommended to obtain target features of any item to be recommended, wherein the independent features are used to characterize the interest preferences of a search object, and the search object is used to search for any item to be recommended; determining a target item from at least one item to be recommended based on the object features and the target features of at least one item to be recommended; and outputting the target item.

[0006] According to one aspect of an embodiment of the present application, a method for recommending an item is also provided, comprising: responding to an input instruction on an operation interface, displaying on the operation interface the initial features of at least one item to be recommended and the object features of a requesting object, wherein the requesting object is used to characterize the object that sends the item recommendation request; responding to a processing instruction on the operation interface, displaying a target item on the operation interface, wherein the target item is determined from at least one item to be recommended by an object feature and a target feature of at least one item to be recommended, the target feature is obtained by enhancing the initial features of any item to be recommended by an independent feature of any item to be recommended, the independent feature is used to characterize the interest preference of a search object, and the search object is used to search for any item to be recommended.

[0007] According to one aspect of an embodiment of the present application, a method for recommending an item is also provided, comprising: obtaining object characteristics of a request object and initial characteristics of at least one item to be recommended by calling a first interface, wherein the request object is used to characterize the object that sends the item recommendation request, and the first interface includes a first parameter, and the parameter value of the first parameter includes the object characteristics and the initial characteristics; based on the independent characteristics of any item to be recommended, the initial characteristics of any item to be recommended are enhanced to obtain the target characteristics of any item to be recommended, wherein the independent characteristics are used to characterize the interest preferences of the search object, and the search object is used to search for any item to be recommended; based on the object characteristics and the target characteristics of at least one item to be recommended, determining the target item from at least one item to be recommended; and outputting the target item by calling a second interface, wherein the second interface includes a second parameter, and the parameter value of the second parameter includes the target item.

[0008] According to one aspect of an embodiment of the present application, there is provided an item recommendation device, comprising: a feature extraction module, for, in response to receiving an item recommendation request, performing feature extraction on object information of a request object to obtain object features, and performing feature extraction on at least one item to be recommended to obtain initial features of at least one item to be recommended, wherein the request object is used to characterize the object that sends the item recommendation request; a feature enhancement module, for performing feature enhancement on the initial features of any item to be recommended based on the independent features of any item to be recommended to obtain target features of any item to be recommended, wherein the independent features are used to characterize the interest preferences of a search object, and the search object is used to search for any item to be recommended; an item determination module, for determining a target item from at least one item to be recommended based on the object features and the target features of at least one item to be recommended; and an item output module, for outputting the target item.

[0009] According to one aspect of an embodiment of the present application, there is also provided an item recommendation device, comprising: a first display module, for responding to an input instruction acting on an operation interface, and displaying on the operation interface the initial features of at least one item to be recommended, and the object features of a requesting object, wherein the requesting object is used to characterize the object that sends the item recommendation request; a second display module, for responding to a processing instruction acting on the operation interface, and displaying a target item on the operation interface, wherein the target item is determined from at least one item to be recommended by an object feature and a target feature of at least one item to be recommended, and the target feature is obtained by enhancing the initial features of any item to be recommended by an independent feature of any item to be recommended, and the independent feature is used to characterize the interest preference of a search object, and the search object is used to search for any item to be recommended.

[0010] According to one aspect of an embodiment of the present application, there is also provided an item recommendation device, comprising: a first calling module, used to obtain object characteristics of a request object and initial characteristics of at least one item to be recommended by calling a first interface, wherein the request object is used to characterize the object that sends the item recommendation request, and the first interface includes a first parameter, and the parameter value of the first parameter includes the object characteristics and the initial characteristics; an enhancement module, used to perform feature enhancement on the initial characteristics of any item to be recommended based on the independent characteristics of any item to be recommended, to obtain a target characteristic of any item to be recommended, wherein the independent characteristic is used to characterize the interest preference of a search object, and the search object is used to search for any item to be recommended; a determination module, used to determine a target item from at least one item to be recommended based on the object characteristics and the target characteristics of at least one item to be recommended; a second calling module, used to output the target item by calling a second interface, wherein the second interface includes a second parameter, and the parameter value of the second parameter includes the target item.

[0011] According to another aspect of the embodiments of the present application, a computing device is further provided, including: a memory storing an executable program; and a processor for running the program, wherein the method in each embodiment of the present application is executed when the program is running.

[0012] According to another aspect of the embodiments of the present application, an electronic device is also provided, including: a memory storing an executable program; a processor connected to the memory via a bus and used to run the program, wherein the method of each embodiment of the present application is executed when the program is running.

[0013] According to another aspect of an embodiment of the present application, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored executable program, wherein when the executable program is running, the device where the computer-readable storage medium is located is controlled to execute the methods in each embodiment of the present application.

[0014] According to another aspect of the embodiments of the present application, a computer program product is also provided, including a computer program, and when the computer program is executed by a processor, the method in each embodiment of the present application is implemented.

[0015] According to another aspect of an embodiment of the present application, a computer program product is also provided, including a non-volatile computer-readable storage medium, wherein the non-volatile computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method in each embodiment of the present application is implemented.

[0016] According to another aspect of the embodiments of the present application, a computer program is also provided, and when the computer program is executed by a processor, the methods in the various embodiments of the present application are implemented.

[0017] In an embodiment of the present application, in response to receiving an item recommendation request, feature extraction is performed on the object information of the request object to obtain object features, and feature extraction is performed on at least one item to be recommended to obtain initial features of at least one item to be recommended; feature enhancement is performed on the initial features of any item to be recommended based on the independent features of any item to be recommended to obtain target features of any item to be recommended; based on the object features and the target features of at least one item to be recommended, a target item is determined from at least one item to be recommended; a method of outputting the target item is used to enhance the initial features of the item to be recommended by using the independent features of the item to be recommended to obtain target features for determining the user's interest in the recommended item, and the target item is determined by using the target features and the object features of the request object, which can greatly improve the degree of match between the determined target item and the user's preferences, thereby improving the user's experience of using the item recommendation system, and further solving the technical problem in the related art that the items recommended to the user cannot meet the user's preferences.

[0018] It is easy to notice that the above general description and the following detailed description are only for the purpose of exemplifying and explaining the present application, and do not constitute a limitation of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0020] Figure 1 is a schematic diagram of an application scenario of an item recommendation method according to an embodiment of the present application;

[0021] Figure 2 is a flow chart of an item recommendation method according to an embodiment of the present application;

[0022] Figure 3 is a schematic diagram of an item search process according to an embodiment of the present application;

[0023] Figure 4 is a schematic diagram of an item recommendation process according to an embodiment of the present application;

[0024] Figure 5 is a flow chart of another item recommendation method according to an embodiment of the present application;

[0025] Figure 6 is a flow chart of another item recommendation method according to an embodiment of the present application;

[0026] Figure 7 is a structural block diagram of an item recommendation device according to an embodiment of the present application;

[0027] Figure 8 is a structural block diagram of another item recommendation device according to an embodiment of the present application;

[0028] Fig. 9 is a structural block diagram of another item recommendation device according to an embodiment of the present application;

[0029] Fig.10 is a structural block diagram of a computing device according to an embodiment of the present application;

[0030] Fig.11 It is a structural block diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0031] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0032] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application 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 data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising 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.

[0033] First, some nouns or terms that appear in the description of the embodiments of the present application are subject to the following explanations:

[0034] Query: In some recommendation systems, users initiate search terms or images to query items based on these search terms or images.

[0035] Features: Data that is fed into the model and can be used to make recommendations using the model.

[0036] Counterfactual learning: A technique for exploring the effects of a quantitative intervention on a target before and after the intervention.

[0037] According to an embodiment of the present application, a method for recommending an item is 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.

[0038] Considering that the model parameters of the model used in the above-mentioned item recommendation method may be very large, and the computing resources of the mobile terminal are limited, the above-mentioned method provided in the embodiment of the present application can be applied to Figure 1 The application scenarios shown are not limited to these. Figure 1 is a schematic diagram of an application scenario of an item recommendation method according to an embodiment of the present application. Figure 1In the application scenario shown, the model used in the above-mentioned item recommendation method can be deployed in the server 10, and the server 10 can be connected to one or more client devices 20 through a local area network connection, a wide area network connection, an Internet connection, or other types of data networks. The client device 20 here can include but is not limited to: smart phones, tablet computers, laptops, PDAs, personal computers, smart home devices, car-mounted devices, etc. The client device 20 can interact with the user through a graphical user interface to call the model, thereby implementing the method provided in the embodiment of the present application.

[0039] In the embodiment of the present application, the system composed of the client device and the server can perform the following steps: the client device performs input of an item recommendation request; outputs a target item. The server performs feature extraction of the object information of the request object to obtain object features, and performs feature extraction of at least one item to be recommended to obtain corresponding initial features; based on the independent features of any item to be recommended, the initial features of any item to be recommended are feature enhanced to obtain the target features of any item to be recommended; based on the object features and the target features of at least one item to be recommended, the target item is determined from at least one item to be recommended.

[0040] It should be noted that, with the rapid development of high-performance computing units, in other application scenarios, the above method provided in the embodiment of the present application can also be applied to the model-in-one machine. In an optional embodiment, a plurality of models are built into the model-in-one machine, and the user can choose to adjust a model as needed to obtain the user's own model, so that the high-performance computing unit built into the model-in-one machine can directly call the adjusted model to execute the above method provided in the embodiment of the present application. In another optional embodiment, a trained model is built into the model-in-one machine, so that the high-performance computing unit built into the model-in-one machine can directly call the model to execute the above method provided in the embodiment of the present application.

[0041] Furthermore, when users need to train their own models, they can also upload their own data sets through the client, which are sent by the client to the server, so that the server can adjust the pre-trained model with the data set to obtain the user's own model and then deploy it to the production environment. In order to facilitate users' needs for model adjustment, the server can provide complete adjustment tools, development frameworks and processes, and can support multiple adjustment strategies, so that the adjusted model can better adapt to applications in different fields and achieve a high degree of customization.

[0042] Under the above operating environment, this application provides Figure 2 The recommended method for the items shown. Figure 2 FIG. 1 is a flowchart of an item recommendation method according to an embodiment of the present application. Figure 2 As shown, the method may include the following steps:

[0043] Step S202, in response to receiving an item recommendation request, extracting features from the object information of the request object to obtain object features, and extracting features from at least one item to be recommended to obtain initial features of at least one item to be recommended.

[0044] The request object is used to represent the object that sends the item recommendation request.

[0045] The request object may refer to other servers that have a communication relationship with the item recommendation system, or a user who is currently using the item recommendation system. The following takes the user as an example for explanation. The items to be recommended may refer to items searched from the database according to the request object, such as the needs of the user, and may include but are not limited to: physical goods in real scenes such as clothing, vehicles, and furniture, and virtual goods in virtual environments such as digital collections, accounts, and online courses.

[0046] In an optional scheme of the present embodiment, when the current item recommendation system recommends items to users, it can first clarify the user's needs based on the user's search behavior, and search for the above-mentioned items to be recommended from a database storing different items according to the user's needs, and then integrate the relevant information of the searched items to be recommended, such as the attribute characteristics of the items to be recommended, the search data generated when searching for the items to be recommended, and other information, and display the above-mentioned items to be recommended to the user based on the integrated information, so that the items to be recommended finally displayed to the user can meet the user's needs. Although the item recommendation system takes into account the user's relevant information, such as the behavioral characteristics corresponding to the user's search behavior, and displays the recommended items to the user based on the behavioral characteristics to ensure the final item display effect, it should be noted that when users view different items, in addition to considering their own objective needs for different items, they will also have certain subjective ideas, such as the degree of interest in different items. As shown above, the current item recommendation system only applies the behavioral characteristics corresponding to the user's search behavior to the item recommendation process, while ignoring the difference between the user's search intention and general interest. For example, assuming that the user has a need for items A and B, but is not interested in item A, then when selecting items, the user may not choose item A. At this time, if the item recommendation system displays item A to the user first and then displays item B to the user, it is very likely to affect the user's experience of using the item recommendation system.

[0047] Therefore, when actually recommending the searched items to be recommended to the user, the item recommendation system can additionally introduce the user's interest level in different items to ensure that when the items are finally recommended to the user, the user's needs can be met while improving the user's experience. In order to accurately determine the user's interest level in different items, the item recommendation system can, after receiving the item recommendation request, first obtain the object information of the requested object and the item information of at least one item to be recommended that is currently prepared to be displayed to the user, and then perform feature extraction on the obtained object information to obtain the object features of the corresponding requested object, and at the same time perform feature extraction on the item information to obtain the corresponding initial features of at least one item to be recommended. Based on the object features of the requested object and the initial features of different items to be recommended, the recommendation system can determine the association between the requested object and the different items to be recommended, thereby determining the request object's interest level in the different items to be recommended.

[0048] Step S204: enhancing the initial features of any item to be recommended based on the independent features of any item to be recommended, to obtain the target features of any item to be recommended.

[0049] Among them, the independent features are used to characterize the interest preferences of the search object, and the search object is used to search for any item to be recommended.

[0050] The search object may refer to an object that performs keyword search, item query, etc., and may be the same as the request object, for example, the user who sends the item recommendation request. The independent feature may refer to a feature related to the user's interest level, which may reflect the interest preference of the search object. By changing the independent feature of the item, the user's interest level in the item may be changed accordingly.

[0051] In an optional solution of this embodiment, considering that under normal circumstances, the characteristics of an item can be divided into two parts: dynamic characteristics and static characteristics. Among them, the dynamic characteristics can also be called query characteristics, which can refer to characteristics related to the user's query behavior and can be applied to the process of searching for the above-mentioned recommended items. They have certain subjectivity and immediacy. For example, assuming that the keyword entered by the user is "metal glasses frame", the above-mentioned dynamic characteristics can be the two characteristics of "metal" and "glasses frame". If the keyword entered by the user is "plastic glasses frame", the above-mentioned dynamic characteristics can be the two characteristics of "plastic" and "glasses frame". The dynamic characteristics can change with the change of the user's query behavior. The item recommendation system can judge whether the item can meet the user's needs based on the dynamic characteristics of the item to conduct item search; static characteristics can also be used to determine whether the item can meet the user's needs based on the dynamic characteristics of the item. The so-called independent features mentioned above may refer to the inherent features of the item itself, which have a certain objectivity. Static features will not change with the change of user query behavior. Different static features may have different appeal to users, and the corresponding user's interest in an item will be affected by the static features of the item. For example, assuming that the keyword entered by the user is "metal glasses frame", the item recommendation system can search for multiple metal glasses frames based on the keyword, but the static features of these glasses frames, such as color, texture, style and other features may be different, and the corresponding users may also have different preferences for these static features. Through these static features, it can be determined that the metal glasses frames that users are more likely to choose, so that the items finally recommended to users can meet the user's expectations well.

[0052] Based on this, in order to better recommend different items to users, after extracting the initial features of different items to be recommended, the item recommendation system can further obtain independent features of different items to be recommended, and use the independent features to enhance the extracted initial features to obtain a rich and informative item representation, that is, to obtain the above-mentioned target features, so that the item recommendation system can provide users with more accurate item recommendation results based on the target features of different items to be recommended.

[0053] Step S206: determining a target item from the at least one item to be recommended based on the object feature and the target feature of the at least one item to be recommended.

[0054] In an optional scheme of the present embodiment, after obtaining the target features of different items to be recommended, considering that the object features of the request object and the target features of the items to be recommended can both reflect the user's interest in the different items to be recommended, therefore, in order to be able to recommend items to users more accurately, the item recommendation system can combine the extracted object features and the acquired target features to judge the user's interest in the different items to be recommended based on the object features and the target features, so as to ensure the accuracy of the determined interest level, and then determine the target item that will be liked by the user from the at least one item to be recommended based on the determined interest level, so as to ensure the accuracy of the determined target item.

[0055] Step S208, outputting the target item.

[0056] In an optional solution of this embodiment, after the target item is determined, the item recommendation system can output the target item in the operation interface for the convenience of the user to view. In order to ensure the user's experience when viewing the target item, the item recommendation system can first display the items with high user interest in the target frequently browsed area, and display the items with low user interest in other areas, as well as other items in the recommended items except the determined target item. Alternatively, considering that the target item is an item that can meet the user's preferences, in order to improve the efficiency of displaying items to the user, the item recommendation system can first display the target item in the operation interface, and then display other items in the recommended items except the target item to the user when the user has the need to continue viewing the items.

[0057] In an embodiment of the present application, in response to receiving an item recommendation request, feature extraction is performed on the object information of the request object to obtain object features, and feature extraction is performed on at least one item to be recommended to obtain initial features of at least one item to be recommended; feature enhancement is performed on the initial features of any item to be recommended based on the independent features of any item to be recommended to obtain target features of any item to be recommended; based on the object features and the target features of at least one item to be recommended, a target item is determined from at least one item to be recommended; a method of outputting the target item is used to enhance the initial features of the item to be recommended by using the independent features of the item to be recommended to obtain target features for determining the user's interest in the recommended item, and the target item is determined by using the target features and the object features of the request object, which can greatly improve the degree of match between the determined target item and the user's preferences, thereby improving the user's experience of using the item recommendation system, and further solving the technical problem in the related art that the items recommended to the user cannot meet the user's preferences.

[0058] In an embodiment of the present application, before enhancing the initial features of any item to be recommended based on the independent features of any item to be recommended, the method further includes: extracting features from the item recommendation request to obtain request features of the item recommendation request; constructing a feature evaluation score corresponding to any recommended item based on the initial features, request features and object features, wherein the feature evaluation score is used to reflect the association relationship between any item to be recommended, the item recommendation request and the request object; inputting the initial features and the feature evaluation scores into a feature evaluation model, and using the feature evaluation model to determine independent features from the initial features.

[0059] In an optional solution of this embodiment, in order to accurately enhance the initial features, the item recommendation system may first determine the corresponding independent features from the initial features of the item to be recommended before using the independent features to enhance the initial features. Considering that the independent features of an item are usually not affected by the user's query behavior, and when the independent features of an item change, the user's interest in the item will change accordingly, the item recommendation system may establish a mapping curve between object features, initial features, and request features for different initial features of the item, so as to reflect the association between the different initial features of the item and the item search process and the user's interest, and by controlling the change of the initial features, determine the change of the corresponding constructed association relationship, and thus determine the type of the item feature according to the change of the association relationship, that is, whether it is an independent feature. Based on this, when determining the independent features of the items to be recommended, the item recommendation system can first perform feature extraction on the item recommendation request to obtain the corresponding request features, and then construct an association relationship between any of the above-mentioned items to be recommended, item recommendation requests and request objects based on the extracted initial features of the items to be recommended, the object features of the request object and the request features of the item recommendation request to obtain a feature evaluation score for any item to be recommended, and then input the constructed feature evaluation score into a pre-constructed feature evaluation model to use the feature evaluation score to determine the corresponding independent features from the extracted initial features.

[0060] In an embodiment of the present application, the feature evaluation score includes: a first evaluation score between any item to be recommended and the requested object, and a second evaluation score between any item to be recommended and the item recommendation request; based on the initial features, the request features and the object features, constructing a feature evaluation score corresponding to any recommended item, including: matching the object features with the initial features to obtain a first evaluation score; matching the request features with the initial features to obtain a second evaluation score.

[0061] In an optional scheme of the present embodiment, in order to clearly reflect the association relationship between any item to be recommended, the item recommendation request and the request object, the above-mentioned first evaluation score can be set for any item to be recommended and the request object, so as to use the first evaluation score to reflect the association relationship between the item to be recommended and the request object, for example, to reflect the user's interest in the item to be recommended, and at the same time, the above-mentioned second evaluation score can be set for any item to be recommended and the item recommendation request, so as to use the second evaluation score to reflect the association relationship between the item to be recommended and the item recommendation request, for example, to reflect whether the item to be recommended can meet the user's needs. Based on this, when constructing the above-mentioned feature evaluation score, the item recommendation system can match the object features of the request object with the initial features of different items to be recommended to obtain the first evaluation scores corresponding to different items to be recommended, and match the request features of the item recommendation request with the initial features of different items to be recommended to obtain the second evaluation scores corresponding to different items to be recommended.

[0062] In an embodiment of the present application, the initial features and feature evaluation scores are input into a feature evaluation model, and the feature evaluation model is used to determine independent features from the initial features, including: removing any initial feature in a preset manner, and obtaining the score change of the feature evaluation score; based on the score change, determining independent features from the initial features.

[0063] In an optional scheme of this embodiment, as shown above, considering that the independent features of an item usually affect the user's interest in the item, when the independent features of an item increase, the corresponding user's interest in the item will increase, and when the independent features of an item decrease, the corresponding user's interest in the item will decrease. Therefore, the feature evaluation model for determining independent features can be a counterfactual decoupling model, which can evaluate the type of item features by counterfactual learning. The item recommendation system can observe the changes in the association between any item to be recommended, the item recommendation request and the request object by changing the features of the item, and determine the independent features based on the changes. Based on this, when using the feature evaluation model to determine independent features, the item recommendation system can first remove any initial feature in a preset manner, such as polling, random selection, etc., and obtain the score change of the above feature evaluation score after removing the feature, and then determine the type of the removed initial feature based on the score change. For example, if after removing the initial feature, the feature evaluation score shows the user's interest in the item to be recommended, such as the above first evaluation score has not changed, then it can be initially determined that the feature will not affect the user's interest and is not an independent feature; if after removing the initial feature, the feature evaluation score shows the user's interest in the item to be recommended, such as the above first evaluation score has changed, then it can be initially determined that the feature will affect the user's interest and is an independent feature.

[0064] In an embodiment of the present application, any initial feature is removed in a preset manner, and the score change of the feature evaluation score is obtained, including: selecting the feature to be removed from the initial features in a preset manner; removing the feature to be removed from the initial features, and constructing a test evaluation score corresponding to any item to be recommended based on other features, request features and object features, wherein the other features are used to characterize the features remaining in the initial features after removing the feature to be removed; and determining the score change based on the test evaluation score and the feature evaluation score.

[0065] In an optional scheme of the present embodiment, when determining the score change of the feature evaluation score, the item recommendation system may first select the feature to be removed from the initial features according to the above-mentioned preset method, and then after removing the feature to be removed, reconstruct the association relationship between the item recommendation request, the item to be recommended and the requested object according to other features remaining in the initial features and the request features of the item recommendation request, as well as the object features of the requested object, to obtain the corresponding test evaluation score to determine the user's interest in the recommended item and whether the satisfaction of the item to be recommended with the item recommendation request will change in the absence of the feature to be removed. Correspondingly, based on the test evaluation score and the feature evaluation score, the item recommendation system can determine the above-mentioned score change.

[0066] In an embodiment of the present application, the test evaluation score includes: a first test score between any item to be recommended and the request object, and a second test score between any item to be recommended and the item recommendation request when the feature to be removed is removed; based on the score change, an independent feature is determined from the initial feature, including: based on the score change, determining a first comparison result between the first evaluation score and the first test score, and a second comparison result between the second evaluation score and the second test score; in response to the first comparison result being that the first test score is greater than the first evaluation score, and the second comparison result being that the second test score is less than the second evaluation score, determining that the feature to be removed is an independent feature.

[0067] Corresponding to the feature evaluation score, the test evaluation score may include: when the feature to be removed is removed, the first test score between any item to be recommended and the request object, and the second test score between any item to be recommended and the item recommendation request. When determining independent features from the initial features based on the score changes, the item recommendation situation can first determine the change between the first evaluation score and the first test score based on the score changes, that is, determine the above-mentioned first comparison result, and at the same time determine the change between the second evaluation score and the second score at this time, that is, determine the above-mentioned second comparison result, and then determine whether the above-mentioned removed feature to be removed is an independent feature based on these two comparison results. For example, if the first comparison result is that the first test score is greater than the first evaluation score, and the second comparison result is that the second score at this time is less than the second evaluation score, then the item recommendation system can determine that the feature to be removed is an independent feature of the item to be recommended, otherwise it can be determined that the feature to be removed is a query feature of the item to be recommended.

[0068] In an embodiment of the present application, the above method also includes: obtaining a training evaluation score corresponding to the training item, wherein the training evaluation score is used to reflect the association relationship between the training item, the training item recommendation request and the test request object; inputting multiple training features and the training evaluation score of the training item into the initial evaluation model, and using the initial evaluation model to determine the initial independent features from the training features; matching the initial independent features with the target independent features corresponding to the training items to construct a feature loss function, wherein the feature loss function includes: a first loss function between the training item and the test request object, and a second loss function between the training item and the training item recommendation request; adjusting the initial evaluation model based on the feature loss function to obtain a feature evaluation model.

[0069] In an optional scheme of the present embodiment, in order to accurately determine the independent features from the initial features of the items to be recommended, the item recommendation system may acquire more training items in advance and construct corresponding training evaluation scores for different training items. For example, according to the aforementioned process of determining the feature evaluation scores, the above-mentioned training evaluation scores are constructed for the item features of the training items, the request features of the training item recommendation request and the object features of the test request object to reflect the association relationship between the training items, the training item recommendation request and the test request object, so as to obtain the corresponding training evaluation scores. Then, the multiple training features of the training items and the corresponding training evaluation scores are input into the initial evaluation model to determine the initial independent features from the training features of different training items using the initial evaluation model. Then, for the initial independent features and the target independent features corresponding to the training items, a corresponding feature loss function is constructed. The feature loss function corresponds to the feature evaluation score and may include a first loss function between the training items and the test request object, i.e., a loss function corresponding to the first evaluation score, and a second loss function between the training items and the training item recommendation request, i.e., a loss function corresponding to the second evaluation score. After constructing the feature loss function, the item recommendation system can use the feature loss function to adjust the initial evaluation model to obtain a feature evaluation model with higher accuracy.

[0070] In an embodiment of the present application, the initial features of any item to be recommended are enhanced based on the independent features of any item to be recommended to obtain the target features of any item to be recommended, including: splicing the independent features and the initial features according to the initial weights to obtain the spliced ​​features; inputting the spliced ​​features into a multi-layer perceptron, adjusting the initial weights using the multi-layer perceptron to obtain the target weights; splicing the independent features and the initial features based on the target weights to obtain the target features.

[0071] In an optional solution of this embodiment, in order to accurately determine the target item based on the target feature, when constructing the target features corresponding to different items to be recommended, the item recommendation system may first splice the independent features and the initial features according to the preset initial weights to obtain the above-mentioned spliced ​​features, so as to increase the proportion of the independent features in the target features, so that the item recommendation system can more accurately determine the user's interest in different items to be recommended based on the target features. Considering that the attractiveness of the independent features of different recommended items to users may be different, a simple initial weight may not be able to reflect this. Therefore, after constructing the above-mentioned spliced ​​features, the item recommendation system can further input the spliced ​​features into the above-mentioned multi-layer perceptron, so as to use the multi-layer perceptron to adjust the above-mentioned initial weights, and obtain the target weights that can better reflect the independent features of the items to be recommended, and then use the target weights to splice the independent features and the initial features, so as to obtain more accurate target features.

[0072] In an embodiment of the present application, feature extraction is performed on at least one item to be recommended to obtain initial features of at least one item to be recommended, including: obtaining an item identifier of at least one item to be recommended and an item attribute of at least one item to be recommended; splicing the item identifier and the item attribute to obtain first splicing information; encoding the first splicing information based on a recommendation domain item encoder to obtain initial features of at least one item to be recommended, wherein the recommendation domain item encoder is used to represent an encoder for encoding features of at least one item to be recommended used when recommending items to a requesting object.

[0073] The above-mentioned recommendation domain item encoder may refer to an encoder used for feature encoding of initial features when making item recommendations. Considering that the features of items are needed both in the item search process and in the item recommendation process, at least one item encoder may be set in the item recommendation system to encode the features of items, and considering that in the recommendation domain, the item encoder is mainly used to learn the features of items to predict the user's preference for items, thereby achieving personalized recommendations, while in the search domain, the item encoder is mainly used to understand the relationship between items and user queries, and to improve the relevance and accuracy of search results. The two have different focuses. Therefore, in order to ensure the accuracy when performing item search and item recommendation, corresponding recommendation domain item encoders and search domain item encoders may be set for the recommendation domain and search domain, respectively.

[0074] In an optional scheme of the present embodiment, in order to accurately and richly represent the features of the item to be recommended in the item recommendation process, when extracting the above-mentioned initial features, the item recommendation system may first obtain at least one item identification of the item to be recommended, such as item ID (Identification), SKU (Stock Keeping Unit) and other identifications, as well as item attributes of the item to be recommended, and then concatenate the item identification and the item attributes to obtain the above-mentioned first concatenation information, and use a pre-configured recommendation domain item encoder to encode the first concatenation information to obtain the initial features of the item to be recommended when making item recommendations.

[0075] In an embodiment of the present application, feature extraction is performed on the object information of the request object to obtain object features, including: obtaining the identity identifier of the request object, the object attributes of the request object, and the request type of the item recommendation request; splicing the identity identifier and the object attributes to obtain second splicing information; based on the request type of the item recommendation request, encoding the second splicing information using a recommendation domain request object encoder to obtain object features, wherein the recommendation domain request object encoder is used to represent an encoder that encodes features of the request object when making item recommendations to the request object.

[0076] The above-mentioned recommendation domain request object encoder may refer to an encoder used for feature encoding of object features when recommending items. Similar to the configuration of the recommendation domain item encoder, considering that the features of the request object need to be used in both the item search process and the item recommendation process, at least one request object encoder may be set in the item recommendation system to encode the features of the item, and considering that in the recommendation domain, the request object encoder is mainly used to determine the relationship between the item to be recommended and the request object, so as to determine the request object's interest in the recommended item, while in the search domain, the request object encoder is mainly used to ensure the matching degree between the request object and different items, so as to determine whether different items can meet the needs of the request object. The two have different focuses. Therefore, in order to ensure the accuracy when performing item search and item recommendation, corresponding recommendation domain request object encoders and search domain request object encoders may be set for the recommendation domain and the search domain, respectively.

[0077] In an optional solution of this embodiment, in order to accurately describe the object characteristics of the request object in the item recommendation process, the item recommendation system may first obtain the identity of the request object and the object attributes of the request object, and then splice the obtained identity and object attributes to obtain the above-mentioned second splicing information, and then use the recommendation domain request object encoder to encode the second splicing information to obtain the object characteristics of the request object when recommending items. Considering that under different recommendation requests, the user's needs and interests in different items may be different, therefore, when actually determining the object characteristics of the request object, the item recommendation system may also introduce the request type of the item recommendation request to determine the above-mentioned object characteristics based on the request type of the item recommendation request.

[0078] In an embodiment of the present application, request types include: collaborative filtering type and sequence recommendation type; based on the request type of the item recommendation request, the second splicing information is encoded using the recommendation domain request object encoder to obtain object features, including: in response to the request type being a collaborative filtering type, the second splicing information is encoded based on the recommendation domain request object encoder to obtain object features; in response to the request type being a sequence recommendation type, a historical interaction sequence of the requested object is obtained, and the historical interaction sequence and the second splicing information are encoded based on the recommendation domain request object encoder to obtain object features, wherein the historical interaction sequence is used to characterize a sequence of items that interacted with the requested object within a historical time period.

[0079] The above collaborative filtering type may refer to a method of recommending items to a requesting object based on other objects or other items, which can be divided into two parts: user-based collaborative filtering and item-based collaborative filtering. User-based collaborative filtering can be understood as: determining the interest of the requesting object in different items based on the interest of other objects similar to the requesting object in different items; item-based collaborative filtering can be understood as: recommending an item to other users similar to the object based on the interest of an object in an item. The above sequence recommendation type may refer to a recommendation technology that considers the user behavior sequence, which can predict the content that the user may be interested in next by analyzing the user's historical behavior sequence.

[0080] In an optional scheme of the present embodiment, in order to accurately extract the object features of the requested object, the item recommendation system may, when determining that the request type of the item recommendation request is the above-mentioned collaborative filtering type, directly encode the second splicing information using the recommendation domain request object to obtain the corresponding object features. When determining that the request type of the item recommendation request is the above-mentioned sequence recommendation type, first obtain the historical interaction sequence of the requested object, such as the item sequence corresponding to the items clicked and viewed by the user in the historical time period, and then use the recommendation domain request object encoder to encode the historical interaction sequence and the second splicing information, so as to obtain object features that can further reflect the user's preferences.

[0081] In an embodiment of the present application, a target item is determined from at least one item to be recommended based on an object feature and a target feature of at least one item to be recommended, including: obtaining an object feature vector corresponding to the object feature, and a target feature vector corresponding to the target feature; obtaining a target value based on an inner product of the object feature vector and the target feature vector; constructing item preference scores corresponding to different recommended items based on the target value, wherein the item preference scores are used to reflect the degree of preference of the requesting object for different items to be recommended; and selecting a target item from at least one item to be recommended based on the item preference scores.

[0082] In an optional scheme of the present embodiment, in order to accurately determine the above-mentioned target item from at least one item to be recommended, the item recommendation system can first obtain the object feature vector corresponding to the above-mentioned object feature, and at the same time obtain the target feature vector corresponding to the target feature, and then determine the request object's preference for different items to be recommended based on the inner product of these two vectors, that is, the above-mentioned target value, that is, determine the above-mentioned item preference score, and finally, based on the item preference score, the item recommendation system can filter out the above-mentioned target item from the above-mentioned items to be recommended.

[0083] For ease of understanding, the item recommendation process proposed in this application can be divided into two parts: one part is a process in which the search object obtains the at least one item to be recommended through the search system, and the other part is a process in which the request object displays the target item to the user through the recommendation system. Figure 3 is a schematic diagram of an item search process according to an embodiment of the present application. Figure 4 is a schematic diagram of an item recommendation process according to an embodiment of the present application. First, Figure 3As shown, in the process of searching for items, the item recommendation system can first obtain the user embedding of the user, such as user ID, attributes and other information, the item embedding of the item, such as item ID, attributes and other information, and the request embedding of the item recommendation request, and then respectively use the search domain user encoder, the search domain item encoder and the search domain request encoder to encode these three embeddings to obtain the corresponding first user feature, first item feature and first request feature, wherein, considering that item features can be divided into query features and independent features, therefore, before encoding the item embedding using the search domain item encoder, the item embedding can be processed first to preliminarily separate the information related to the query feature and independent feature of the item from the item embedding to ensure the accuracy of the first item feature encoded by the search domain item encoder. After obtaining the corresponding first user feature, first item feature and first request feature, the item recommendation system can combine the first item feature and the first user feature to determine the first association relationship between the item and the user, and at the same time combine the first item feature and the first request feature to determine the second association relationship between the item and the query request. Based on these two association relationships, the item recommendation system can further determine the query feature and independent feature of the item from the item features. At the same time, the item recommendation system can also search for recommended items that can meet user needs from the database based on these two association relationships.

[0084] After searching for the recommended items, Figure 4As shown, the item recommendation system can use the recommendation domain user encoder and the recommendation domain item encoder to encode the user embedding and the item embedding respectively to obtain the corresponding second user feature and the second item feature. After obtaining the second item feature, the item recommendation system can introduce the first item feature encoded by the search domain item encoder in the aforementioned search process, as well as the independent feature of the determined item, and use these two features to enhance the second item feature to obtain the corresponding target feature. After constructing the target feature, the item recommendation system can determine the user's interest in different items to be recommended based on the target feature and the second user feature, and then determine the target item from the items to be recommended based on the determined interest level to display to the user. In order to ensure the accuracy of the determined interest level, the item recommendation system can additionally configure a score determination model to use the score determination model to determine the user's preference score for different items to be recommended based on the target feature and the second user feature, thereby determining the user's interest in different items to be recommended based on the preference score. Whether it is the aforementioned feature evaluation model or the score determination model mentioned here, an end-to-end training strategy can be adopted during model training to optimize the recommendation task and the search auxiliary task simultaneously by using a joint learning method controlled by a balanced parameter. During the training process, the item recommendation system can use the preference scores between users and items and the matching scores between recommendation requests and items as supervisory signals to use these signals to provide the model with rich semantic information, thereby helping the model to better understand the user's behavior and preferences. In addition, the accuracy of the model can be improved by jointly optimizing multiple loss functions.

[0085] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0086] It should be noted that, for the aforementioned method embodiments, for the sake of simplicity, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the present application is not limited by the described order of actions, because according to the present application, 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 the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present application.

[0087] Through the description of the above implementation methods, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods of each embodiment of the present application.

[0088] According to an embodiment of the present application, a method for recommending items is also provided. Figure 5 is a flow chart of another item recommendation method according to an embodiment of the present application. Figure 5 As shown, the method may include the following steps:

[0089] Step S502, in response to an input instruction on the operation interface, displaying initial features of at least one item to be recommended and object features of the requesting object on the operation interface.

[0090] The request object is used to represent the object that sends the item recommendation request.

[0091] Step S504, responding to the processing instruction acting on the operation interface, displaying the target item on the operation interface.

[0092] Among them, the target item is determined from at least one item to be recommended by object features and target features of at least one item to be recommended, the target feature is obtained by enhancing the initial features of any independent feature of any item to be recommended, the independent feature is used to characterize the interest preference of the search object, and the search object is used to search for any item to be recommended.

[0093] The input instruction may refer to an instruction for obtaining an item to be recommended. The processing instruction may refer to an item recommendation request input by a requesting object.

[0094] In an optional solution of this embodiment, the item recommendation system can display at least one initial feature of the item to be recommended and the object feature of the request object in the above operation interface when receiving the input instruction. When receiving the processing instruction, the item recommendation system can determine the target item based on the initial feature and the object feature according to the above content, and display the target item in the operation interface for user viewing. The specific process of determining the target item can be as shown above, and will not be repeated here.

[0095] According to an embodiment of the present application, a method for recommending items is also provided. Figure 6is a flow chart of another item recommendation method according to an embodiment of the present application. Figure 6 As shown, the method may include the following steps:

[0096] Step S602: acquiring object features of a request object and initial features of at least one item to be recommended by calling a first interface.

[0097] The request object is used to represent the object that sends the item recommendation request, the first interface includes a first parameter, and the parameter value of the first parameter includes object characteristics and initial characteristics.

[0098] Step S604: enhance the initial features of any item to be recommended based on the independent features of any item to be recommended, so as to obtain the target features of any item to be recommended.

[0099] Among them, the independent features are used to characterize the interest preferences of the search object, and the search object is used to search for any item to be recommended.

[0100] Step S606: determining a target item from the at least one item to be recommended based on the object feature and the target feature of the at least one item to be recommended.

[0101] Step S608, outputting the target item by calling the second interface.

[0102] The second interface includes a second parameter, and a parameter value of the second parameter includes a target item.

[0103] In an optional solution of this embodiment, when recommending items to users, the item recommendation system can first obtain the first parameter through the first interface, that is, obtain the object characteristics of the request object and the initial characteristics of the item to be recommended, and use the independent characteristics of any item to be recommended to enhance the initial characteristics of the item to be recommended to obtain the target characteristics of the item to be recommended, and then use the object characteristics and target characteristics to determine the target item from the items to be recommended, and finally output the second parameter by calling the second interface, that is, output the target item, for the convenience of user viewing. The specific process of determining the target item can be as shown above, and will not be repeated here.

[0104] According to an embodiment of the present application, there is also provided an item recommendation device for implementing the above-mentioned item recommendation method. Figure 7 is a structural block diagram of an item recommendation device according to an embodiment of the present application. Figure 7 As shown, the device includes: a feature extraction module 702, a feature enhancement module 704, an item determination module 706 and an item output module 708.

[0105] Among them, the feature extraction module 702 is used to extract features from the object information of the request object in response to receiving an item recommendation request, obtain object features, and extract features from at least one item to be recommended to obtain initial features of at least one item to be recommended, wherein the request object is used to characterize the object that sends the item recommendation request; the feature enhancement module 704 is used to enhance the initial features of any item to be recommended based on the independent features of any item to be recommended, and obtain target features of any item to be recommended, wherein the independent features are used to characterize the interest preferences of the search object, and the search object is used to search for any item to be recommended; the item determination module 706 is used to determine the target item from at least one item to be recommended based on the object features and the target features of at least one item to be recommended; the item output module 708 is used to output the target item.

[0106] In an embodiment of the present application, the above-mentioned device also includes: a request feature extraction unit, which is used to extract features from the item recommendation request to obtain request features of the item recommendation request; an evaluation score construction unit, which is used to construct a feature evaluation score corresponding to any recommended item based on the initial features, request features and object features, wherein the feature evaluation score is used to reflect the association relationship between any item to be recommended, the item recommendation request and the request object; an independent feature determination unit, which is used to input the initial features and the feature evaluation scores into a feature evaluation model, and use the feature evaluation model to determine the independent features from the initial features.

[0107] In an embodiment of the present application, the feature evaluation score includes: a first evaluation score between any item to be recommended and the request object, and a second evaluation score between any item to be recommended and the item recommendation request; the evaluation score construction unit is also used to match the object feature with the initial feature to obtain a first evaluation score; and match the request feature with the initial feature to obtain a second evaluation score.

[0108] In an embodiment of the present application, the independent feature determination unit is further used to: remove any initial feature in a preset manner and obtain a score change of the feature evaluation score; and determine an independent feature from the initial features based on the score change.

[0109] In an embodiment of the present application, the independent feature determination unit is also used to: select features to be removed from the initial features according to a preset method; remove the features to be removed from the initial features, and construct a test evaluation score corresponding to any item to be recommended based on other features, request features and object features, wherein the other features are used to characterize the features remaining in the initial features after removing the features to be removed; determine the score change based on the test evaluation score and the feature evaluation score.

[0110] In an embodiment of the present application, the test evaluation score includes: a first test score between any item to be recommended and the request object, and a second test score between any item to be recommended and the item recommendation request when the feature to be removed is removed; the independent feature determination unit is also used to: determine a first comparison result between the first evaluation score and the first test score, and a second comparison result between the second evaluation score and the second test score based on the score change; in response to the first comparison result being that the first test score is greater than the first evaluation score, and the second comparison result being that the second test score is less than the second evaluation score, determine that the feature to be removed is an independent feature.

[0111] In an embodiment of the present application, the above-mentioned device also includes: a score acquisition module, which is used to obtain a training evaluation score corresponding to a training item, wherein the training evaluation score is used to reflect the association relationship between the training item, the training item recommendation request and the test request object; a feature determination module, which is used to input multiple training features and training evaluation scores of the training item into an initial evaluation model, and use the initial evaluation model to determine the initial independent features from the training features; a function construction module, which is used to match the initial independent features and the target independent features corresponding to the training items, and construct a feature loss function, wherein the feature loss function includes: a first loss function between the training item and the test request object, and a second loss function between the training item and the training item recommendation request; a model adjustment module, which is used to adjust the initial evaluation model based on the feature loss function to obtain a feature evaluation model.

[0112] In the embodiment of the present application, the feature enhancement module 704 includes: a first splicing unit, which is used to splice the independent features and the initial features according to the initial weights to obtain the spliced ​​features; a weight adjustment unit, which is used to input the spliced ​​features into the multi-layer perceptron, and use the multi-layer perceptron to adjust the initial weights to obtain the target weights; a second splicing unit, which is used to splice the independent features and the initial features based on the target weights to obtain the target features.

[0113] In the embodiment of the present application, the feature extraction module 702 includes: an attribute acquisition unit, used to obtain the item identification of at least one item to be recommended, and the item attributes of at least one item to be recommended; a third splicing unit, used to splice the item identification and the item attributes to obtain first splicing information; a first encoding unit, used to encode the first splicing information based on a recommendation domain item encoder to obtain the initial features of at least one item to be recommended, wherein the recommendation domain item encoder is used to represent an encoder for encoding the features of at least one item to be recommended when recommending items to a requesting object.

[0114] In the embodiment of the present application, the feature extraction module 702 includes: a parameter acquisition unit, which is used to obtain the identity of the request object, the object attributes of the request object, and the request type of the item recommendation request; a fourth splicing unit, which is used to splice the identity and the object attributes to obtain second splicing information; a second encoding unit, which is used to encode the second splicing information based on the request type of the item recommendation request using a recommendation domain request object encoder to obtain object features, wherein the recommendation domain request object encoder is used to represent an encoder that encodes the features of the request object when recommending items to the request object.

[0115] In an embodiment of the present application, the second encoding unit is further used to: in response to the request type being a collaborative filtering type, encode the second splicing information based on the recommendation domain request object encoder to obtain object features; in response to the request type being a sequence recommendation type, obtain the historical interaction sequence of the requested object, and encode the historical interaction sequence and the second splicing information based on the recommendation domain request object encoder to obtain object features, wherein the historical interaction sequence is used to characterize the sequence of items that interacted with the requested object within a historical time period.

[0116] In the embodiment of the present application, the item determination module 706 includes: a vector acquisition unit, which is used to obtain an object feature vector corresponding to the object feature, and a target feature vector corresponding to the target feature; a value determination unit, which is used to obtain a target value based on the inner product of the object feature vector and the target feature vector; a score construction unit, which is used to construct item preference scores corresponding to different recommended items based on the target value, wherein the item preference scores are used to reflect the degree of preference of the requesting object for different items to be recommended; and an item selection module, which is used to select a target item from at least one item to be recommended based on the item preference scores.

[0117] It should be noted that the feature extraction module 702, feature enhancement module 704, item determination module 706 and item output module 708 correspond to steps S202 to S208 in the above embodiment, and the four modules and corresponding steps implement the same examples and application scenarios, but are not limited to the contents disclosed in the above embodiment. It should be noted that the above modules or units can be hardware components or software components stored in a memory and processed by one or more processors, and the above modules can also be run in the server 10 provided in the above embodiment as part of the device.

[0118] According to an embodiment of the present application, there is also provided an item recommendation device for implementing the above-mentioned item recommendation method. Figure 8 is a structural block diagram of another item recommendation device according to an embodiment of the present application, such as Figure 8 As shown, the device includes: a first display module 802 and a second display module 804 .

[0119] Among them, the first display module 802 is used to respond to input instructions acting on the operation interface, and display at least one initial feature of the item to be recommended and the object feature of the requesting object on the operation interface, wherein the requesting object is used to characterize the object that sends the item recommendation request; the second display module 804 is used to respond to processing instructions acting on the operation interface, and display the target item on the operation interface, wherein the target item is determined from at least one item to be recommended by the object feature and the target feature of at least one item to be recommended, and the target feature is obtained by enhancing the initial feature of any item to be recommended by the independent feature of any item to be recommended, and the independent feature is used to characterize the interest preference of the search object, and the search object is used to search for any item to be recommended.

[0120] It should be noted that the first display module 802 and the second display module 804 correspond to steps S502 to S504 in the above embodiment, and the examples and application scenarios implemented by the two modules and the corresponding steps are the same, but are not limited to the contents disclosed in the above embodiment. It should be noted that the above modules or units can be hardware components or software components stored in a memory and processed by one or more processors, and the above modules can also be run in the server 10 provided in the above embodiment as part of the device.

[0121] According to an embodiment of the present application, there is also provided an item recommendation device for implementing the above-mentioned item recommendation method. Fig. 9 is a structural block diagram of another item recommendation device according to an embodiment of the present application, such as Fig. 9 As shown, the device includes: a first calling module 902, an enhancing module 904, a determining module 906 and a second calling module 908.

[0122] Among them, the first calling module 902 is used to obtain the object characteristics of the request object and the initial characteristics of at least one item to be recommended by calling the first interface, wherein the request object is used to characterize the object that sends the item recommendation request, the first interface includes a first parameter, and the parameter value of the first parameter includes the object characteristics and the initial characteristics; the enhancement module 904 is used to perform feature enhancement on the initial characteristics of any item to be recommended based on the independent characteristics of any item to be recommended to obtain the target characteristics of any item to be recommended, wherein the independent characteristics are used to characterize the interest preferences of the search object, and the search object is used to search for any item to be recommended; the determination module 906 is used to determine the target item from at least one item to be recommended based on the object characteristics and the target characteristics of at least one item to be recommended; the second calling module 908 is used to output the target item by calling the second interface, wherein the second interface includes a second parameter, and the parameter value of the second parameter includes the target item.

[0123] It should be noted that the first calling module 902, the enhancing module 904, the determining module 906 and the second calling module 908 correspond to steps S602 to S608 in the above embodiment, and the four modules and the corresponding steps implement the same examples and application scenarios, but are not limited to the contents disclosed in the above embodiment. It should be noted that the above modules or units can be hardware components or software components stored in a memory and processed by one or more processors, and the above modules can also be run in the server 10 provided in the above embodiment as part of the device.

[0124] It should be noted that the preferred implementation scheme involved in the above embodiments of the present application is the same as the scheme provided in the above embodiments, as well as the application scenario and implementation process, but is not limited to the scheme provided in the above embodiments.

[0125] An embodiment of the present application may provide a computing device. Fig.10 is a structural block diagram of a computing device according to an embodiment of the present application. Fig.10 As shown, the computing device 1000 may include: one or more (only one is shown in the figure) processors 1002, a memory 1004, a storage controller, and a peripheral interface.

[0126] The above-mentioned computing device can be understood as an integrated intelligent terminal, including but not limited to a server, a desktop computer, a PC (Personal Computer), a model all-in-one machine, etc., and the computing device can be pre-installed with the model in the above-mentioned embodiment of the present application.

[0127] Specifically, the computing device can pre-set multiple types of models, including but not limited to models in the fields of natural language processing, visual processing, speech processing, code processing, multimodal task processing, etc., so as to provide a variety of model choices. In different product forms, the computing device can support one or more model usage methods, including but not limited to model training, model calling, model fine-tuning, model deployment, model reasoning and application, etc. In some product forms, the computing device also supports model management, including but not limited to multi-type model management (supporting the management of multiple types of models such as discriminants and generative models), model version control (supporting the control of different model versions), model evaluation (based on model evaluation tools, evaluating the performance and effect of the model), etc. In other product forms, the computing device can also create applications based on the model, provide API calling capabilities, and can call the model to the created application through the API interface, while providing application management tools to achieve application management and control.

[0128] Furthermore, the computing device can also include data management (supporting the creation and management of model tuning data sets), a training center (providing rich training resources to help users learn and master AI technology), and basic management and control capabilities (providing enterprise-level basic management and control capabilities to ensure the security and efficient operation of the system). Through the above functions, a comprehensive, integrated AI development, training, deployment and application device is provided.

[0129] Among them, the memory can be used to store software programs and modules, such as program instructions / modules corresponding to the methods and devices in the embodiments of the present application. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, that is, implementing the methods in the above embodiments. 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 arranged relative to the processor, and these remote memories may be connected to the terminal A 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.

[0130] The processor may call the executable program stored in the memory through the transmission device to execute any method in the above embodiments.

[0131] An embodiment of the present application may provide an electronic device. Fig.11 1 is a structural block diagram of an electronic device according to an embodiment of the present application. Fig.11 As shown, the electronic device may include: an input / output device 1102 ; a memory 1104 , and a processor 1106 , wherein the processor 1106 is connected to the input / output device 1102 and the memory 1104 via a bus 1108 .

[0132] Among them, the memory can be used to store software programs and modules, such as program instructions / modules corresponding to the methods and devices in the embodiments of the present application. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, that is, implementing the methods in the above embodiments. 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 arranged relative to the processor, and these remote memories may be connected to the terminal A 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.

[0133] The processor may call the executable program stored in the memory through the transmission device to execute any method in the above embodiments.

[0134] It will be understood by those skilled in the art that the structure shown in the figure is for illustration only, and the computing device may also be a terminal device such as a smart phone (such as an Android phone, an iOS phone, etc.), a tablet computer, a PDA, and a mobile Internet device (Mobile Internet Devices, MID), PAD, etc. The figure does not limit the structure of the above computing device. For example, the computing device 1000 may also include more or fewer components (such as a network interface, a display device, etc.) than those shown in the figure, or have a configuration different from that shown in the figure.

[0135] A person of ordinary skill in the art can 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, and the storage medium may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0136] The embodiment of the present application 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 method provided in the above embodiment.

[0137] Optionally, in this embodiment, the above storage medium may be located in a computing device.

[0138] Optionally, in this embodiment, the computer-readable storage medium is configured to store an executable program, and when the executable program is running, the device where the computer-readable storage medium is located is controlled to execute any one of the methods in the above embodiments.

[0139] The embodiment of the present application further provides a computer program product. Optionally, in this embodiment, the computer program product may include a computer program, and the computer program implements the method provided in the embodiment when executed by a processor.

[0140] The embodiments of the present application also provide a computer program product. Optionally, the computer program product may include a non-volatile computer-readable storage medium, which may be used to store a computer program, and when the computer program is executed by a processor, the method provided in the embodiments is implemented.

[0141] The embodiment of the present application further provides a computer program. Optionally, in this embodiment, when the computer program is executed by a processor, the method provided in the above embodiment is implemented.

[0142] In the above embodiments of the present application, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0143] 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 units is only a logical function division, and there may be other division methods in actual implementation, for example, 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.

[0144] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0145] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0146] 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 application, 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, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, server or network device, etc.) to perform all or part of the steps of each embodiment method of the present application. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, disk or optical disk, etc. Various media that can store program codes.

[0147] The above are only preferred implementations of the present application. 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 application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. An item recommendation method, characterized in that: include: In response to receiving an item recommendation request, extracting features from the object information of the request object to obtain object features, and extracting features from at least one item to be recommended to obtain initial features of the at least one item to be recommended, wherein the request object is used to represent the object that sent the item recommendation request; Based on the independent features of any one of the items to be recommended, feature enhancement is performed on the initial features of the item to be recommended to obtain the target features of the item to be recommended, wherein the independent features are used to characterize the interest preference of the search object, and the search object is used to search for the item to be recommended; Determining a target item from the at least one item to be recommended based on the object feature and the target feature of the at least one item to be recommended; Output the target item.

2. The method according to claim 1, characterized in that Before enhancing the initial features of any one of the items to be recommended based on the independent features of the items to be recommended, the method further includes: Extracting features of the item recommendation request to obtain request features of the item recommendation request; Based on the initial features, the request features and the object features, construct a feature evaluation score corresponding to any one of the recommended items, wherein the feature evaluation score is used to reflect the association relationship between the any one of the recommended items, the item recommendation request and the request object; The initial features and the feature evaluation scores are input into a feature evaluation model, and the independent features are determined from the initial features using the feature evaluation model.

3. The method according to claim 2, characterized in that The feature evaluation score includes: a first evaluation score between the any one of the items to be recommended and the request object, and a second evaluation score between the any one of the items to be recommended and the item recommendation request; constructing the feature evaluation score corresponding to the any one of the recommended items based on the initial features, the request features and the object features includes: Matching the object feature with the initial feature to obtain the first evaluation score; The request feature is matched with the initial feature to obtain the second evaluation score.

4. The method according to claim 2, characterized in that: The step of inputting the initial features and the feature evaluation scores into a feature evaluation model, and using the feature evaluation model to determine the independent features from the initial features comprises: Remove any initial feature in a preset manner, and obtain a score change of the feature evaluation score; Based on the score change, the independent feature is determined from the initial feature.

5. The method according to claim 4, characterized in that The removing of any initial feature in a preset manner and obtaining a score change of the feature evaluation score includes: Selecting features to be removed from the initial features according to the preset method; The feature to be removed is removed from the initial feature, and a test evaluation score corresponding to any one of the items to be recommended is constructed based on other features, the request feature and the object feature, wherein the other features are used to characterize the features remaining in the initial feature after the feature to be removed is removed; Based on the test evaluation score and the feature evaluation score, the score change is determined.

6. The method according to claim 5, characterized in that The test evaluation score includes: a first test score between the any one of the items to be recommended and the request object, and a second test score between the any one of the items to be recommended and the item recommendation request, when the feature to be removed is removed; The step of determining the independent feature from the initial feature based on the score change includes: Based on the score changes, determining a first comparison result between the first evaluation score and the first test score, and a second comparison result between the second evaluation score and the second test score; In response to the first comparison result being that the first test score is greater than the first evaluation score, and the second comparison result being that the second test score is less than the second evaluation score, it is determined that the feature to be removed is the independent feature.

7. The method according to claim 2, characterized in that The method further comprises: Obtaining a training evaluation score corresponding to the training item, wherein the training evaluation score is used to reflect the association relationship between the training item, the training item recommendation request, and the test request object; Inputting the plurality of training features of the training object and the training evaluation score into an initial evaluation model, and using the initial evaluation model to determine initial independent features from the training features; Based on matching the initial independent features with the target independent features corresponding to the training items, a feature loss function is constructed, wherein the feature loss function includes: a first loss function between the training items and the test request object, and a second loss function between the training items and the training item recommendation request; The initial evaluation model is adjusted based on the feature loss function to obtain the feature evaluation model.

8. The method according to claim 1, characterized in that The step of enhancing the initial features of any one of the items to be recommended based on the independent features of the items to be recommended to obtain the target features of the items to be recommended includes: Concatenate the independent features and the initial features according to the initial weights to obtain concatenated features; Inputting the concatenated features into a multi-layer perceptron, and using the multi-layer perceptron to adjust the initial weights to obtain target weights; The independent feature and the initial feature are concatenated based on the target weight to obtain the target feature.

9. The method according to claim 1, characterized in that: The extracting features of at least one item to be recommended to obtain initial features of the at least one item to be recommended includes: Acquire an item identifier of the at least one item to be recommended and an item attribute of the at least one item to be recommended; splicing the item identifier and the item attribute to obtain first splicing information; The first splicing information is encoded based on a recommendation domain item encoder to obtain initial features of the at least one item to be recommended, wherein the recommendation domain item encoder is used to represent an encoder for encoding features of the at least one item to be recommended when recommending items to the requesting object.

10. The method according to claim 1, characterized in that The extracting features of the object information of the request object to obtain the object features includes: Obtaining the identity of the request object, the object attributes of the request object, and the request type of the item recommendation request; Concatenate the identity identifier and the object attribute to obtain second concatenated information; Based on the request type of the item recommendation request, the second splicing information is encoded using a recommendation domain request object encoder to obtain the object features, wherein the recommendation domain request object encoder is used to represent an encoder used to encode the features of the request object when making item recommendations to the request object.

11. The method according to claim 10, characterized in that The request types include collaborative filtering type and sequence recommendation type; the request type based on item recommendation request, using the recommendation domain request object encoder to encode the second splicing information to obtain the object features, includes: In response to the request type being the collaborative filtering type, requesting an object encoder to encode the second concatenation information based on the recommendation domain to obtain the object feature; In response to the request type being the sequence recommendation type, a historical interaction sequence of the requested object is obtained, and based on the recommendation domain request object encoder, the historical interaction sequence and the second splicing information are encoded to obtain the object feature, wherein the historical interaction sequence is used to characterize a sequence of items that interacted with the requested object within a historical time period.

12. The method according to claim 1, characterized in that Determining a target item from the at least one item to be recommended based on the object feature and the target feature of the at least one item to be recommended includes: Obtaining an object feature vector corresponding to the object feature and a target feature vector corresponding to the target feature; Obtaining a target value based on an inner product of the object feature vector and the target feature vector; Based on the target value, constructing item preference scores corresponding to different recommended items, wherein the item preference scores are used to reflect the degree of preference of the requesting object for different items to be recommended; The target item is selected from the at least one item to be recommended based on the item preference score.

13. An item recommendation method, characterized in that: include: In response to an input instruction acting on the operation interface, displaying on the operation interface at least one initial feature of an item to be recommended and an object feature of a requesting object, wherein the requesting object is used to represent an object sending the item recommendation request; In response to a processing instruction acting on the operation interface, a target item is displayed on the operation interface, wherein the target item is determined from the at least one item to be recommended by the object feature and the target feature of the at least one item to be recommended, and the target feature is obtained by enhancing the initial feature of any one of the items to be recommended using an independent feature of the item to be recommended, and the independent feature is used to characterize the interest preference of a search object, and the search object is used to search for any one of the items to be recommended.

14. An item recommendation method, characterized in that: include: Acquire object characteristics of a request object and initial characteristics of at least one item to be recommended by calling a first interface, wherein the request object is used to represent the object sending the item recommendation request, the first interface includes a first parameter, and a parameter value of the first parameter includes the object characteristics and the initial characteristics; Based on the independent features of any one of the items to be recommended, feature enhancement is performed on the initial features of the item to be recommended to obtain the target features of the item to be recommended, wherein the independent features are used to characterize the interest preference of the search object, and the search object is used to search for the item to be recommended; Determining a target item from the at least one item to be recommended based on the object feature and the target feature of the at least one item to be recommended; The target item is output by calling a second interface, wherein the second interface includes a second parameter, and a parameter value of the second parameter includes the target item.

15. A computing device, characterized in that: include: A memory storing an executable program; A processor, configured to run the program, wherein the program executes the method according to any one of claims 1 to 13 when running.

16. An electronic device, characterized in that: include: A memory storing an executable program; A processor, connected to the memory via a bus, and configured to run the program, wherein the program executes the method described in any one of claims 1 to 13 when running.

17. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored executable program, wherein when the executable program is executed, the device where the computer-readable storage medium is located is controlled to execute the method according to any one of claims 1 to 13.

18. A computer program product, characterized in that The invention comprises a computer program which, when executed by a processor, implements the method according to any one of claims 1 to 13.

Images