Data processing method and device, equipment and storage medium

Abstract

The invention provides a data processing method and device, equipment and a storage medium, and relates to the field of data processing, in particular to the fields of artificial intelligence, intelligent recommendation, big data and the like. The specific implementation scheme is as follows: acquiring initial sample data of a target object; the initial sample data represents the degree of interest of the target object in first resource data; on the basis of the initial sample data, estimating to obtain N pieces of second resource data in which the target object is interested; n is a positive integer greater than or equal to 1; based on the object features of the target object and the resource features of the second resource data in the N pieces of second resource data, N pieces of newly-added sample data of the target object are obtained, and the newly-added sample data in the N pieces of newly-added sample data can represent the interest degree of the target object in the second resource data.

Description

Technical Field

[0001] The present disclosure relates to the field of data processing technology, and in particular to the fields of artificial intelligence, intelligent recommendation, big data, etc. Background Art

[0002] In the aforementioned recommendation scenarios, mainstream ranking modules in the industry often use supervised deep learning methods to score resources. The effectiveness of deep learning depends crucially on the scale and quality of data samples. In other words, the ranking module's deep learning model requires massive amounts of historical behavior and resource data for training to improve scoring accuracy. However, for less active objects in the recommendation system, the sparsity of their historical behavior data prevents the model from fully learning their behavior, resulting in lower accuracy in the estimated scores, impacting recommendation effectiveness and even reducing the user experience. Summary of the Invention

[0003] The present disclosure provides a data processing method, apparatus, device, and storage medium.

[0004] According to one aspect of the present disclosure, there is provided a data processing method, comprising:

[0005] Acquire initial sample data of the target object; the initial sample data represents the interest level of the target object in the first resource data;

[0006] Based on the initial sample data, N second resource data of interest to the target object are estimated; N is a positive integer greater than or equal to 1;

[0007] Based on the object characteristics of the target object and the resource characteristics of each second resource data in the N second resource data, N new sample data of the target object are obtained, wherein the new sample data in the N new sample data can represent the target object's interest level in the second resource data.

[0008] According to another aspect of the present disclosure, there is provided a data processing apparatus, comprising:

[0009] A sample acquisition unit, configured to acquire initial sample data of a target object; the initial sample data represents the interest level of the target object in the first resource data;

[0010] A resource expansion unit, configured to estimate N second resource data of interest to the target object based on the initial sample data; wherein N is a positive integer greater than or equal to 1;

[0011] A sample generation unit is used to obtain N new sample data of the target object based on the object characteristics of the target object and the resource characteristics of each second resource data in the N second resource data, wherein the new sample data in the N new sample data can represent the degree of interest of the target object in the second resource data.

[0012] According to another aspect of the present disclosure, there is provided an electronic device, comprising:

[0013] at least one processor; and

[0014] a memory communicatively connected to the at least one processor; wherein,

[0015] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform any method in the embodiments of the present disclosure.

[0016] According to another aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to enable the computer to execute any method according to the embodiments of the present disclosure.

[0017] According to another aspect of the present disclosure, a computer program product is provided, including a computer program. When the computer program is executed by a processor, the computer program implements any one of the methods according to the embodiments of the present disclosure.

[0018] In this way, the disclosed solution can make full use of the initial sample data of the target object to expand and obtain new sample data for the target object. In other words, the disclosed solution provides a specific solution for increasing sample data, thereby effectively increasing the amount of sample data for the target object, thereby providing strong support for effectively solving the recommendation problem of low-activity objects.

[0019] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present disclosure, nor are they intended to limit the scope of the present disclosure. Other features of the present disclosure will become readily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The accompanying drawings are provided to facilitate a better understanding of the present invention and do not constitute a limitation of the present disclosure.

[0021] Figure 1 This is a schematic flow chart of a data processing method according to an embodiment of the present application. Figure 1 ;

[0022] Figure 2 This is a schematic flow chart of a data processing method according to an embodiment of the present application. Figure 2 ;

[0023] Figure 3 is a flowchart of a data processing method in a specific example according to an embodiment of the present application;

[0024] Figure 4 1 is a schematic diagram of a scenario of a sample data processing method based on large language model reasoning according to an embodiment of the present application;

[0025] Figure 5 Schematic diagram of a scenario for recommending model training according to an embodiment of the present application;

[0026] Figure 6 is a structural diagram of a data processing device according to an embodiment of the present application;

[0027] Figure 7 It is a block diagram of an electronic device used to implement the data processing method according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0028] The following description of exemplary embodiments of the present disclosure is made in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be appreciated by those skilled in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0029] The term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. The term "at least one" in this article means any combination of at least two of any one or more of a plurality of. For example, including at least one of A, B, and C, can mean including any one or more elements selected from the set consisting of A, B, and C. The terms "first" and "second" in this article refer to multiple similar technical terms and distinguish them, and do not mean to limit the order or to limit to only two. For example, the first feature and the second feature refer to two categories / two features. The first feature can be one or more, and the second feature can also be one or more.

[0030] In addition, numerous specific details are provided in the following detailed description to better illustrate the present disclosure. Those skilled in the art will appreciate that the present disclosure can be practiced without certain specific details. In some instances, methods, means, components, and circuits well known to those skilled in the art are not described in detail in order to highlight the main points of the present disclosure.

[0031] The following describes the related technologies of the embodiments of the present disclosure. The following related technologies are optional solutions that can be combined with the technical solutions of the embodiments of the present disclosure in any way, and all of them fall within the protection scope of the embodiments of the present disclosure.

[0032] The recommendation process of the recommendation system corresponding to the client is as follows:

[0033] First, the client sends a resource recommendation request to the resource aggregation module, which then sends the request to the downstream queue module. Each queue in the queue module adopts a corresponding recall strategy based on the resource's attributes (different queues can have different resource attributes, and resources with different attributes can also have different recall strategies). The sorting module then performs coarse sorting, fine sorting, and re-ranking on the recall results to score the resources in the recall results. The scoring results are then returned to the resource aggregation module.

[0034] Secondly, the resource aggregation module deduplicates, merges, filters, and sorts the resources returned by each queue, generates a list of resources that meets the object's interests, and sends it to the client.

[0035] Finally, display the resources in the resource list in order.

[0036] In the aforementioned recommendation scenarios, mainstream ranking modules in the industry often use supervised deep learning methods to score resources. The effectiveness of deep learning depends crucially on the scale and quality of data samples. In other words, the ranking module's deep learning model requires massive amounts of historical behavior and resource data for training to improve scoring accuracy. However, for less active objects in the recommendation system, the sparsity of their historical behavior data prevents the model from fully learning their behavior, resulting in lower accuracy in the estimated scores, impacting recommendation effectiveness and even reducing the user experience.

[0037] Based on this, the disclosed solution provides a data processing method that can generate new data for the less active object based on the historical data of the less active object, thereby increasing the data volume of the sample data of the less active object.

[0038] Specifically, Figure 1 This is a schematic flow chart of a data processing method according to an embodiment of the present application. Figure 1 The method may be optionally applied to electronic devices, such as personal computers, servers, server clusters, and other electronic devices.

[0039] Furthermore, the method includes at least part of the following contents. Figure 1 Shown, including:

[0040] Step S101: Acquire initial sample data of a target object; the initial sample data represents the interest level of the target object in first resource data.

[0041] In one example, the initial sample data may be historical behavior data of the target object.

[0042] Step S102: Based on the initial sample data, N second resource data that the target object is interested in are estimated.

[0043] Here, N is a positive integer greater than or equal to 1. In other words, the first or second resource data can be estimated based on the initial sample data.

[0044] It should be pointed out that, in one example, the resource type of the second resource data is different from the resource type of the first resource data. In this way, on the basis of effectively increasing the amount of sample data for the target object, the resource type of the resource data for the target object can be further enriched, thereby enriching the data expression dimension of the sample data for the target object.

[0045] Furthermore, in another example, the resource types of different second resource data may be the same or different, and the disclosed solution does not impose any limitation on this.

[0046] Here, it can be understood that in order to further enrich the resource types in the sample data, the resource types of different second resource data may also be different.

[0047] Step S103: Obtain N newly added sample data of the target object based on the object feature of the target object and the resource feature of each second resource data of the N second resource data.

[0048] Here, the newly added sample data in the N newly added sample data can represent the interest level of the target object in the second resource data.

[0049] In this way, the disclosed solution can make full use of the initial sample data of the target object to expand and obtain new sample data for the target object. In other words, the disclosed solution provides a specific solution for increasing sample data, thereby effectively increasing the amount of sample data for the target object, thereby providing strong support for effectively solving the recommendation problem of low-activity objects.

[0050] In one example, the initial sample data described in the disclosed solution can be the historical behavior data of the target object, and the newly added sample data obtained by the disclosed solution can be understood as the actual behavior data of the non-target object, which is estimated virtual data. In this way, the problem of insufficient sample data of low-activity objects can be effectively solved.

[0051] Specifically, Figure 2 This is a schematic flow chart of a data processing method according to an embodiment of the present application. Figure 2 The method can be optionally applied to electronic devices, such as personal computers, servers, server clusters and other electronic devices. It is understood that the above Figure 1 The relevant contents of the method shown can also be applied to this example, and this example will not elaborate on the relevant contents.

[0052] Furthermore, the method includes at least part of the following contents. Figure 2 Shown, including:

[0053] Step S201: When it is determined that the total amount (e.g., the total number of items) of sample data (e.g., historical behavior data) for the target object is less than a preset number, initial sample data is selected from all sample data of the target object. Here, the initial sample data represents the interest level of the target object in the first resource data.

[0054] For example, a piece of data is randomly selected from all the historical behavior data of the target object as the initial sample data. It is understandable that this selection method is only for illustrative purposes and the present disclosure does not impose any specific restrictions on the selection method.

[0055] For example, in one example, the target object is a low-activity object. Accordingly, the initial sample data is the historical behavior data of the low-activity object. In this case, due to the small amount of historical behavior data of the low-activity object, the recommendation accuracy is low during the recommendation process, which reduces the user experience. Based on this, the disclosed solution can effectively increase the amount of sample data for the low-activity object.

[0056] In a specific example, the initial sample data is positive sample data. Accordingly, the newly added sample data obtained based on the initial sample data can also be specifically positive sample data. In this way, the disclosed solution can effectively increase the amount of positive sample data.

[0057] Alternatively, in another specific example, the initial sample data is negative sample data, and accordingly, the newly added sample data obtained based on the initial sample data can also be specifically negative sample data; in this way, the disclosed solution can effectively increase the amount of negative sample data.

[0058] Alternatively, in another specific example, the initial sample data is positive sample data, and accordingly, the newly added sample data obtained based on the initial sample data can also be specifically negative sample data; in this way, the disclosed solution can effectively increase the amount of negative sample data.

[0059] Alternatively, in another specific example, the initial sample data is negative sample data, and accordingly, the newly added sample data obtained based on the initial sample data can also be specifically positive sample data; in this way, the disclosed solution can effectively increase the amount of positive sample data.

[0060] It should be noted that positive sample data can be understood as: the target object is interested in the first resource data. Conversely, negative sample data can be understood as: the target object is not interested in the first resource data.

[0061] In this way, the disclosed solution provides a flexible solution for generating new sample data. For example, new positive sample data or new negative sample data can be generated based on the initial sample data. The process is simple, flexible and efficient, and further supports the increase in the amount of sample data for the target object. At the same time, it also provides strong support for further effectively solving the recommendation problem of low-activity objects.

[0062] Step S202: Based on the initial sample data, estimate and obtain N second resource data that the target object is interested in. Here, N is a positive integer greater than or equal to 1.

[0063] It should be noted that the relevant description about the second resource data can be found in the above description and will not be repeated here.

[0064] Step S203: Obtain N newly added sample data of the target object based on the object feature of the target object and the resource feature of each second resource data of the N second resource data.

[0065] Here, the newly added sample data in the N newly added sample data can represent the interest level of the target object in the second resource data.

[0066] For example, in one example, the resource characteristics of the second resource data can be spliced ​​with the object characteristics of the target object to obtain new sample data. At this time, the spliced ​​new sample data can represent the degree of interest of the target object in the second resource data, for example, representing that the target object is interested in the second resource data, or is not interested in the second resource data.

[0067] It should be pointed out that the initial sample data can also be understood as the data obtained by splicing the object characteristics of the target object and the resource characteristics of the first resource data, and can represent the degree of interest of the target object in the first resource data, for example, representing whether the target object is interested in the first resource data, or is not interested in the first resource data.

[0068] Thus, the disclosed solution provides a specific application scenario, namely, when it is determined that the total amount of sample data for the target object is less than a preset amount, the disclosed solution is used to increase the sample data. This solves the problem of insufficient sample data and inaccurate recommendations for low-activity objects. At the same time, it further enhances the intelligence level of the disclosed solution, and the process is simple and efficient, effectively avoiding the problem of wasted resources caused by adding samples when no additional samples are needed.

[0069] In a specific example of the disclosed solution, the second resource data may be estimated in the following manner. Specifically, the above-described method of estimating N second resource data of interest to the target object based on the initial sample data (e.g., step S102 or step S202) specifically includes:

[0070] Step 1-1: Based on the initial sample data, obtain the object features of the target object.

[0071] That is, the initial sample data includes the object features of the target object. In one example, the object features of the target object may be any features associated with the target object, such as the target object's behavior features, browsing features, and operation features, and the present disclosure does not impose any specific restrictions on this.

[0072] Step 1-2: Based on the object features of the target object, estimate and obtain N second resource data that the target object is interested in.

[0073] In this way, the disclosed solution can use the object features of the target object carried in the initial sample data to estimate the second resource data that the target object is sensitive to. This method is simple, efficient, and highly interpretable, and can provide support for increasing the amount of sample data for the target object. At the same time, it also provides strong support for further effectively solving the recommendation problem of low-activity objects.

[0074] Furthermore, in a specific example, the above-mentioned estimating and obtaining N second resource data of interest to the target object based on the object features of the target object (for example, step S1-2) may specifically include:

[0075] Step 1-2-1: Based on the object features of the target object, determine the target category to which the target object belongs.

[0076] Here, in a specific example, the target category to which the target object belongs may specifically refer to the group to which the target object belongs, or further refer to the attribute characteristics of the group to which the target object belongs.

[0077] Step 1-2-2: Based on the target category to which the target object belongs, estimate and obtain N second resource data that the target object is interested in.

[0078] In other words, we can first determine the category to which the target object's features belong, i.e., the target category, based on the target object's object features contained in the initial sample data. Then, based on the target category to which the target object belongs, we can estimate the resource data that the target object is interested in. This way, we can leverage the common preferences among objects belonging to the same category to estimate the second resource data that the target object is interested in, thereby laying the foundation for improving recommendation accuracy.

[0079] In this way, the disclosed solution can utilize the object characteristics of the target object carried in the initial sample data to obtain the classification to which the object belongs, such as the group to which it belongs, and then estimate the second resource data that the target object is sensitive to based on the group to which the target object belongs. This method is simple, efficient, and highly interpretable, and provides strong support for increasing the amount of sample data for the target object. At the same time, it also provides strong support for further effectively solving the recommendation problem of low-activity objects.

[0080] Furthermore, in one example, the above-described method of estimating N second resource data of interest to the target object based on the target classification to which the target object belongs (e.g., step 1-2-2) specifically includes:

[0081] Step 1-2-2-1: Obtain a first mapping relationship between a preset category and a preset extended resource type.

[0082] That is, the first mapping relationship represents the correspondence between the preset classification and the preset extended resource type.

[0083] Here, the preset classification can be understood to specifically refer to the group to which the object belongs, or further to the attribute characteristics of the group to which the object belongs. Accordingly, the preset extended resource type can specifically refer to the inferred resource type of resource data that the preset classification (e.g., a certain group) may be interested in. Accordingly, in one example, the first mapping relationship can specifically represent the mapping relationship between the attribute characteristics of the group and the inferred resource type of resource data that the group may be interested in.

[0084] That is to say, the first mapping relationship can characterize the mapping relationship between the preset classification and the extended resource type. In this way, after obtaining the target classification of the target object, the first mapping relationship can be used to determine the extended resource type corresponding to the target classification of the target object, that is, to determine the resource type that the target object may be interested in, thereby laying the foundation for quickly estimating the second resource data that the target object is interested in.

[0085] Here, it should be noted that the preset object and the preset extended resource type can be a one-to-one relationship, or a one-to-many relationship. For example, the first mapping relationship records an extended resource type corresponding to the preset object, or records two or more extended resource types corresponding to the preset object. The present disclosure does not impose any restrictions on this.

[0086] Here, in one example, the first mapping relationship may be pre-established in the following manner, i.e., the above-mentioned acquisition of the first mapping relationship between the preset classification and the preset extended resource type (e.g., the above-mentioned step 1-2-2-1), including:

[0087] Step 1-2-2-1-1: Input the preset classification and the common resource types corresponding to the preset classification into the large language model to infer at least one preset extended resource type for the preset classification.

[0088] Here, it should be pointed out that the disclosed solution does not limit the number of common resource types input into the large language model. For example, the common resource type input into the large language model can be one or more.

[0089] Furthermore, in one example, the common resource types corresponding to the preset classification can be obtained in the following manner: within a preset time period, the resource types of resource data that multiple objects belonging to the preset classification are interested in are counted to obtain the common preferences of multiple objects belonging to the preset classification, that is, the common interest types.

[0090] Step 1-2-2-1-2: Based on a preset classification and at least one target extended resource type for the preset classification, obtain the first mapping relationship.

[0091] Furthermore, in one example, the above-mentioned inputting of the preset categories and the common resource types corresponding to the preset categories into the large language model may specifically include:

[0092] Based on the preset categories and the common resource types corresponding to the preset categories, a target prompt word question matching the preset prompt word template is generated; and the target prompt word question is input into the large language model.

[0093] Here, the preset prompt word template may specifically include relevant content such as preset categories, common interest types corresponding to preset categories, template examples, and special personalized needs. For example, in one example, the preset prompt word template may specifically include: the attribute characteristics of the group, the common interest types of the group, all candidate interest types, template examples (for example, since the group has attribute A, it may be interested in interest type A), and personalized needs (for example, inferring the interest types that the group may be interested in from the candidate interest types, and not including the common interest types of the group).

[0094] It is understood that the specific attributes of the preset categories and the common interest types of the preset categories in the preset prompt word template are variables. Thus, after a specific category is given, the variables of the preset prompt word template can be assigned values ​​based on the specific attributes of the category and the common interest types of the category to obtain the target prompt word question.

[0095] This ensures that the large language model searches and infers within a specified range, thereby effectively improving the accuracy of the inference results.

[0096] In this way, the disclosed solution utilizes the powerful language induction and reasoning capabilities of the large language model to infer the possible extended interest types of the preset classification, thereby laying the foundation for quickly estimating the second resource data that the target object is interested in. At the same time, it provides strong support for increasing the amount of sample data for the target object, thereby effectively solving the recommendation problem of low-activity objects.

[0097] Step 1-2-2-2: Based on the first mapping relationship, determine M target extended resource types corresponding to the target category to which the target object belongs. Here, M is a positive integer greater than or equal to 1.

[0098] Step 1-2-2-3: Based on the determined M target extended resource types, estimate and obtain N second resource data that the target object is interested in.

[0099] In this way, after obtaining the target classification of the target object, the first mapping relationship can be used to determine the extended resource type corresponding to the target classification of the target object, thereby laying the foundation for quickly estimating the second resource data of interest to the target object.

[0100] Furthermore, in one example, the above-described process of estimating N second resource data items of interest to the target object based on the determined M target extended resource types (e.g., steps 1-2-2-3) may specifically include:

[0101] Step 1-2-2-3-1: From the second mapping relationship between resource data and resource types, select at least one extended resource data corresponding to each preset extended resource type in the M target extended resource types.

[0102] For example, based on the target extended resource type, at least one extended resource data is randomly selected from the second mapping relationship. In this way, when M is greater than or equal to 2, at least M extended resource data can be randomly selected.

[0103] Step 1-2-2-3-2: Based on at least one extended resource data corresponding to each selected preset extended resource type, obtain N second resource data.

[0104] For example, in one example, all the determined extended resource data are used as the second resource data, thereby further increasing the amount of sample data for the target object, and further providing strong support for effectively solving the recommendation problem of low-activity objects.

[0105] In a specific example of the disclosed solution, after obtaining N new sample data for the target object, the new sample data can also be used for model training. This effectively solves the problem of inaccurate recommendations for low-activity objects and further improves the user experience. Specifically, at least the initial sample data for the target object and the N new sample data are input into the recommendation model to perform model training on the recommendation model.

[0106] The following combination Figure 3 、 Figure 4 and Figure 5 The disclosed solution is further described in detail. Specifically, Figure 3 As shown, the sample data processing method based on large language model reasoning specifically includes:

[0107] Step S301: Obtain a preset category.

[0108] Here, in actual applications, different objects may be classified based on specific attributes of the objects to obtain preset classifications. For example, different objects may be assigned to different groups based on the specific attributes of the objects.

[0109] Step S302: Count the resource types of all resource data. For example, record the resource types of all resource data as candidate classification set A.

[0110] Step S303: Count the frequencies of resource types of interest to all objects in each preset classification (e.g., each group) within a preset time period, and sort the resource types in descending order based on the frequencies to obtain a common interest set a for each preset classification (e.g., each group). For example, the common interest set a includes resource types ranked in the top X (e.g., 5, 10, etc.). The resource types ranked in the top X can be considered as the common resource types of the preset classification.

[0111] Step S304: performing large language model inference to infer a plurality of preset extended resource types that the object in each preset category is most likely to be interested in.

[0112] Here, a preset prompt word template can be constructed in advance. For example, the specific attributes of each preset category, the common interest set of the preset category, etc. can be used to design the preset prompt word template; at this time, the specific attributes of the preset category, the common interest set of the preset category, etc. can be used as variables of the preset prompt word template, and then after determining the specific category, the target prompt word question corresponding to the specific category can be obtained, that is, the target prompt word question corresponding to each preset category can be obtained, so as to ask questions to the large language model, so as to obtain the extended interest type that the preset category may be interested in through inference by the large language model.

[0113] Here, when designing the preset prompt word template, you can also add personalized requirements. For example, you can remind the large language model to filter out the common interest set a from the candidate classification set A. In this way, the extended interest types of the preset classification finally inferred are different from the common resource types included in the common interest set a. In other words, the extended interest types of the preset classification are selected from the remaining candidate sets (e.g., set A-set a).

[0114] For example, in a specific example, the preset prompt word template is:

[0115] Please simulate a recommendation system that can memorize and infer content that a preset category might be interested in. The preset category has specific attributes (e.g., age group). Given the set of shared interests a (e.g., entertainment, social news, etc.) that the preset category is most interested in, and a set of candidate categories A (e.g., all resource types that the recommendation system can promote), infer from the candidate set A the content that the subjects in the preset category are most likely to be interested in (e.g., the X most likely resource types). Note that the known shared interests set a must not be included.

[0116] Furthermore, taking group A as an example, the target prompt word problem of group A can be specifically as follows:

[0117] Imagine a recommendation system that memorizes and infers content that Group A might be interested in. Group A's age range is [a1-a2]. It's known that Group A's most common interests, set a, include entertainment and social news. The candidate category set A represents all resource types that the recommendation system can promote. Infer the content that Group A is most likely to be interested in from candidate category set A, but note that this set should not include the known common interests set a.

[0118] Furthermore, in a specific example, the output of the large language model may be:

[0119] Based on the attribute characteristics of group A and the known most interesting resource types, it is inferred from the candidate classification set A that group A may also be interested in the following:

[0120] Resource type 1: Since the group has attribute feature A, the group may also be interested in resource type 1.

[0121] Resource type 2: Since the group has attribute feature B, the group may also be interested in resource type 2.

[0122] It should be noted that after obtaining the inference results of the large language model, corresponding processing can be performed, such as formatting, to obtain multiple preset extended resource types that the objects in the preset classification are most likely to be interested in. For example, the output results of the above-mentioned large language model are processed to extract "resource type 1 and resource type 2", and the resource type 1 and resource type 2 are respectively used as the preset extended resource types of the group A inferred by the large language model.

[0123] Step S305: Obtain an interest classification table, which represents the mapping relationship between preset classifications and preset extended resource types (for example, the mapping relationship between the attribute characteristics of a group and the preset extended resource type). Also obtain a resource table, which represents the mapping relationship between resource data and resource types. Here, the resource table may also include resource characteristics of each resource data, such as resource title, cover image, content quality, etc.

[0124] Step S306: For low-activity objects, newly added sample data is obtained.

[0125] It should be noted that, in this example, for low-activity objects, the newly added sample data can be newly added positive sample data or newly added negative sample data. The generation logic of the two is similar, and the present disclosure does not impose specific restrictions on this.

[0126] The following is an example of generating new positive sample data:

[0127] Before training the ranking model, for low-activity objects, such as Figure 4 As shown, first, the positive sample data is obtained (for example, the low-activity object clicks on a certain video resource as the positive sample data). Here, the positive sample data may include object features and resource features of the clicked video resource.

[0128] Secondly, object features based on the positive sample data are obtained to determine the target category to which the low-activity object belongs, and based on a pre-built interest classification table, multiple target extended resource types corresponding to the target category are obtained.

[0129] Finally, based on the pre-constructed resource table, at least one resource data corresponding to the target extended resource type is selected (for example, randomly selected), and the selected resource data is used as the extended resource data corresponding to the low-activity object (that is, the second resource data mentioned above), and the resource characteristics of each extended resource data are spliced ​​with the object characteristics of the low-activity object to obtain multiple new sample data for the low-activity object.

[0130] Step S307: Based on the positive sample data of the low-activity object and the obtained newly added sample data, the positive sample data are used as training data to perform model training on the recommendation model, such as (the ranking model in the recommendation model).

[0131] In this way, the disclosed solution can train the recommendation model (such as the ranking model in the recommendation model) based on the expanded sample data of low-activity objects (i.e., the newly added sample data). In this way, it can effectively solve the problem of inaccurate recommendations of low-activity objects, thereby improving the user experience.

[0132] The present disclosure also provides a data processing device, such as Figure 6 Shown, including:

[0133] The sample acquisition unit 601 is configured to acquire initial sample data of a target object; the initial sample data represents the interest level of the target object in the first resource data;

[0134] The resource expansion unit 602 is configured to estimate N second resource data of interest to the target object based on the initial sample data; N is a positive integer greater than or equal to 1;

[0135] The sample generation unit 603 is used to obtain N new sample data of the target object based on the object characteristics of the target object and the resource characteristics of each second resource data in the N second resource data, wherein the new sample data in the N new sample data can represent the degree of interest of the target object in the second resource data.

[0136] In a specific example of the disclosed solution, the sample acquisition unit is specifically configured to:

[0137] When it is determined that the total amount of sample data for the target object is less than a preset amount, initial sample data is selected from all the sample data of the target object.

[0138] In a specific example of the disclosed solution, the initial sample data and the newly added sample data meet one of the following conditions:

[0139] The initial sample data is positive sample data, and accordingly, the newly added sample data obtained based on the initial sample data is positive sample data;

[0140] The initial sample data is negative sample data, and accordingly, the newly added sample data obtained based on the initial sample data is negative sample data;

[0141] The initial sample data is positive sample data, and correspondingly, the newly added sample data obtained based on the initial sample data is negative sample data;

[0142] The initial sample data is negative sample data, and accordingly, the newly added sample data obtained based on the initial sample data is positive sample data.

[0143] In a specific example of the disclosed solution, the resource extension unit is specifically configured to:

[0144] Based on the initial sample data, obtaining object features of the target object;

[0145] Based on the object features of the target object, N second resource data of interest to the target object are estimated.

[0146] In a specific example of the disclosed solution, the resource extension unit is specifically configured to:

[0147] Determining the target category to which the target object belongs based on the object characteristics of the target object;

[0148] Based on the target category to which the target object belongs, N second resource data that the target object is interested in are estimated.

[0149] In a specific example of the disclosed solution, the resource extension unit is specifically configured to:

[0150] Obtaining a first mapping relationship between a preset category and a preset extended resource type;

[0151] Based on the first mapping relationship, determining M target extended resource types corresponding to the target category to which the target object belongs; wherein M is a positive integer greater than or equal to 1;

[0152] Based on the determined M target extended resource types, N second resource data that the target object is interested in are estimated.

[0153] In a specific example of the disclosed solution, the resource extension unit is specifically configured to:

[0154] Selecting at least one extended resource data corresponding to each target extended resource type in the M target extended resource types from the second mapping relationship between resource data and resource types;

[0155] Based on the at least one extended resource data corresponding to each selected target extended resource type, N second resource data are obtained.

[0156] In a specific example of the disclosed solution, the resource extension unit is specifically configured to:

[0157] Inputting the preset classification and the common resource types corresponding to the preset classification into the large language model to infer at least one preset extended resource type for the preset classification;

[0158] The first mapping relationship is obtained based on a preset category and at least one preset extended resource type for the preset category.

[0159] In a specific example of the disclosed solution, the resource extension unit is specifically configured to:

[0160] Based on the preset categories and the common resource types corresponding to the preset categories, generating target prompt word questions that match the preset prompt word templates;

[0161] The target cue word question is input into the large language model.

[0162] In a specific example of the disclosed solution, the device further includes: a model training unit; wherein,

[0163] The model training unit is used to input at least the initial sample data of the target object and N newly added sample data into the recommendation model to perform model training on the recommendation model.

[0164] For the description of specific functions and examples of each unit of the device in the embodiment of the present disclosure, please refer to the relevant description of the corresponding steps in the above method embodiment, which will not be repeated here.

[0165] In the technical solutions disclosed herein, the acquisition, storage, and application of user personal information involved are all data that can be obtained after the user's authorization, comply with the provisions of relevant laws and regulations, and do not violate public order and good morals.

[0166] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0167] Figure 7 A schematic block diagram of an example electronic device 700 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0168] like Figure 7 As shown, the device 700 includes a computing unit 701, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 702 or a computer program loaded from a storage unit 708 into a random access memory (RAM) 703. Various programs and data required for the operation of the device 700 can also be stored in the RAM 703. The computing unit 701, the ROM 702, and the RAM 703 are connected to each other via a bus 704. An input / output (I / O) interface 705 is also connected to the bus 704.

[0169] Various components in device 700 are connected to I / O interface 705, including an input unit 706, such as a keyboard, mouse, etc.; an output unit 707, such as various types of displays, speakers, etc.; a storage unit 708, such as a magnetic disk, optical disk, etc.; and a communication unit 709, such as a network card, modem, wireless communication transceiver, etc. The communication unit 709 allows device 700 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0170] The computing unit 701 can be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 701 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units that run machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 701 performs the various methods and processes described above, such as the data processing method. For example, in some embodiments, the data processing method can be implemented as a computer software program that is tangibly contained in a machine-readable medium, such as the storage unit 708. In some embodiments, part or all of the computer program can be loaded and / or installed on the device 700 via the ROM 702 and / or the communication unit 709. When the computer program is loaded into the RAM 703 and executed by the computing unit 701, one or more steps of the data processing method described above can be performed. Alternatively, in other embodiments, the computing unit 701 can be configured to perform the data processing method by any other appropriate means (e.g., by means of firmware).

[0171] Various embodiments of the systems and techniques described herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), system-on-chip systems (SOCs), programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include being implemented in one or more computer programs that are executable and / or interpreted on a programmable system comprising at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0172] The program code for implementing the method of the present disclosure can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device so that when the program code is executed by the processor or controller, the functions / operations specified in the flow chart and / or block diagram are implemented. The program code can be executed entirely on the machine, partially on the machine, as a stand-alone software package, partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0173] In the context of the present disclosure, a machine-readable medium can be a tangible medium that can contain or store a program for use by or in conjunction with an instruction execution system, device or equipment. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium can include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0174] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0175] The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (LAN), a wide area network (WAN), and the Internet.

[0176] A computer system may include a client and a server. The client and server are generally remote from each other and typically interact through a communication network. The client-server relationship arises through computer programs running on the respective computers and having a client-server relationship with each other. The server may be a cloud server, a server in a distributed system, or a server integrated with a blockchain.

[0177] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this disclosure can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved. This is not limited herein.

[0178] The above specific embodiments do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the principles of this disclosure shall be included within the scope of protection of this disclosure.

Claims

1. A data processing method, comprising: Get the initial sample data of the target object; The initial sample data represents the target object's interest in the first resource data; Based on the initial sample data, N second resource data of interest to the target object are estimated; N is a positive integer greater than or equal to 1; Based on the object characteristics of the target object and the resource characteristics of each second resource data in the N second resource data, N new sample data of the target object are obtained, wherein the new sample data in the N new sample data can represent the target object's interest level in the second resource data.

2. The method according to claim 1, wherein The obtaining of initial sample data of the target object includes: When it is determined that the total amount of sample data for the target object is less than a preset amount, initial sample data is selected from all the sample data of the target object.

3. The method according to claim 1 or 2, wherein The initial sample data and the newly added sample data meet one of the following conditions: The initial sample data is positive sample data, and accordingly, the newly added sample data obtained based on the initial sample data is positive sample data; The initial sample data is negative sample data, and accordingly, the newly added sample data obtained based on the initial sample data is negative sample data; The initial sample data is positive sample data, and correspondingly, the newly added sample data obtained based on the initial sample data is negative sample data; The initial sample data is negative sample data, and accordingly, the newly added sample data obtained based on the initial sample data is positive sample data.

4. The method according to any one of claims 1 to 3, wherein: The estimating N second resource data of interest to the target object based on the initial sample data includes: Based on the initial sample data, obtaining object features of the target object; Based on the object features of the target object, N second resource data of interest to the target object are estimated.

5. The method according to claim 4, wherein The estimating N second resource data of interest to the target object based on the object feature of the target object includes: Determining the target category to which the target object belongs based on the object characteristics of the target object; Based on the target category to which the target object belongs, N second resource data that the target object is interested in are estimated.

6. The method according to claim 5, wherein: The estimating N second resource data that the target object is interested in based on the target classification to which the target object belongs includes: Obtaining a first mapping relationship between a preset category and a preset extended resource type; Based on the first mapping relationship, determining M target extended resource types corresponding to the target category to which the target object belongs; wherein M is a positive integer greater than or equal to 1; Based on the determined M target extended resource types, N second resource data that the target object is interested in are estimated.

7. The method according to claim 6, wherein: The estimating N second resource data of interest to the target object based on the determined M target extended resource types includes: Selecting at least one extended resource data corresponding to each target extended resource type in the M target extended resource types from the second mapping relationship between resource data and resource types; Based on the at least one extended resource data corresponding to each selected target extended resource type, N second resource data are obtained.

8. The method according to claim 6 or 7, wherein: The obtaining of a first mapping relationship between a preset classification and a preset extended resource type includes: Inputting the preset classification and the common resource types corresponding to the preset classification into the large language model to infer at least one preset extended resource type for the preset classification; The first mapping relationship is obtained based on a preset category and at least one preset extended resource type for the preset category.

9. The method according to claim 8, wherein Inputting the preset categories and the common resource types corresponding to the preset categories into the large language model includes: Based on the preset categories and the common resource types corresponding to the preset categories, generating target prompt word questions that match the preset prompt word templates; The target cue word question is input into the large language model.

10. The method according to any one of claims 1 to 9, further comprising: At least the initial sample data of the target object and N newly added sample data are input into the recommendation model to perform model training on the recommendation model.

11. A data processing device comprising: A sample acquisition unit, used to acquire initial sample data of the target object; The initial sample data represents the target object's interest in the first resource data; A resource expansion unit, configured to estimate N second resource data of interest to the target object based on the initial sample data; wherein N is a positive integer greater than or equal to 1; A sample generation unit is used to obtain N new sample data of the target object based on the object characteristics of the target object and the resource characteristics of each second resource data in the N second resource data, wherein the new sample data in the N new sample data can represent the degree of interest of the target object in the second resource data.

12. The device according to claim 11, wherein The sample acquisition unit is specifically used to: When it is determined that the total amount of sample data for the target object is less than a preset amount, initial sample data is selected from all the sample data of the target object.

13. The device according to claim 11 or 12, wherein: The initial sample data and the newly added sample data meet one of the following conditions: The initial sample data is positive sample data, and accordingly, the newly added sample data obtained based on the initial sample data is positive sample data; The initial sample data is negative sample data, and accordingly, the newly added sample data obtained based on the initial sample data is negative sample data; The initial sample data is positive sample data, and correspondingly, the newly added sample data obtained based on the initial sample data is negative sample data; The initial sample data is negative sample data, and accordingly, the newly added sample data obtained based on the initial sample data is positive sample data.

14. The device according to any one of claims 11 to 13, wherein: The resource extension unit is specifically used to: Based on the initial sample data, obtaining object features of the target object; Based on the object features of the target object, N second resource data of interest to the target object are estimated.

15. The device according to claim 14, wherein The resource extension unit is specifically used to: Determining the target category to which the target object belongs based on the object characteristics of the target object; Based on the target category to which the target object belongs, N second resource data that the target object is interested in are estimated.

16. The device according to claim 15, wherein The resource extension unit is specifically used to: Obtaining a first mapping relationship between a preset category and a preset extended resource type; Based on the first mapping relationship, determining M target extended resource types corresponding to the target category to which the target object belongs; wherein M is a positive integer greater than or equal to 1; Based on the determined M target extended resource types, N second resource data that the target object is interested in are estimated.

17. The device according to claim 16, wherein The resource extension unit is specifically used to: Selecting at least one extended resource data corresponding to each target extended resource type in the M target extended resource types from the second mapping relationship between resource data and resource types; Based on the at least one extended resource data corresponding to each selected target extended resource type, N second resource data are obtained.

18. The device according to claim 16 or 17, wherein The resource extension unit is specifically used to: Inputting the preset classification and the common resource types corresponding to the preset classification into the large language model to infer at least one preset extended resource type for the preset classification; The first mapping relationship is obtained based on a preset category and at least one preset extended resource type for the preset category.

19. The device according to claim 18, wherein The resource extension unit is specifically used to: Based on the preset categories and the common resource types corresponding to the preset categories, generating target prompt word questions that match the preset prompt word templates; The target cue word question is input into the large language model.

20. The device according to any one of claims 11 to 19, further comprising: Model training unit; where, The model training unit is used to input at least the initial sample data of the target object and N newly added sample data into the recommendation model to perform model training on the recommendation model.

21. An electronic device comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 10.

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

23. A computer program product comprising a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 10.

Images