A recommendation information acquisition method and device, electronic equipment and storage medium

Abstract

Embodiments of the present application provide a recommendation information acquisition method and device, electronic equipment and storage medium, based on the operation category, grouping each historical operation collected, obtaining at least one corresponding historical operation combination, and performing corresponding feature extraction on each obtained historical operation combination, obtaining the historical operation features corresponding to each historical operation combination, further, based on the preset feature dimension, dimension reduction processing is performed on at least one obtained historical operation feature, effectively reducing the operation complexity required for feature aggregation, thereby ensuring full use of each historical operation while reducing the operation time required to obtain the target operation feature, improving the efficiency of obtaining the recommendation information.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method, apparatus, electronic device, and storage medium for obtaining recommendation information. Background Technology

[0002] With the continuous development of the information age, a massive amount of online information, such as product display information, is rapidly emerging on internet platforms. However, when faced with a large volume of this online information, target users often need to spend a significant amount of time searching to accurately obtain the display information related to the products they are interested in. Therefore, large information display or exchange platforms typically employ Information Recommendation Systems (IRS). These systems, after requesting permission from the target user, analyze and predict the target user's information preferences based on collected user activity information. Based on these determined preferences, targeted information recommendations are then sent to the target user to meet their personalized information needs.

[0003] Specifically, in related technologies, feature analysis is typically performed on historical operation data collected from a target object within a specified time range, based on a pre-defined neural network. To independently distinguish each historical operation within the historical operation data, the historical operation data is further divided into sets of historical operations triggered for each specified operation attribute, according to the operation attributes of each historical operation, such as the corresponding operation medium and operation category. In these related technologies, feature analysis is performed on each of the aforementioned historical operation sets to obtain corresponding operation features. These features are then aggregated based on a pre-defined connection network layer to obtain corresponding target operation features. Furthermore, using these target operation features, the information recommendation system can analyze and match at least one associated target recommendation information from a specified target database, thereby achieving targeted information recommendation for the target object based on the obtained at least one target recommendation information.

[0004] However, the above methods still have the following drawbacks:

[0005] In related technologies, the operation attributes of each historical operation are usually defined based on the operation medium and operation category of each historical operation. However, in reality, the number of operation mediums and operation categories triggered by the target object is often large. This approach often leads to an increase in the number of historical operation sets when dividing historical operation data. Consequently, the number of operation features determined for each historical operation set in the connection network layer is usually large. Therefore, in related technologies, when aggregating the operation features, the large number of determined operation features often results in a large computational load, causing a lot of waiting time during the acquisition of target operation features. As a result, the efficiency of acquiring recommendation information is usually low in reality. Summary of the Invention

[0006] This application provides a method, apparatus, electronic device, and storage medium for obtaining recommendation information, which can improve the efficiency of obtaining recommendation information.

[0007] On the one hand, embodiments of this application provide a method for obtaining recommendation information, the method comprising:

[0008] Retrieves a set of historical operations triggered by the target object for at least one specified operation medium.

[0009] In the historical operation set, each historical operation is grouped according to the operation category to obtain at least one corresponding historical operation combination, and feature extraction is performed on each of the at least one historical operation combination to obtain the corresponding historical operation features.

[0010] Based on a preset dimension mapping function, at least one historical operation feature is subjected to dimensionality reduction processing to obtain corresponding dimensionality-reduced operation features. The at least one dimensionality-reduced operation features are then aggregated to obtain the target operation feature.

[0011] In the recommendation information database, based on the information features of each recommendation, the similarity between the information features and the target operation features is determined, and when the similarity is greater than a preset similarity threshold, the recommendation is used as the target recommendation for matching the target operation features.

[0012] On one hand, embodiments of this application provide a recommendation information acquisition device, which includes:

[0013] The acquisition module is used to acquire a set of historical operations triggered by the target object for at least one specified operation medium.

[0014] The extraction module is used to group each historical operation in the historical operation set according to the operation category, obtain at least one corresponding historical operation combination, and extract features from each of the at least one historical operation combination to obtain the corresponding historical operation features.

[0015] The mapping module performs dimensionality reduction on at least one historical operation feature based on a preset dimensionality mapping function to obtain corresponding dimensionality-reduced operation features, and then aggregates these dimensionality-reduced operation features to obtain the target operation feature. The matching module determines the similarity between the information features of each recommendation and the target operation feature in the recommendation information database. If the similarity is greater than a preset similarity threshold, the recommendation is used as the target recommendation for matching the target operation feature.

[0016] Optionally, when extracting features from at least one combination of historical operations to obtain the corresponding historical operation features, the extraction module is specifically used for:

[0017] For at least one combination of historical operations, perform the following operations respectively:

[0018] Based on the operation medium of each historical operation in a historical operation combination, the corresponding historical operation template is determined from the preset operation template set.

[0019] Based on the obtained historical operation templates, the parameter information associated with the corresponding historical operations is aggregated to generate corresponding historical description information.

[0020] Feature extraction is performed on the obtained historical description information to obtain the corresponding historical operation features.

[0021] Optionally, feature extraction is performed on the obtained historical description information to obtain the corresponding historical operation features. Specifically, the extraction module is used for:

[0022] The obtained historical description information is embedded and encoded to obtain the corresponding historical description vectors.

[0023] Based on the operational medium associated with each historical description information, the historical description vectors are grouped to obtain at least one corresponding combination of description vectors.

[0024] In at least one combination of description vectors, the corresponding historical description vectors are aggregated to obtain at least one combined description vector.

[0025] Aggregate at least one combined description vector to obtain the corresponding historical operation features.

[0026] Optionally, the historical operation set is associated with at least one historical time interval, wherein each historical time interval is determined based on a preset division duration, after dividing the corresponding historical time range when the historical operation set is triggered.

[0027] When extracting features from at least one combination of historical operations to obtain the corresponding historical operation features, the extraction module is specifically used for:

[0028] For at least one combination of historical operations, perform the following operations respectively:

[0029] Based on the trigger time of each historical operation in a historical operation combination, the historical operations are grouped according to at least one determined historical time interval to obtain at least one corresponding interval operation combination.

[0030] Feature extraction is performed on at least one combination of interval operations to obtain corresponding interval operation features, and the at least one interval operation features are aggregated to obtain corresponding historical operation features.

[0031] Optionally, after aggregating at least one dimensionality reduction operation feature to obtain the target operation feature, the extraction module is further used to: obtain the target basic features of the target object, wherein the target basic features are used to describe the basic attributes of the target object.

[0032] In the recommendation information database, when determining the similarity between the information features and the target operation features based on the information features of each recommendation, the matching module is specifically used for:

[0033] The target's basic features and operational features are aggregated to obtain the corresponding target recommendation features.

[0034] In the recommendation information database, based on the information features of each recommendation, the similarity between the information features and the target operation features is determined. On one hand, embodiments of this application provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps of the above-described recommendation information acquisition method.

[0035] On one hand, embodiments of this application provide a computer-readable storage medium storing a computer program executable by a computer device, which, when run on the computer device, causes the computer device to perform the steps of the above-described method for obtaining recommendation information.

[0036] In one aspect, embodiments of this application also provide a computer program product that, when invoked by a computer, causes the computer to execute the method described in the first aspect.

[0037] This application provides a method, apparatus, electronic device, and storage medium for obtaining recommendation information. Based on operation categories, the collected historical operations are grouped to obtain at least one corresponding historical operation combination. Then, feature extraction is performed on each historical operation combination to obtain the historical operation features corresponding to each historical operation combination. Through the above method, the effective merging of historical operations of the same category is achieved, making the management of the obtained historical operations by the information recommendation system more reasonable and efficient, thereby improving the scalability of the information recommendation system.

[0038] Furthermore, based on the preset feature dimensions, at least one historical operation feature is subjected to dimensionality reduction processing. This method significantly reduces the computational complexity during feature aggregation, thereby ensuring full utilization of each historical operation while reducing the computation time required to obtain the target operation feature and improving the efficiency of recommendation information acquisition. Attached Figure Description

[0039] Figure 1 This is a schematic diagram of a related technical process in an embodiment of this application;

[0040] Figure 2 This is a schematic diagram of an application scenario in an embodiment of this application;

[0041] Figure 3 This is a schematic diagram of an interactive window in an embodiment of this application;

[0042] Figure 4 This is a flowchart of a recommendation information acquisition method in an embodiment of this application;

[0043] Figure 5 This is a schematic diagram of a possible historical operation in an embodiment of this application;

[0044] Figure 6 This is a schematic diagram of a method for obtaining historical operation features in an embodiment of this application;

[0045] Figure 7 This is a schematic diagram of a target operation feature acquisition method in an embodiment of this application;

[0046] Figure 8 This is a logical diagram illustrating a recommendation information acquisition scenario in an embodiment of this application;

[0047] Figure 9 This is a schematic diagram of the structure of a recommendation information acquisition device according to an embodiment of this application;

[0048] Figure 10 This is a schematic diagram of an electronic device using an embodiment of this application. Detailed Implementation

[0049] The technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this invention.

[0050] For ease of understanding, the terms used in the embodiments of this invention are explained below.

[0051] Operation medium: refers to the actual carrier that carries the historical operations triggered by the target object, such as news information, marketing information, etc.

[0052] Operation category: refers to the type of actual interactive action performed on the target object, such as browsing, clicking, etc.

[0053] Information recommendation: refers to the process of matching and displaying relevant information to the target object based on the target object's historical operations triggered by a specified operating medium, with the target object's own permission. For example, marketing information push.

[0054] The design concept of the embodiments of this application will be briefly introduced below.

[0055] This application's embodiments relate to information recommendation scenarios, where information recommendation refers to the process of matching and displaying relevant information to the target object based on the target object's historical operations triggered by a specified operation medium, provided that the target object has its own permission.

[0056] In related technologies, the above information recommendations are typically made in the following ways:

[0057] In related technologies, historical operation data is typically divided based on the operational attributes of each historical operation triggered by a target object within a specified time range. Each historical operation set corresponds to a unique operational attribute, which is composed of the operational medium and category of the corresponding historical operation. In these technologies, feature analysis is typically performed on each of the divided historical operation sets to obtain corresponding operational features. These features are then aggregated based on a pre-defined connection network layer to obtain target operational features. Furthermore, based on these target operational features, relevant information is matched and displayed to the target object to achieve targeted information recommendation.

[0058] For example, see Figure 1As shown, assuming the target object triggers various historical operations for n different operation media within a specified time range, and the operation categories of the above historical operations are statistically analyzed to determine that there are a total of m operation categories, then in the related technology, by determining the corresponding m*n operation attributes, the historical operation data collected for the target object is divided into a specified m*n historical operation set, thereby aggregating the operation features of each of the above historical operation sets, so as to realize the targeted information recommendation of the target object based on the corresponding target operation features.

[0059] However, the above methods often face the following shortcomings:

[0060] Firstly, in related technologies, it is usually necessary to divide the collected historical operation data according to the defined operation attributes. However, in reality, the number of operation media and operation categories triggered by the target object is often large, and the number of corresponding operation attributes is also large. As a result, the computational complexity required to aggregate various operation features is usually large. Therefore, the above methods often require a lot of waiting time to obtain the corresponding target operation features of the target object, resulting in the problem of low efficiency in obtaining recommendation information in related technologies.

[0061] Secondly, in related technologies, the number of defined operation categories and operation media are often intersected to determine the total number of operation attributes for the target object. In this way, when the number of operation categories or operation media increases, the number of corresponding operation attributes will inevitably increase as well. This makes it difficult for the managed object to expand the operation categories or operation media defined in the system while ensuring the load of the information recommendation system itself, thus affecting the optimization efficiency of the information recommendation system.

[0062] In view of this, embodiments of this application provide a method, apparatus, electronic device, and storage medium for obtaining recommendation information. In this method, based on the operation category of historical operations, each collected historical operation is grouped to obtain at least one corresponding historical operation combination. Then, feature extraction is performed on each historical operation combination to obtain the historical operation features corresponding to each historical operation combination. In the above manner, each determined operation category is defined as a unique historical operation feature, thereby improving the scalability of the information recommendation system during the system optimization process.

[0063] Furthermore, based on the preset feature dimensions, each historical operation feature is subjected to dimensionality reduction processing. In this way, the target operation features of the target object are determined through the obtained dimensionality-reduced operation features. The above method makes the dimensions of each obtained dimensionality-reduced operation feature fixed and controllable, thereby significantly reducing the computational complexity during feature aggregation. This method effectively reduces the large amount of waiting time required to obtain target operation features and improves the efficiency of obtaining recommendation information.

[0064] The preferred embodiments of this application are described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit this application. Furthermore, the embodiments and features in the embodiments of this application can be combined with each other without conflict.

[0065] See Figure 2 The diagram shown is an application scenario illustration of an embodiment of this application. The application scenario diagram includes a terminal device 210 and a server 220.

[0066] In this embodiment, the terminal device 210 includes, but is not limited to, mobile phones, tablets, laptops, desktop computers, e-book readers, smart voice interaction devices, smart home appliances, and in-vehicle terminals. The terminal device may have a client installed related to obtaining recommendation information. This client can be software (e.g., a browser), a webpage, or a mini-program. The server 220 is the backend server corresponding to the software, webpage, or mini-program, or a server specifically used for obtaining recommendation information; this application does not impose specific limitations. The server 220 can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN (Content Delivery Network), and big data and artificial intelligence platforms.

[0067] It should be noted that the recommendation information acquisition method in this application embodiment can be executed by an electronic device, which can be a server 220 or a terminal device 210. That is, the method can be executed by the server 220 or the terminal device 210 alone, or by the server 220 and the terminal device 210 together.

[0068] In one alternative implementation, the terminal device 210 and the server 220 can communicate via a communication network.

[0069] In one alternative implementation, the communication network is a wired network or a wireless network.

[0070] It should be noted that, Figure 2 The examples shown are merely illustrative; in reality, the number of terminal devices and servers is unlimited and is not specifically limited in the embodiments of this application.

[0071] In this embodiment of the application, when there are multiple servers, the multiple servers can form a blockchain, and the servers are nodes on the blockchain; as disclosed in the recommendation information acquisition method of this embodiment, the set of historical operations triggered by the target object for at least one specified operation medium can be stored on the blockchain, such as browsing click information, following and liking information, etc.

[0072] The following description, in conjunction with the application scenarios described above and with reference to the accompanying drawings, illustrates the method for obtaining recommendation information provided by exemplary embodiments of this application. It should be noted that the above application scenarios are merely illustrative for the purpose of understanding the spirit and principles of this application, and the embodiments of this application are not limited in any way. Furthermore, the embodiments of this application can be applied to various scenarios, including not only information recommendation scenarios, but also scenarios involving cloud technology, artificial intelligence, smart transportation, and assisted driving.

[0073] In order to record the various historical operations triggered by the target object within a specified time range with the target object's permission, it is often necessary for the terminal device to make corresponding operation recording requests to the target object based on one or more interactive windows set in the time range. For example, in an optional embodiment, the terminal device can present a specified interactive window to the target object through the aforementioned installed client related to obtaining recommendation information. The interactive window can be a software interface or mini-program interface corresponding to the aforementioned client, or it can be one or more information display windows with interactive functions included in the aforementioned software interface or mini-program interface.

[0074] For example, see Figure 3 As shown, the aforementioned terminal device can present a specified interactive window to the target object through a deployed client. This interactive window can contain descriptive information related to the operation record request, as well as interactive buttons that can be associated with the descriptive information. In an optional embodiment, the terminal device can respond to the relevant operation triggered by the target object in response to the aforementioned interactive button, obtain the corresponding operation record permission, and based on the obtained operation record permission, record the various historical operations triggered by the target object within the corresponding time range in a short period of time.

[0075] Furthermore, in order to independently record each historical operation triggered by the aforementioned target object, each historical operation is represented as a combination of the corresponding operation medium and operation category, specifically as follows:

[0076] 1. Operation Category

[0077] In reality, the types of actual interactive actions performed by the target object are often different. In order to more accurately identify the operation attributes of each historical operation, the type of actual interactive action corresponding to each historical operation is regarded as the operation category of that historical operation. Thus, each historical operation can be independently distinguished based on its operation category.

[0078] For example, in one possible scenario, if a target object performs a "browse" operation on product display information A that it is interested in, then when recording the historical operations triggered by the target object on product display information A, the type of the actual interaction action of the historical operation can be identified by recording the operation category "browse".

[0079] Furthermore, in information recommendation scenarios, to improve the efficiency of acquiring recommendation information, the aforementioned specified operation categories can be determined by considering the actual operation categories that might be triggered when a target object has an information preference for a certain product display information. In this embodiment, the aforementioned operation categories specifically include:

[0080] 1) Browse

[0081] In reality, when a target audience has a preference for certain product display information, they will often actively or passively view the relevant content during the process of displaying the product information to the target audience.

[0082] For example, when product display information is passively or actively exposed to a target object, it can be regarded as the target object that has triggered a corresponding browsing operation in response to the exposed product display information, indicating that the target object may be interested in the product display information. In this embodiment of the application, the "browse" operation category can be used as a specified operation category mentioned above, so that when the target object triggers the relevant historical operation of this type, the corresponding operation category is recorded.

[0083] 2) Click

[0084] In practice, when a target audience has a preference for a particular product's displayed information, they will often click on related information after viewing it to request more details. This demonstrates the target audience's level of interest in the displayed content.

[0085] For example, in some news information display interfaces, after viewing the displayed content, the target object can further trigger a corresponding click operation on the access point related to obtaining more information in the display interface, indicating that the target object requests to obtain more information content. In this embodiment of the application, the "click" operation category can be used as a specified operation category mentioned above, so that when the target object triggers the relevant historical operation of this type, the corresponding operation category is recorded.

[0086] 3) Transformation

[0087] In practice, when a target audience has a preference for a certain product display information, they will often take further action after the product display information is shown to them, in order to express their level of interest in the displayed content.

[0088] For example, in some marketing information display interfaces, target users can install an application associated with the product display information or purchase a specific product associated with the product display information after viewing the relevant product display information. In this embodiment of the application, the "conversion" operation category can be used as a specified operation category mentioned above, so that when the target user triggers the relevant historical operation of this type, the corresponding operation category is recorded.

[0089] 4) Comments

[0090] In practice, when a target audience has a preference for a particular product display, they will often comment on the displayed content after it is shown to them, thus expressing their interest in the content.

[0091] For example, in some news information display interfaces, the target audience can leave comments in the corresponding comment section after viewing the relevant content. In this embodiment of the application, the "comment" operation category can be used as a specified operation category mentioned above, so that when the target audience triggers the relevant historical operation of this type, the corresponding operation category is recorded.

[0092] 5) Like

[0093] In practice, when a target audience has a preference for a particular product display, they will often express their interest in the displayed content by "liking" or doing other actions after the product display is shown to them.

[0094] For example, in some news information display interfaces, the target object can trigger corresponding "like" or other operations on the display interface to indicate that they are satisfied with the displayed content. In this embodiment of the application, the "like" operation category can be used as a specified operation category mentioned above, so that when the target object triggers the relevant historical operation of this type, the corresponding operation category is recorded.

[0095] 6) Stay

[0096] In practice, when a target audience has a preference for certain product display information, they will often spend a relatively long time on the relevant display interface presented by that product display information. This also expresses the target audience's level of interest in the displayed content.

[0097] For example, in some news information display interfaces, the target audience may spend a lot of time carefully reading the content displayed on the interface. This indicates that the target audience is interested in the content. In this embodiment of the application, the "stay" operation category can be used as a specified operation category mentioned above, so that when the target audience triggers the relevant historical operation of this type, the corresponding operation category is recorded.

[0098] 7) Share

[0099] In practice, when a target audience has a preference for a particular product display, they will often share that product display with other audiences after it has been shown to them, in order to express their interest in the displayed content.

[0100] For example, after a target object reads some news information, it can send it to other objects it specifies by triggering an operation such as "share". This indicates that the target object is interested in the content it displays. In this embodiment of the application, the "share" operation category can be used as a specified operation category mentioned above, so that when the target object triggers the relevant historical operation of this type, the corresponding operation category is recorded.

[0101] 2. Operating medium

[0102] In practice, different historical operations often target different actual carriers. In order to more accurately identify the operational attributes of each historical operation, the actual carrier targeted by each historical operation is regarded as the operation medium of that historical operation. Thus, each historical operation can be independently distinguished based on the operation medium of each historical operation.

[0103] For example, in one possible scenario, if a target object performs a "browse" operation on product display information A that it is interested in, then when recording the historical operations triggered by the target object on product display information A, the actual carrier targeted by the historical operation can be identified by recording the operation medium "product display information A".

[0104] Furthermore, in real-world scenarios, the operational medium triggered by the target object might be a news article, a video, or product display information presented for a specific product during the aforementioned process. Specifically, by organizing and categorizing different types of operational mediums in information recommendation scenarios, these operational mediums can include:

[0105] 1) Display medium

[0106] In an optional embodiment, the aforementioned operating medium may include a display medium, such as product display information describing related products. Through the product-related information described by such a display medium, the types of products that the target object may be interested in can be further determined, indicating that the target object may have information preferences for the determined product types.

[0107] For example, when a target audience sees an e-commerce advertisement, based on the information about the e-commerce product described in the advertisement, such as clothing, it can be determined that the target audience may have an information preference for information displayed about clothing e-commerce products.

[0108] 2) Program medium

[0109] In an optional embodiment, the aforementioned operating medium may include program media, such as parts of software or applications. Through the relevant information of such program media, the types of products that the target object may be interested in can be further determined, indicating that the target object may have information preferences for the aforementioned determined product types.

[0110] For example, when a target audience views content presented by a specific app, based on relevant information in the app such as the app carrying a "game" attribute identifier, it can be determined that the target audience may have an information preference for apps with game attributes.

[0111] 3) Information media

[0112] In an optional embodiment, the aforementioned operating medium may include information media, such as news information, articles, etc. Through the relevant information of such information media, the types of products that the target object may be interested in can be further determined, so as to indicate that the target object may have information preferences for the determined product types.

[0113] For example, when a target audience views relevant content presented by a news article, based on the relevant information carried by the news article, such as the news category being sports, it can be determined that the target audience may have an information preference for sports news.

[0114] 4) Marketing media

[0115] In an optional embodiment, the aforementioned operating medium may include marketing media, such as marketing information. Through the relevant information of such marketing media, the types of products that the target audience may be interested in can be further determined, indicating that the target audience may have information preferences for the aforementioned determined product types.

[0116] For example, when a target audience sees marketing information presented to a specific entity, based on relevant information such as the entity described in the marketing information belonging to the e-commerce category "home appliances," it can be determined that the target audience may have an information preference for home appliance marketing information.

[0117] Furthermore, in order to more accurately capture the relevant information of the various types of operating media identified above, in this embodiment of the application, corresponding operating templates are configured for each type of operating media identified above, specifically including:

[0118] 1. Display operation template

[0119] Specifically, since the aforementioned display media typically represent product display information for a specific product, a more granular description can be made based on the basic attributes of the specified product, such as its industry and brand. Therefore, the corresponding display operation template can include any one or a combination of the following:

[0120] 1) Product Industry

[0121] Specifically, when a certain operating medium is a display medium, the industry to which the relevant products described by the above-mentioned operating medium belong can be determined, such as e-commerce, education, games, etc., and the operating medium can be described accordingly, thereby making a clear distinction between operating media of different industries.

[0122] In an optional embodiment, each of the above-mentioned designated industries can be assigned a unique industry identifier based on the actual situation. For example, the industry identifier for "e-commerce" can be determined as "237". In actual circumstances, the above-mentioned industries are also represented by their corresponding industry identifiers, which will not be elaborated further here.

[0123] 2) Product Brand

[0124] Specifically, when a certain operating medium is a display medium, the brand of the relevant product described by the above-mentioned operating medium can be identified, such as Apple, Nike, etc., and the operating medium can be described accordingly, thereby making a clear distinction between different operating media in the industry.

[0125] 3) Product Categories

[0126] Specifically, when a certain operating medium is a display medium, the category of the relevant products described by the above-mentioned operating medium can be determined, such as jeans, women's clothing, etc., and the operating medium can be described accordingly, thereby making a clear distinction between operating media of different categories.

[0127] 2. Program Operation Template

[0128] Specifically, since the aforementioned program medium is typically represented as a specific software entity or related application, in practice, it can be described in a more granular way based on its fundamental attributes. Therefore, the corresponding program operation template can include any one or a combination of the following:

[0129] 1) Program Identifier

[0130] Specifically, when an operating medium is a program medium, the operating medium can be described accordingly by determining the basic attributes carried by the program medium, such as program identifiers. That is, different program media can be clearly distinguished by the different program identifiers.

[0131] 3. Information Operation Template

[0132] Specifically, since the aforementioned information media are typically represented as relevant news or articles viewed by a target audience, in practice, a more granular description can be made based on the basic attributes of the information media. Therefore, the corresponding information operation template can include any one or a combination of the following:

[0133] 1) Information Identifier

[0134] Specifically, when an operating medium is an information medium, the operating medium can be described accordingly by determining the basic attributes carried by the information medium, such as information identifiers. That is, different information media can be clearly distinguished by the different information identifiers mentioned above.

[0135] 2) Article Theme

[0136] Specifically, when a certain operating medium is an information medium, it can also be described according to the article theme of that information medium, such as politics, sports, economics, entertainment, etc., so as to clearly distinguish operating media with different article themes.

[0137] 4. Marketing Operation Template

[0138] Specifically, since the aforementioned marketing media typically represent marketing information targeting a specific product, in practice, based on the basic attributes of the specified product, the marketing media can be further refined and described. Therefore, the corresponding marketing operation template can include any one or a combination of the following:

[0139] 1) Marketing Category

[0140] Specifically, when a certain operating medium is a marketing medium, the operating medium can be described accordingly by determining the marketing category to which the specified product belongs, such as electrical appliances or food. It is worth noting that in practice, the above marketing categories can be further subdivided into finer-grained categories according to the specified category classification criteria. In this case, the above marketing categories can also be represented as a combination of the corresponding subcategories determined for the specified product.

[0141] For example, when a certain operating medium is a marketing medium, the primary category of the corresponding e-commerce product can be determined as "Appliances" based on preset primary marketing categories, such as electrical appliances, food, and clothing. Further, based on preset secondary marketing categories, such as home appliances, snack foods, and women's clothing, the secondary category of the e-commerce product can be determined as "Home Appliances". On this basis, based on preset tertiary marketing categories, such as refrigerators, potato chips, and women's jeans, the tertiary category of the e-commerce product can be determined as "Refrigerators". The marketing category to which the above e-commerce products belong can then be represented as "Appliances-Home Appliances-Refrigerators", providing a more granular description of the marketing categories to which different marketing media belong.

[0142] For further details, please refer to [link / reference]. Figure 4 As shown, after obtaining permission to access the target object's operation records, based on the various operation media and operation categories determined above, this application embodiment provides a method for obtaining recommendation information, including:

[0143] S401: Obtain the set of historical operations triggered by the target object for at least one specified operation medium.

[0144] Specifically, the aforementioned operation medium can be the interactive interface presented by any APP on the terminal device, such as relevant news, information or product display information. The aforementioned historical operation set can be characterized as the corresponding historical interactive actions performed by the target object in response to the relevant news or product display information presented in the interactive interface. Each historical operation includes a specified operation medium and a corresponding operation category.

[0145] For example, see Figure 5The diagram illustrates a historical operation triggered by a target object. This historical operation is performed on a specific news app on a terminal device. Specifically, the target object viewed news-related content displayed in the interactive interface of the news app. In one optional embodiment, the historical operation is recorded as "Browsing_Information Media" based on the corresponding operation medium and operation category.

[0146] By using the above method, the historical operations performed by the target object on at least one specified operating medium are effectively recorded, thereby ensuring the readability of each historical operation. Furthermore, in an optional embodiment, it is assumed that the target object performed corresponding historical operations on at least one specified news information or product display information in various interactive interfaces presented by the terminal device within a specified historical time range. Specifically, as shown in Table 1 below:

[0147] Table 1

[0148]

[0149]

[0150] S402: In the historical operation set, group each historical operation based on the operation category to obtain at least one corresponding historical operation combination, and extract features from each of the at least one historical operation combination to obtain the corresponding historical operation features.

[0151] Furthermore, taking the historical operations shown in Table 1 above as examples, the historical operations are grouped based on the specified operation categories "browse", "like", and "convert". The corresponding grouping results are shown in Table 2 below:

[0152] Table 2

[0153]

[0154] As can be seen from the above, by specifying each operation category, the original 5 historical operations in Table 1 are further merged into corresponding combinations of 2 historical operations. Based on this coarse-grained merging method, the historical operation set is mapped to each specified operation category, thereby achieving preliminary adjustment of the corresponding feature dimensions during the feature extraction process.

[0155] It is worth noting that, in practice, the parameter information associated with different types of operating media often differs. For example, for the aforementioned display media, the industry and brand of the corresponding designated product can be used as the parameter information associated with the corresponding historical operation; while for program media, the identified program identifier is usually used as the parameter information associated with the corresponding historical operation. Therefore, in an optional embodiment, the historical operation characteristics of a combination of historical operations can be obtained in the following ways:

[0156] S4021: Based on the operation medium of each historical operation in a historical operation combination, determine the corresponding historical operation template from the preset operation template set.

[0157] Specifically, each of the preset operation templates describes the basic attributes of each specified operation medium. In this embodiment, the four operation templates mentioned above are used as examples to determine the basic attributes for each type of operation medium.

[0158] It is understood that the above operation templates are only examples. In one possible scenario, the content of the basic attributes recorded in each operation template can be adjusted accordingly. For example, for the marketing operation template, the specific classification rules of the marketing categories it contains can be adjusted to make it suitable for some more special marketing scenarios, which will not be elaborated here.

[0159] Furthermore, taking the historical operation combination corresponding to the "browse" operation category as an example, for its included historical operation 1, based on the operation medium (display medium a) corresponding to the historical operation, its corresponding historical operation template is determined to be the aforementioned display operation template, thereby determining that the basic attributes corresponding to the operation medium may include: product industry, product brand, product category, etc. Similarly, for its included historical operation 2, based on the operation medium (information medium b) corresponding to the historical operation, its corresponding historical operation template is determined to be the aforementioned information operation template, thereby determining that the basic attributes corresponding to the operation medium may include: information identifier, article theme, etc., which will not be elaborated further here.

[0160] S4022: Based on the obtained historical operation templates, aggregate the parameter information associated with the corresponding historical operations to generate corresponding historical description information.

[0161] Specifically, for each historical operation template, the corresponding basic attributes are determined and aggregated with the corresponding parameter information associated with each historical operation when it is actually triggered. Based on the generated historical description information, each historical operation contained in the historical operation combination is clearly described.

[0162] For example, in an optional embodiment, assuming that the parameter information associated with historical operation 1 includes "e-commerce-Nike-jeans", the parameter information associated with historical operation 2 includes "123-sports", and the parameter information associated with historical operation 3 includes "234-sports", then the corresponding historical description information is shown in Table 3 below:

[0163] Table 3

[0164]

[0165]

[0166] S4023: Extract features from the obtained historical description information to obtain the corresponding historical operation features.

[0167] Specifically, for each given combination of historical operations, a pre-defined encoding method can be used to transform it into a corresponding historical description vector based on its defined historical description information.

[0168] For example, in an ANN neural network, based on the historical description information determined above, it can be converted into corresponding historical description vectors by embedding encoding, which will not be elaborated further.

[0169] Furthermore, since the historical description vectors obtained above have high similarity among those with the same associated operation medium, in order to reduce the computational load, the historical description vectors with the same operation medium are aggregated.

[0170] For example, for each historical operation shown in Table 3 above, corresponding historical description vectors 1 to 3 are obtained respectively. Among them, historical description vectors 2 and 3, which are associated with the same operation medium, can be aggregated and then further aggregated with the aforementioned historical description vector 1 to obtain the corresponding historical operation features.

[0171] Based on the above explanation, please refer to Figure 6 As shown, when determining the operation category and operation medium of each historical operation triggered by the target object, the collected historical operations are grouped based on the operation medium. In each corresponding historical operation combination, the corresponding historical operations are further divided and aggregated based on the operation medium. Based on this feature structure after division, the information recommendation system manages the acquired historical operations more reasonably and efficiently, thereby improving the scalability of the information recommendation system.

[0172] Furthermore, in real-world scenarios, the aforementioned set of historical operations may also be associated with at least one specified historical time interval. Each historical time interval is determined by dividing the corresponding historical time range when the aforementioned set of historical operations is triggered, based on a preset division duration.

[0173] For example, if the target object triggers at least one historical operation on a specified operation medium within the historical time range [12:00, 13:00], then in order to improve computational efficiency, the above historical time range is often divided based on a preset division duration.

[0174] As an example, assuming the time interval is 30 minutes, the corresponding historical time intervals are [12:00, 12:30] and [12:30, 13:00]. During the execution of S402 above, each historical operation combination can be further divided into time intervals based on the trigger time of each historical operation. Taking the historical operation combinations associated with the "Browse" operation category in Table 2 as an example, assuming the trigger time for historical operation 1 is 12:00, the trigger time for historical operation 2 is 12:10, and the trigger time for historical operation 3 is 12:40, the corresponding interval operation combinations are shown in Table 4 below:

[0175] Table 4

[0176]

[0177] Based on the above method, feature extraction is performed on each obtained interval operation combination to obtain corresponding interval operation features. Each interval operation feature corresponds to a specified historical time interval. By aggregating at least one obtained interval operation feature, corresponding historical operation features are obtained. This reduces the overall system load and ensures the scalability of the information recommendation system while making full use of the operation attributes of each obtained historical operation.

[0178] S403: Based on a preset dimension mapping function, perform dimensionality reduction processing on at least one historical operation feature to obtain corresponding dimensionality-reduced operation features, and aggregate the at least one dimensionality-reduced operation features to obtain the target operation feature. Further, see [link to documentation]. Figure 7 As shown, in order to reduce the computational complexity required for feature aggregation, each historical operation feature is dimensionality-reduced based on a preset feature dimension. This method significantly reduces the computational complexity during feature aggregation, thereby ensuring full utilization of each historical operation while reducing the computation time required to obtain the target operation feature and improving the efficiency of obtaining recommendation information.

[0179] S404: In the recommendation information database, based on the information features of each recommendation information, determine the similarity between the information features and the target operation features, and when the similarity is greater than a preset similarity threshold, use the recommendation information as the target recommendation information for matching the target operation features.

[0180] Furthermore, based on the constructed target operation features mentioned above, at least one target recommendation information matching the target operation features is obtained from the specified recommendation information database, so as to use each obtained target recommendation information as the information recommendation result for the target object.

[0181] In an optional embodiment, after obtaining the target operation features, the target basic features of the target object can be further obtained. These target basic features are used to describe the basic attributes of the target object, such as specific identifiers determined for the target object. By aggregating the above-mentioned target basic features with the target operation features, corresponding target recommendation features are constructed. In scenarios with multiple objects, at least one corresponding target recommendation information is matched with the constructed target recommendation features, and the obtained at least one target recommendation information is used to provide targeted information recommendations to the specified target object associated with the target recommendation features.

[0182] In an optional embodiment, based on the information features of each recommendation in the above-mentioned recommendation information database, the similarity between each information feature and the target operation feature can be determined respectively, so that each recommendation that meets the preset similarity threshold is used as the target recommendation information that matches the obtained target operation feature.

[0183] See Figure 8 The diagram below illustrates the logic of a recommendation information acquisition method provided in this application. As shown in Table 5, assuming the target object triggers various historical operations for a specified operation medium within a specified time range [12:00, 13:00], the determined historical operations are grouped according to their operation categories. Then, the historical operation features of each historical operation are reduced in dimensionality using a preset feature dimension. By aggregating the obtained reduced-dimensional operation features, the target operation features of the target object within the above time range can be obtained quickly and accurately, ensuring the timeliness and accuracy of the obtained target recommendation information.

[0184] Based on the same technical concept, this application provides a schematic diagram of the structure of a recommendation information acquisition device, such as... Figure 9 As shown, the device 900 includes:

[0185] The acquisition module 901 is used to acquire a set of historical operations triggered by the target object for at least one specified operation medium.

[0186] The extraction module 902 is used to group each historical operation in the historical operation set according to the operation category to obtain at least one corresponding historical operation combination, and to extract features from each of the at least one historical operation combination to obtain the corresponding historical operation features.

[0187] The mapping module 903 is used to perform dimensionality reduction processing on at least one historical operation feature obtained based on a preset dimensionality mapping function to obtain corresponding dimensionality reduction operation features, and to aggregate the at least one dimensionality reduction operation features to obtain target operation features; wherein the dimensionality reduction operation features are in a preset fixed feature dimension.

[0188] The matching module 904 is used to determine the similarity between the information features and the target operation features of each recommendation information in the recommendation information database, and when the similarity is greater than a preset similarity threshold, the recommendation information is used as the target recommendation information for matching the target operation feature.

[0189] Optionally, when extracting features from at least one combination of historical operations to obtain the corresponding historical operation features, the extraction module 902 is specifically used for:

[0190] For at least one combination of historical operations, perform the following operations respectively:

[0191] Based on the operation medium of each historical operation in a historical operation combination, the corresponding historical operation template is determined from the preset operation template set.

[0192] Based on the obtained historical operation templates, the parameter information associated with the corresponding historical operations is aggregated to generate corresponding historical description information.

[0193] Feature extraction is performed on the obtained historical description information to obtain the corresponding historical operation features.

[0194] Optionally, when extracting features from the obtained historical description information to obtain the corresponding historical operation features, the extraction module 902 is specifically used for:

[0195] The obtained historical description information is embedded and encoded to obtain the corresponding historical description vectors.

[0196] Based on the operational medium associated with each historical description information, the historical description vectors are grouped to obtain at least one corresponding combination of description vectors.

[0197] In at least one combination of description vectors, the corresponding historical description vectors are aggregated to obtain at least one combined description vector.

[0198] Aggregate at least one combined description vector to obtain the corresponding historical operation features.

[0199] Optionally, the historical operation set is associated with at least one historical time interval, wherein each historical time interval is determined based on a preset division duration, after dividing the corresponding historical time range when the historical operation set is triggered.

[0200] When extracting features from at least one combination of historical operations to obtain the corresponding historical operation features, the extraction module 902 is specifically used for:

[0201] For at least one combination of historical operations, perform the following operations respectively:

[0202] Based on the trigger time of each historical operation in a historical operation combination, the historical operations are grouped according to at least one determined historical time interval to obtain at least one corresponding interval operation combination.

[0203] Feature extraction is performed on at least one combination of interval operations to obtain corresponding interval operation features, and the at least one interval operation features are aggregated to obtain corresponding historical operation features.

[0204] Optionally, after aggregating the obtained at least one dimensionality reduction operation feature to obtain the target operation feature, the extraction module 902 is further configured to: obtain the target basic features of the target object, wherein the target basic features are used to describe the basic attributes of the target object.

[0205] In the recommendation information database, when determining the similarity between the information features and the target operation features based on the information features of each recommendation, the matching module 904 is specifically used for:

[0206] The target's basic features and operational features are aggregated to obtain the corresponding target recommendation features.

[0207] In the recommendation information database, the similarity between the information features and the target operation features is determined based on the information features of each recommendation.

[0208] In this embodiment, based on the operation category, the collected historical operations are grouped to obtain at least one corresponding historical operation combination. Then, the corresponding features are extracted for each historical operation combination to obtain the historical operation features corresponding to each historical operation combination. Through the above method, the effective merging of historical operations of the same category is achieved, making the information recommendation system more capable of managing the obtained historical operations, thereby further ensuring the scalability of the system.

[0209] Furthermore, based on the preset feature dimensions, at least one historical operation feature is subjected to dimensionality reduction processing. This method significantly reduces the computational complexity during feature aggregation, thereby ensuring full utilization of each historical operation while reducing the computation time required to obtain the target operation feature and improving the efficiency of recommendation information acquisition.

[0210] Based on the same technical concept, embodiments of this application provide a computer device, which can be... Figure 2 The terminal devices and / or servers shown, such as Figure 10 As shown, it includes at least one processor 1001 and a memory 1002 connected to at least one processor. In this embodiment, the specific connection medium between the processor 1001 and the memory 1002 is not limited. Figure 10 Taking the connection between processor 1001 and memory 1002 via a bus as an example. The bus can be divided into address bus, data bus, control bus, etc.

[0211] In this embodiment of the application, the memory 1002 stores instructions that can be executed by at least one processor 1001. By executing the instructions stored in the memory 1002, at least one processor 1001 can perform the steps of the above-described method for obtaining recommendation information.

[0212] The processor 1001 is the control center of the computer device. It can connect to various parts of the computer device using various interfaces and lines. It retrieves recommendation information by running or executing instructions stored in the memory 1002 and calling data stored in the memory 1002. Optionally, the processor 1001 may include one or more processing units. The processor 1001 may integrate an application processor and a modem processor. The application processor mainly handles the operating system, user interface, and applications, while the modem processor mainly handles wireless communication. It is understood that the modem processor may not be integrated into the processor 1001. In some embodiments, the processor 1001 and the memory 1002 can be implemented on the same chip; in some embodiments, they can also be implemented on separate chips.

[0213] The processor 1001 can be a general-purpose processor, such as a central processing unit (CPU), digital signal processor, application-specific integrated circuit (ASIC), field-programmable gate array (FPGA), or other programmable logic device, discrete gate or transistor logic device, or discrete hardware component, capable of implementing or executing the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this application can be directly manifested as being executed by a hardware processor, or executed by a combination of hardware and software modules within the processor.

[0214] Memory 1002, as a non-volatile computer-readable storage medium, can be used to store non-volatile software programs, non-volatile computer-executable programs, and modules. Memory 1002 may include at least one type of storage medium, such as flash memory, hard disk, multimedia card, card-type memory, random access memory (RAM), static random access memory (SRAM), programmable read-only memory (PROM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), magnetic memory, magnetic disk, optical disk, etc. Memory 1002 can be any other medium capable of carrying or storing desired program code in the form of instructions or data structures that can be accessed by a computer, but is not limited thereto. In the embodiments of this application, memory 1002 can also be a circuit or any other device capable of implementing storage functions for storing program instructions and / or data.

[0215] Based on the same inventive concept, embodiments of this application provide a computer-readable storage medium storing a computer program executable by a computer device, which, when run on the computer device, causes the computer device to perform the steps of the above-described method for obtaining recommended information.

[0216] Based on the same inventive concept, this application provides a computer program product, which includes a computer program stored on a computer-readable storage medium. The computer program includes program instructions, which, when executed by a computer, cause the computer to perform the steps of the above-described method for obtaining recommendation information.

[0217] Those skilled in the art will understand that embodiments of the present invention can be provided as methods or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0218] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0219] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0220] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0221] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0222] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for obtaining recommendation information, characterized in that, include: Obtain a set of historical operations triggered by a target object in response to at least one specified operation medium; the operation medium is the actual carrier that carries the historical operations triggered by the target object. In the set of historical operations, each historical operation is grouped according to the operation category to obtain at least one corresponding combination of historical operations, and features are extracted from each of the at least one combination of historical operations to obtain corresponding historical operation features. For each of the at least one historical operation combination, perform the following operations respectively: Based on the operation medium of each historical operation in a historical operation combination, the corresponding historical operation template is determined from the preset operation template set; wherein, each historical operation template is used to describe the basic attributes of the corresponding operation medium. Based on the obtained historical operation templates, the parameter information associated with the corresponding historical operations is aggregated to generate corresponding historical description information. Feature extraction is performed on the obtained historical description information to obtain the corresponding historical operation features; The step of extracting features from the obtained historical description information to obtain corresponding historical operation features includes: The obtained historical description information is embedded and encoded to obtain the corresponding historical description vectors. Based on the operational medium associated with each of the historical description information, the historical description vectors are grouped to obtain at least one corresponding combination of description vectors. Each historical description vector in the at least one combination of description vectors is aggregated to obtain at least one combined description vector. Aggregate at least one combined description vector to obtain the corresponding historical operation features; Based on a preset dimension mapping function, at least one historical operation feature is subjected to dimensionality reduction processing to obtain corresponding dimensionality-reduced operation features. The at least one dimensionality-reduced operation feature is then aggregated to obtain the target operation feature. Each dimensionality-reduced operation feature is located in a preset fixed feature dimension, and each dimensionality-reduced operation feature contains the basic attributes of the corresponding historical operation feature. In the recommendation information database, based on the information features of each recommendation information, the similarity between the information features and the target operation features is determined, and when the similarity is determined to be greater than a preset similarity threshold, the recommendation information is used as the target recommendation information for matching the target operation features.

2. The method as described in claim 1, characterized in that, The historical operation set is associated with at least one historical time interval, wherein each historical time interval is determined based on a preset division duration, after dividing the corresponding historical time range when the historical operation set is triggered. The step of extracting features from the at least one historical operation combination to obtain corresponding historical operation features includes: For each of the at least one historical operation combination, perform the following operations respectively: Based on the trigger time of each historical operation in a historical operation combination, the historical operations are grouped according to at least one determined historical time interval to obtain at least one corresponding interval operation combination. Feature extraction is performed on at least one combination of interval operations to obtain corresponding interval operation features, and the at least one interval operation features are aggregated to obtain corresponding historical operation features.

3. The method as described in claim 1, characterized in that, After aggregating the obtained at least one dimensionality reduction operation feature to obtain the target operation feature, the method further includes: obtaining the target basic features of the target object, wherein the target basic features are used to describe the basic attributes of the target object. The step of determining the similarity between the information features and the target operation features based on the information features of each recommendation in the recommendation information database includes: The target basic features and the target operational features are aggregated to obtain corresponding target recommendation features; In the recommendation information database, based on the information features of each recommendation, the similarity between the information features and the target operation features is determined.

4. A device for obtaining recommendation information, characterized in that, include: The acquisition module is used to acquire a set of historical operations triggered by a target object in response to at least one specified operation medium; the operation medium is the actual carrier that carries the historical operations triggered by the target object. The extraction module is used to group each historical operation in the historical operation set based on the operation category to obtain at least one corresponding historical operation combination, and to extract features from the at least one historical operation combination to obtain corresponding historical operation features. The extraction module is specifically used to perform the following operations for each of at least one historical operation combination: Based on the operation medium of each historical operation in a historical operation combination, the corresponding historical operation template is determined from the preset operation template set. Based on the obtained historical operation templates, the parameter information associated with the corresponding historical operations is aggregated to generate corresponding historical description information. Feature extraction is performed on the obtained historical description information to obtain the corresponding historical operation features; The extraction module is specifically used for: The obtained historical description information is embedded and encoded to obtain the corresponding historical description vectors. Based on the operational medium associated with each historical description information, the historical description vectors are grouped to obtain at least one corresponding combination of description vectors. In at least one combination of description vectors, the corresponding historical description vectors are aggregated to obtain at least one combined description vector. The mapping module is used to perform dimensionality reduction processing on at least one historical operation feature obtained based on a preset feature dimension mapping function to obtain corresponding dimensionality reduction operation features, and to aggregate the at least one dimensionality reduction operation features to obtain target operation features; wherein, each dimensionality reduction operation feature is located in a preset fixed feature dimension, and each dimensionality reduction operation feature contains the basic attributes of the corresponding historical operation feature. The matching module is used to determine the similarity between the information features of each recommendation and the target operation feature in the recommendation information database, and when the similarity is greater than a preset similarity threshold, the recommendation is used as the target recommendation for matching the target operation feature.

5. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the method of any one of claims 1-3.

6. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that, when executed on a computer, cause the computer to perform the method as described in any one of claims 1-3.

7. A computer program product, characterized in that, When the computer program product is invoked by a computer, it causes the computer to perform the method as described in any one of claims 1-3.

Images