Data processing method and related device

Abstract

The invention discloses a data processing method and a related device, and the method comprises the steps: firstly obtaining object interaction data which identifies an interaction behavior between a target object and content, and generating N interaction data sequences of the target object according to the object interaction data and N interaction behavior types; the interaction data sequence and the interaction behavior type have a corresponding relation, and the interaction behavior type is determined according to the interaction frequency. And when a data acquisition request for the description set of the target object is acquired from the recommendation server, selecting a corresponding basic data sequence from the N interactive data sequences according to data demand parameters of the description set carried by the data acquisition request, and returning the basic data sequence for generating the description set to the recommendation server. The interaction type can be determined through the number of times of interaction, and the interaction sequence obtained based on the interaction type can adapt to most description set construction requirements for recommendation, so that the storage cost of basic data is reduced.

Description

Technical Field

[0001] The present application relates to the field of data processing, and in particular to a data processing method and related devices. Background Art

[0002] In the recommendation scenario, content recommendation needs to be made based on the description set of the target object. The description set may include a comprehensive description of the target object's interests, preferences, behavior patterns, and characteristics. Therefore, content recommendation for the target object based on the description set is more targeted.

[0003] In related technologies, when determining a description set for a target object, different types of target object interaction data must be prepared based on different description requirements. Different description requirements correspond to different types of target object interaction data. In this case, each type of interaction data is captured and stored specifically based on the description requirements.

[0004] Therefore, if the above method of determining the description set of the target object is adopted, a large amount of interaction data will need to be stored, which greatly increases the storage cost. Summary of the Invention

[0005] In order to solve the above technical problems, the present application provides a data processing method and related devices, which can determine the interaction type by the number of interactions. The interaction sequence obtained based on the interaction type can adapt to most of the description set construction requirements for recommendation, thereby reducing the storage cost of basic data.

[0006] The embodiments of this application disclose the following technical solutions:

[0007] In one aspect, an embodiment of the present application provides a data processing method, the method comprising:

[0008] Acquire object interaction data of a target object, where the object interaction data is used to identify an interaction behavior between the target object and content;

[0009] Generate N interaction data sequences for the target object based on the object interaction data and N interaction behavior types, wherein the interaction data sequences correspond to the interaction behavior types in a one-to-one manner, and the N interaction behavior types are determined based on the number of interactions, with different interaction behavior types corresponding to different numbers of interactions. The number of interactions is used to identify the number of interactions of the target object with respect to the same content, where N>1.

[0010] When a data acquisition request for the description set of the target object is obtained from the recommendation server, a corresponding basic data sequence is selected from the N interaction data sequences according to the data requirement parameter of the description set carried in the data acquisition request, where the basic data sequence is at least one of the N interaction data sequences;

[0011] The basic data sequence used to generate the description set is returned to the recommendation server.

[0012] On the other hand, an embodiment of the present application provides another data processing method, the method comprising:

[0013] When starting the recommendation service, according to the description set required by the recommendation service, a data acquisition request for the description set of the target object is sent to the data server, wherein the data acquisition request carries the data requirement parameters of the description set;

[0014] Obtaining a basic data sequence from the data server, the basic data sequence being used to generate a description set of the target object, wherein, for the target object, the basic data sequence is selected from N interaction data sequences according to the data requirement parameters, the basic data sequence being at least one of the N interaction data sequences, the N interaction data sequences being generated according to object interaction data of the target object and N interaction behavior types, the interaction data sequences corresponding to the interaction behavior types being one-to-one, the N interaction behavior types being determined according to a number of interactions, with different interaction behavior types corresponding to different numbers of interactions, the number of interactions being used to identify the number of interaction behaviors of the target object with respect to the same content, where N>1;

[0015] The description set is constructed according to the basic data sequence, and the description set is used to provide content recommendation service for the target object.

[0016] On the other hand, an embodiment of the present application provides a data processing device, the device comprising: an acquisition module, a generation module, a selection module, and a return module;

[0017] The acquisition module is used to acquire object interaction data of the target object, where the object interaction data is used to identify the interaction behavior between the target object and the content;

[0018] The generating module is configured to generate N interaction data sequences of the target object based on the object interaction data and N interaction behavior types, wherein the interaction data sequences correspond to the interaction behavior types in a one-to-one manner, and the N interaction behavior types are determined based on the number of interactions, with different interaction behavior types corresponding to different numbers of interactions. The number of interactions is used to identify the number of interactions of the target object with respect to the same content, where N>1;

[0019] The selection module is configured to, when receiving a data acquisition request for the description set of the target object from the recommendation server, select a corresponding basic data sequence from the N interaction data sequences based on the data requirement parameter of the description set carried in the data acquisition request, where the basic data sequence is at least one of the N interaction data sequences;

[0020] The returning module is configured to return the basic data sequence used to generate the description set to the recommendation server.

[0021] On the other hand, an embodiment of the present application provides a data processing device, the device comprising: a sending module, a generating module, and a constructing module;

[0022] The sending module is configured to send a data acquisition request for the description set of the target object to the data server according to the description set required by the recommendation service when the recommendation service is started, wherein the data acquisition request carries the data requirement parameters of the description set;

[0023] The generation module is configured to obtain a basic data sequence from the data server, the basic data sequence being used to generate a description set of the target object, wherein, for the target object, the basic data sequence is selected from N interaction data sequences based on the data requirement parameters, the basic data sequence being at least one of the N interaction data sequences, the N interaction data sequences being generated based on the object interaction data of the target object and N interaction behavior types, the interaction data sequences corresponding to the interaction behavior types being one-to-one, the N interaction behavior types being determined based on the number of interactions, with different interaction behavior types corresponding to different numbers of interactions, the number of interactions being used to identify the number of interactions of the target object with respect to the same content, where N>1;

[0024] The construction module is used to construct the description set according to the basic data sequence, and the description set is used to provide content recommendation service for the target object.

[0025] In another aspect, an embodiment of the present application provides a computer device, comprising a processor and a memory:

[0026] Memory is used to store computer programs;

[0027] The processor is configured to execute the above-described method according to the computer program.

[0028] In another aspect, an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium is used to store a computer program, and the computer program is used to execute the method described in the above aspects.

[0029] On the other hand, an embodiment of the present application provides a computer program product including a computer program, which, when executed on a computer device, enables the computer device to execute the method described in the above aspects.

[0030] As can be seen from the above technical solution, object interaction data identifying the interactions between a target object and content is divided into N interaction data sequences based on N interaction behavior types. The interaction behavior type is determined based on the number of interactions, which identifies the number of interactions a target object has with the same content. Therefore, N different interaction behavior types can be divided based on the number of interactions. The interaction data sequences corresponding to each interaction behavior type identify which content the target object is willing to interact with more or less frequently. When constructing a description set for content recommendation for a target object, regardless of the content recommendation requirement, the target object is expected to interact with a certain number of times. Therefore, when constructing the data requirement parameters for the description set, the recommendation server can select an interaction data sequence from the N interaction data sequences that meets the requirements as the base data sequence to be returned. Thus, by processing the object interaction data based on the number of interactions, the N interaction data sequences generated can adapt to the generation requirements of different description sets, providing universality. Only N interaction data sequences need to be stored, effectively conserving storage resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0032] Figure 1 A schematic diagram of generating a description set in a related technology provided in an embodiment of the present application;

[0033] Figure 2 A schematic diagram of a data processing scenario provided in an embodiment of the present application;

[0034] Figure 3 A flow chart of a data processing method provided in an embodiment of the present application;

[0035] Figure 4 A schematic diagram of data sequence construction in an application scenario provided in an embodiment of the present application;

[0036] Figure 5A schematic diagram of data acquisition based on log storage provided in an embodiment of the present application;

[0037] Figure 6 A flow chart of another data processing method provided in an embodiment of the present application;

[0038] Figure 7 A schematic diagram of a recommendation process in an application scenario provided by an embodiment of the present application;

[0039] Figure 8 A schematic diagram of the structure of a server provided in an embodiment of the present application;

[0040] Figure 9 An interactive diagram of a data processing method provided in an embodiment of the present application;

[0041] Figure 10 A schematic diagram of a data processing device provided in an embodiment of the present application;

[0042] Figure 11 A schematic diagram of another data processing device provided in an embodiment of the present application;

[0043] Figure 12 A structural diagram of a terminal device provided in an embodiment of the present application;

[0044] Figure 13 A structural diagram of a server provided in an embodiment of the present application. DETAILED DESCRIPTION

[0045] The embodiments of the present application are described below with reference to the accompanying drawings.

[0046] In recommendation scenarios, the target object's description set is used as a basis for content recommendations. The description set mentioned above refers to a comprehensive set of descriptions of the target object's interests, preferences, behavior patterns, and characteristics. This description set facilitates the determination of the target object's content preferences, allowing for targeted content recommendations to be made to the target object, improving recommendation efficiency.

[0047] In the related art, when constructing a description set of a target object, a large amount of different types of object interaction data is often required as basic data. For example, assuming that it is necessary to construct a description set of target objects A, target object B, and target object C, for target object A, type 1 and type 2 object interaction data need to be used as basic data for constructing the description set; for target object B, type 3 and type 4 object interaction data need to be used as basic data for constructing the description set; for target object C, type 5 and type 6 object interaction data need to be used as basic data for constructing the description set. From the above example, it can be seen that when constructing the description set of three target objects, 6 types of object interaction data will be involved, which means that all 6 types of object interaction data need to be stored.

[0048] As another example, let's continue with the construction of the description sets of target objects A, B, and C. For target object A, type 1 and type 2 object interaction data need to be used as basic data for the construction of the description set; for target object B, type 1 and type 4 object interaction data need to be used as basic data for the construction of the description set; and for target object C, type 5 and type 6 object interaction data need to be used as basic data for the construction of the description set. At this time, corresponding data preparation is required for the construction requirements of different description sets. The construction of the description sets of target objects A and B requires the use of type 1 object interaction data, which means that the type 1 object interaction data will be stored twice (the object interaction data corresponding to the construction of different description sets is dedicated storage). At the same time, in actual applications, the order of magnitude of the target objects is very large, and the order of magnitude of the corresponding object interaction data that needs to be stored will increase exponentially, and redundant storage will also be generated.

[0049] Figure 1 This is a schematic diagram of generating a description set in a related technology provided in an embodiment of the present application, see Figure 1 As shown, during the construction process of description set 1 and description set 2, object interaction data needs to be accessed, and the object interaction data is not divided into types, but directly input into the construction process of the description set. Each time a description set is constructed, all object interaction data needs to be loaded and stored. Therefore, the order of magnitude of the object interaction data that needs to be stored will increase exponentially, and redundant storage will inevitably be generated in the process of storing the object interaction data.

[0050] Generally speaking, in related technologies, because different target objects involve different types of object interaction data, when constructing a description set for a target object, it is necessary to store all the object interaction data involved to accommodate the construction requirements of the description sets for different target objects. This results in huge storage costs and a significant increase in storage space.

[0051] To this end, an embodiment of the present application provides a data processing method and related devices, which determine the interaction type by the number of interactions. The interaction sequence obtained based on the interaction type can adapt to most of the description set construction requirements for recommendation, thereby reducing the storage cost of basic data.

[0052] The data processing method provided in the embodiments of the present application can be implemented by a computer device, which can be a terminal device or a server. The server can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services. Terminal devices include but are not limited to mobile phones, computers, intelligent voice interaction devices, smart home appliances, vehicle-mounted terminals, aircraft, etc. The terminal device and the server can be directly or indirectly connected via wired or wireless communication, which is not limited in this application.

[0053] The collection and processing of relevant data (such as object interaction data) in this application should be strictly in accordance with the requirements of relevant national laws and regulations when applied in practice, and the informed consent or separate consent of the personal information subject (such as the target object) should be obtained. Subsequent data use and processing should be carried out within the scope of authorization of laws and regulations and the personal information subject.

[0054] First, several noun terms that may be involved in the embodiments below in this application are explained.

[0055] Recommendation server: Provides personalized recommendations or suggestions based on the target's interests and preferences. It uses the target's basic data sequence, personal information, and other contextual information (which can be understood as the recommendation environment or recommendation scenario) to analyze and mine the basic data sequence to predict the content that the target may be interested in.

[0056] Description set of target object: The description set of target object is a data set that describes the characteristics and behaviors of the target object. It is mainly determined by analyzing the basic data sequence of the target object to determine the interests, preferences, behavior patterns and characteristics of the target object. The description set is constructed based on the recommendation scenario. The description set includes multiple descriptive features. The descriptive features are used to personalize the target object's interactive behavior towards the content in the interactive scenario involved in the recommendation service. Taking the scenario in the video application as an example, the description set of the target object in this scenario is constructed by analyzing the basic data sequence of the target object in the video application. By constructing the description set of the target object, you can understand the target object's areas of interest in video.

[0057] Forward index: A forward index is a data structure that uses an item's unique identifier (such as a video ID) as a key to record various item characteristics. For example, a forward index uses the video ID as a key, allowing users to quickly find detailed information such as the video's title, duration, and tags.

[0058] Recommendation Service: An internal module within the recommendation server, the recommendation service is responsible for acquiring, processing, and providing target object description data to support personalized recommendation tasks within the recommendation algorithm model. The recommendation service retrieves target object description data from storage (such as Redis). Algorithms such as recall, coarse ranking, and fine ranking retrieve the corresponding target object description data based on the target object's identifier (e.g., its ID).

[0059] Figure 2 A schematic diagram of a data processing scenario provided in an embodiment of the present application, wherein the aforementioned computer device is a data server.

[0060] In the recommendation scenario, a data server can be used as the provider of the basic data (i.e., the basic data sequence) in the process of generating the description set. The data server involves six actions that need to be performed during this process: ① Acquire; ② Generate; ③ Receive; ④ Select; ⑤ Obtain; ⑥ Return.

[0061] The following describes each of these actions in detail. First, the data server needs to obtain the target object's object interaction data. Then, based on the object interaction data and interaction behavior type, it generates an interaction data sequence for the target object. It should be noted that the object interaction data mentioned above refers to the interaction behavior between the target object and the content. There is a corresponding relationship between the interaction data sequence and the interaction behavior type, which is determined by the number of interactions.

[0062] The recommendation server sends a data acquisition request for the description set to the data server. When the data server receives the data acquisition request, it selects from the interactive data sequence based on the data acquisition request and obtains the basic data sequence. Finally, it returns the basic data sequence that generates the description set of the target object to the recommendation server.

[0063] By classifying object interaction data into interaction data sequences based on interaction behavior type, which is determined by the number of interactions, which identifies the number of times a target object interacts with the same content, this allows for the classification of interaction behavior types based on the number of interactions. Interaction data sequences classified based on the number of interactions can adapt to the generation requirements of different description sets. Furthermore, the number of interaction data sequences is limited, thus conserving storage resources.

[0064] In the embodiment of the present application, the data server and the recommendation server may be two independent servers respectively set up. In addition, the data server and the recommendation server may be integrated into one device.

[0065] Figure 3 This is a method flow chart of a data processing method provided in an embodiment of the present application. The method can be executed by a computer device. In this embodiment, the computer device is taken as an example of a server.

[0066] The method comprises:

[0067] S201: Acquire object interaction data of a target object.

[0068] The target object can be understood as the person for whom content recommendations are intended, typically referring to a user of an application, product, or similar application. Object interaction data identifies the interactions between the target object and the content. Specific interactions between the target object and the content include, but are not limited to, browsing, clicking, and commenting. When the target object interacts with content, corresponding object interaction data is generated. For example, if user A clicks on content B, a corresponding piece of object interaction data for user A with content B will be generated.

[0069] Content refers to the data content of applications or products recommended to the target object in the recommendation scenario. For example, content can refer to music, videos, pictures, or product links, etc. The purpose of obtaining the target object's object interaction data is to serve as the data source for generating the interaction data sequence.

[0070] Object interaction data can carry the following key information:

[0071] (1) Target object identification: Each target object will have a unique identifier to distinguish different target objects.

[0072] (2) Interaction behavior type: records the type of specific interaction behavior between the target object and the content, such as browsing, playing, liking, commenting, etc.

[0073] (3) Behavior object: records the objects involved in the target object’s interactive behavior, such as pages browsed, videos played, and content liked.

[0074] (4) Timestamp information: records the time when the target object’s interactive behavior occurs, which is used for the subsequent generation of description sets and personalized content recommendations.

[0075] S202: Generate N interaction data sequences of the target object according to the object interaction data and N interaction behavior types.

[0076] The above-mentioned interaction data sequence corresponds to the interaction behavior type one-to-one. The N interaction behavior types are determined based on the number of interactions, and different interaction behavior types correspond to different numbers of interactions. The number of interactions is used to identify the number of times the target object interacts with the same content, and N>1. A higher number of interactions means a greater number of times the target object interacts with the content. It should be noted that in the embodiment of the present application, N is a finite value. The specific value of N can be determined by those skilled in the art based on actual conditions and application scenarios, and is not limited here.

[0077] From the above description, it can be seen that the type of interactive behavior is determined according to the number of interactions, and different numbers of interactions correspond to different types of interactive behavior. For example, assume that there are three types of interactive behaviors, namely exposure, playback, and interaction. As the name implies, exposure can be understood as displaying the content on the terminal device interface of the target object; playback requires the target object to perform a playback operation on the content displayed on the terminal device interface; interaction refers to the target object performing relevant evaluations, likes, or collections on the content displayed on the terminal device interface. From the above introduction, it can be seen that from exposure to playback to interaction, the number of interactive behaviors of the target object corresponding to these three types of interactive behaviors on the same content is increasing in sequence, that is, the depth of the interactive behaviors corresponding to these three types of interactive behaviors is gradually deepening. In other words, from exposure to playback to interaction, the number of interactions corresponding to these three types of interactive behaviors is increasing in sequence.

[0078] In the embodiment of the present application, the interaction behavior type is related to the number of interactions, and the interaction behavior type can be determined based on the number of interactions. For example, the number of interactions can be set to high, medium, and low, and the corresponding interaction behavior types are high, medium, and low. At the same time, the interaction data sequence corresponding to the interaction behavior type identifies which content the target object is willing to interact with more frequently and which content they are willing to interact with less frequently.

[0079] In a recommendation scenario, it is hoped that the recommended content can make the target object pay more or less interactions, that is, it is hoped that the target object can generate interactive behaviors for the recommended content, and the specific number can be determined according to the number of interactive behaviors. When constructing a description set for content recommendation for the target object, it is necessary to use the interaction data sequence. In the embodiment of the present application, the dimension of the number of interactions is introduced in the process of generating the interaction data sequence, and the object interaction data is divided into N interaction data sequences according to the different numbers of interactions. By adapting the generation requirements of different description sets to the divided interaction data sequences and having universality, only N interaction data sequences need to be stored in the server, which reduces the storage cost to a certain extent.

[0080] The aforementioned related technologies use different object interaction data stored in response to the generation requirements of different description sets, and the storage process is targeted storage, that is, the corresponding object interaction data needs to be stored independently for the generation requirements of a separate description set. In the embodiment of the present application, for the object interaction data that identifies the interaction behavior between the target object and the content, the object interaction data is divided into N interaction data sequences according to N interaction behavior types. A finite number of interaction behavior types are divided according to the number of interactions, and a finite number of interaction data sequences are obtained and stored in the server to meet the generation requirements of different description sets. Compared with the related technologies, it provides a universal basic data source for generating the description set of the target object, avoiding the targeted storage of complicated and redundant object interaction data.

[0081] S203: When a data acquisition request for the description set of the target object is obtained from the recommendation server, a corresponding basic data sequence is selected from the N interactive data sequences according to the data requirement parameters of the description set carried in the data acquisition request.

[0082] The above-mentioned basic data sequence is at least one of the N interaction data sequences. When the server obtains a data acquisition request for the description set of the target object from the recommendation server, the data acquisition request carries the data requirement parameters of the description set of the target object. There is an association between the specific data requirement parameters and the description set of the target object. The description set of the target object will involve information such as the time range of the target object's interactive behavior and the type of interactive behavior involved. Then the data requirement parameters corresponding to the description set of the target object will carry information about the time range and the type of interactive behavior. The corresponding basic data sequence can be selected from the aforementioned interactive data sequence through the data requirement parameters.

[0083] Specifically, continuing with the example of interactive behavior types including exposure, playback, and interaction, the interactive data sequences also include three: the exposure sequence, the playback sequence, and the interaction sequence. When the data requirement parameter in the description set of the target object carried in the data acquisition request indicates that object interaction data of an interactive behavior type such as playback of the target object is to be calculated or described, the playback sequence corresponding to the interactive behavior type of playback needs to be used as the base sequence from the previously generated interactive data sequence.

[0084] Of course, if the data requirement parameters in the description set of the target object carried by the data acquisition request indicate that when calculating or describing the object interaction data of interactive behavior types such as playback and exposure of the target object, it is necessary to use the playback sequence and exposure sequence corresponding to the two types of interactive behavior types of playback and exposure from the aforementioned generated interaction data sequence as the basic sequence.

[0085] S204: Return the basic data sequence used to generate the description set to the recommendation server.

[0086] A recommendation server is a server that provides personalized recommendations or suggestions based on the interests and preferences of a target user. It can predict the content that the target user might be interested in. Once the basic data sequence used to generate a description set for the target user is determined, it needs to be sent to the recommendation server so that the server can construct the description set based on the basic data sequence.

[0087] The data processing method provided above divides object interaction data, which identifies interactions between a target object and content, into N interaction data sequences based on N interaction behavior types. The interaction behavior type is determined based on the number of interactions, which identifies the number of interactions a target object has had with the same content. Therefore, N different interaction behavior types can be classified based on the number of interactions. The interaction data sequences, corresponding to each interaction behavior type, identify which content the target object is willing to interact with more or less frequently. When constructing a description set for content recommendation for a target object, regardless of the content recommendation requirement, the target object is expected to have a greater or lesser number of interactions with the recommended content. Therefore, when constructing the description set data requirement parameters for the recommendation server, the recommendation server can select an interaction data sequence from the N interaction data sequences that meets the requirements and return it as the base data sequence. Thus, by processing the object interaction data based on the number of interactions, the N interaction data sequences generated can adapt to the generation requirements of different description sets, providing universality. The server only needs to store N interaction data sequences, effectively conserving storage resources.

[0088] In the aforementioned S202, it is mentioned that "according to the object interaction data and N interaction behavior types, N interaction data sequences of the target object are generated", that is, in the embodiment of the present application, the number of interaction data sequences is N, that is, finite. When processing the interaction data sequence, the concept of time can be introduced, and the interaction data sequence can be divided and sub-interaction data sequences can be constructed according to different time granularities. In a possible implementation method, the interaction data sequence can be constructed as follows: according to the object interaction data and N interaction behavior types, M types of sub-interaction data sequences are constructed for each interaction behavior type through the timestamp information carried by the object interaction data. The M types of sub-interaction data sequences serve as the interaction data sequences of the corresponding interaction behavior types. The time granularity of sub-interaction data sequences of different types is different, and M>1.

[0089] Object interaction data is used to identify the interaction between the target object and the content. When an interaction occurs between the target object and the content, corresponding object interaction data is generated. The type of interaction behavior needs to be determined based on the number of interactions. Different numbers of interactions correspond to different types of interaction behavior. Object interaction data can include data of multiple types of interaction behaviors. For example, the object interaction data generated for target object A can include three types of interaction behaviors: exposure, playback, and interaction. The object interaction data generated for target object B can include two types of interaction behaviors: playback and interaction.

[0090] Object interaction data carries timestamp information. A single interaction behavior type can correspond to different sub-interaction data sequences. These sub-interaction data sequences can be constructed based on the timestamp information carried by the object interaction data, and the sub-interaction data sequences can be divided into different classes based on different time granularities. Time granularity refers to the dimension used to divide different sub-interaction data sequences. This dimension is based on the time range involved. Different sub-interaction data sequences have different time ranges, and the size of the time range corresponds to the size of the time granularity. Classifying sub-interaction data sequences by time granularity provides a standard for processing sub-interaction data sequences based on the time dimension.

[0091] The timestamp information carried by the aforementioned object interaction data can be used to determine the time information of the object interaction data. This time information can indicate the moment the object interaction data was acquired by the server and the moment the object interaction data was generated. When the server acquires the object interaction data and generates the interaction data sequence, it can combine the timestamp information and time granularity carried by the object interaction data to construct a sub-interaction data sequence for each interaction behavior type. The combination of M sub-interaction data sequences constitutes the interaction data sequence for the corresponding interaction behavior type.

[0092] A single sub-interaction data sequence in the aforementioned M sub-interaction data sequences may include multiple sub-interaction data sequences. For example, taking a sub-interaction data sequence with a time granularity of days as the unit, assuming that the object interaction data collected by the terminal device covers a time range of 10 days, then within that time range, 10 sub-interaction data sequences with a time granularity of days will be generated.

[0093] For ease of understanding, the following example illustrates this. Taking the interactive behavior type of play as an example, M sub-interaction data sequences are constructed for the object interaction data with the interactive behavior type of play, using the timestamp information carried by the object interaction data. The difference between different sub-interaction data sequences lies in the different time granularity, assuming that time granularity is divided into high time granularity and low time granularity.

[0094] Among them, high time granularity can be understood as a larger unit for dividing time, such as years, months, and days; low time granularity can be understood as a smaller unit for dividing time, such as minutes and seconds. Assume that high time granularity indicates days as units and low time granularity indicates hours as units. Then, according to the time granularity, two sub-interaction data sequences can be constructed as playback sequences (i.e., interaction data sequences with the interaction behavior type of playback), wherein one sub-interaction data sequence includes object interaction data with the interaction behavior type of playback within one hour, and the other sub-interaction data sequence includes object interaction data with the interaction behavior type of playback within one day.

[0095] The above-mentioned method for generating an interaction data sequence enables the generation of an interaction data sequence from a time granularity perspective. Based on the timestamp information carried in the object interaction data, sub-interaction data sequences of different time granularities are constructed for each interaction behavior type as the interaction data sequence corresponding to the interaction behavior type. This allows the interaction data sequence to include sub-interaction data sequences corresponding to different time granularities. Using time as the division dimension can improve the real-time performance and accuracy of the basic data sequence used to generate the description set, and can better align with the real-time interests and preferences of the target object.

[0096] The aforementioned statement "Using the timestamp information carried by the object interaction data, M sub-interaction data sequences are constructed for each interaction behavior type," which means that time granularity is introduced to construct sub-interaction data sequences. In one possible implementation, three sub-interaction data sequences can be constructed for a single interaction behavior type. These three sub-interaction data sequences include, based on time granularity from smallest to largest, the first sub-interaction data sequence, the second sub-interaction data sequence, and the third sub-interaction data sequence. The specific method for constructing sub-interaction data sequences is as follows:

[0097] A1: Construct the first-category sub-interaction data sequence in real time according to the timestamp information carried by the object interaction data, and store it in a first message queue.

[0098] From the description of A1, it can be seen that the first type of sub-interaction data sequence is constructed in real time. The object interaction data can be the historical interaction behavior data of the target object, such as browsing, playing, liking and commenting interaction behavior data. At the same time, the object interaction data carries timestamp information, which is used to record the time when the target object's interaction behavior occurs. "Real-time" can be understood as "read and build", that is, when the server obtains the object interaction data, it constructs the sub-interaction data sequence for the object interaction data. After the construction and generation of the first type of sub-interaction data sequence are completed, the first type of sub-interaction data sequence will be stored in the first message queue.

[0099] The construction of the specific first-category sub-interaction data sequence can be understood as the server aggregating the acquired object interaction data according to the interaction behavior type to form the first-category sub-interaction data sequence. In an embodiment of the present application, the construction of the first-category sub-interaction data sequence can be performed by the server using the Flink framework. Before aggregating the acquired object interaction data according to the interaction behavior type, the acquired object interaction data needs to be processed. The specific data processing includes operations such as data cleaning and data conversion. The specific steps are as follows:

[0100] 1) Perform data cleaning on the acquired object interaction data, including removing duplicate data, processing abnormal data, filling missing values, etc., to ensure the quality and accuracy of the object interaction data.

[0101] 2) Transform object interaction data for better understanding and analysis. For example, extract useful information such as the time and page of the interaction, encode the interaction type, and standardize the interaction objects. This makes the data easier to process and analyze. The interaction object refers to the object involved in the target object's interaction, such as the page viewed, video played, or content liked.

[0102] After the data processing is completed, the object interaction data is aggregated within the window according to the interaction behavior type to form a first type of sub-interaction data sequence, and stored in the first message queue.

[0103] A2: According to the time granularity of the second-type sub-interaction data sequence, read multiple first-type sub-interaction data sequences from the first message queue to generate the second-type sub-interaction data sequence.

[0104] The time granularity between the first, second, and third sub-interaction data sequences mentioned above is in ascending order. This means that the time granularity corresponding to the second sub-interaction data sequence is greater than that of the first sub-interaction data sequence. The generation process of the second sub-interaction data sequence requires the first sub-interaction data sequence as the base data. The second sub-interaction data sequence is generated by reading the constructed first sub-interaction data sequence from the first message queue. The specific number of reads is determined by the time granularity of the second sub-interaction data sequence.

[0105] Specifically, assuming that the time granularity of the first-category sub-interaction data sequence is 1 second, and the time granularity of the second-category sub-interaction data sequence is in minutes. Then, when it is necessary to generate the second-category sub-interaction data sequence, it is necessary to read the first-category sub-interaction data sequence with a time interval of 60 seconds, that is, 60 first-category sub-interaction data sequences. Then, these 60 first-category sub-interaction data sequences are merged to generate the second-category sub-interaction data sequence. In this embodiment of the present application, Spark Streaming can be used to generate the second-category sub-interaction data sequence.

[0106] A3: splicing the third type of sub-interaction data sequence based on the timestamp information of the object interaction data stored locally offline and the second type of sub-interaction data sequence.

[0107] A3 mentions "local offline stored object interaction data", which is different from the object interaction data mentioned in A1. The object interaction data mentioned here refers to historical object interaction data, while the object interaction data mentioned in A1 refers to real-time object interaction data.

[0108] The third type of sub-interaction data sequence corresponds to the largest time granularity. During the construction of the third type of sub-interaction data, two construction methods can be used based on time granularity. One construction method is to directly construct the third type of sub-interaction data sequence based on the time granularity based on the timestamp information of the object interaction data stored locally offline. The other construction method is to combine the sub-interaction data sequence constructed based on the object interaction data stored locally offline with the aforementioned second type of sub-interaction data sequence to obtain the third type of sub-interaction data sequence.

[0109] For ease of understanding, the two aforementioned construction methods are specifically described below by way of example. The first construction method is applicable to the construction process of the third-category sub-interaction data sequence with a time granularity of days. When the time granularity is in days, the object interaction data corresponding to a time interval of 1 day can be directly aggregated according to the timestamp information of the object interaction data stored locally offline to form a third-category sub-interaction data sequence. The second construction method is applicable to the construction process of the third-category sub-interaction data sequence with a time granularity of hours. When the time granularity is in hours, the second-category sub-interaction data sequence and the aforementioned third-category sub-interaction data sequence formed in days can be spliced ​​according to the timestamp information of the object interaction data stored locally offline to obtain the third-category sub-interaction data sequence. The third-category sub-interaction data sequence formed with a time granularity of days is stored in plain text, which can facilitate the troubleshooting of data problems. In the process of forming the third-category sub-interaction data sequence with a time granularity of hours, the object interaction data can be converted into pb format data.

[0110] Of course, the process of constructing the third type of sub-interaction data sequence also includes the processes of data cleaning and data conversion. Unlike the construction process of the first type of sub-interaction data sequence, the construction process of the third type of sub-interaction data sequence involves the splicing of sub-interaction data sequences with different time granularities. For example, the third type of sub-interaction data sequence with a time granularity of days and the second type of sub-interaction data sequence with a time granularity of minutes are spliced ​​together to obtain the third type of sub-interaction data sequence with a time granularity of hours.

[0111] In practical applications, when M=3, the M types of sub-interaction data sequences include, based on time granularity from small to large, the first type of sub-interaction data sequence, the second type of sub-interaction data sequence, and the third type of sub-interaction data sequence. The first type of sub-interaction data sequence can be a real-time short-term sequence, the second type of sub-interaction data sequence can be a near-line short-term sequence, and the third type of sub-interaction data sequence can be an offline long-term sequence. Figure 4 A schematic diagram of a data sequence construction in an application scenario provided in an embodiment of the present application is shown as follows: Figure 4 As shown, the interactive data sequence includes an exposure sequence, a playback sequence, and an interaction sequence. The above-mentioned interactive data sequences are constructed simultaneously. When constructing the first type of sub-interaction data sequence (i.e., a real-time short-term sequence), the Flink framework (distributed stream processing framework) is used to read the object interaction data in the second message queue in real time, and the object interaction data is processed and constructed in real time to obtain a real-time short-term sequence. The real-time short-term sequence includes a playback sequence, an exposure sequence, and an interaction sequence. The data processing process includes operations such as data cleaning, data conversion, and window aggregation. The specific steps are as follows:

[0112] 1) Before offline merging to generate the first sub-interaction data series (i.e., real-time short-term series), the original object interaction data usually needs to be cleaned. This includes removing duplicate data, processing abnormal data, filling missing values, and other operations to ensure the quality and accuracy of the object interaction data.

[0113] 2) Perform data conversion on object interaction data to facilitate better understanding and analysis. For example, extract useful information such as the time and page where the interaction occurred, encode the interaction type, and standardize the interaction objects. This makes the data easier to process and analyze.

[0114] 3) Window aggregation, that is, aggregating sub-interaction data sequences according to the interaction behavior type within the window to form a real-time short-term sequence.

[0115] When constructing the second type of sub-interaction data sequence (i.e., near-line short-term sequence), the process uses Spark Streaming (Spark streaming data processing) to consume a batch of real-time short-term sequences every 5 minutes (other time intervals are also possible and are not limited here) and merge them into near-line short-term sequences and store them offline. This provides near-line short-term sequences for generating the description set of the target object. The near-line short-term sequences include: playback sequence, exposure sequence, and interaction sequence.

[0116] When constructing the third type of sub-interaction data series (i.e., offline long-term series), the process extracts useful interaction data series (including playback series, exposure series, and interaction series) from the request history and merges them offline to form offline long-term series for persistence. This process includes data cleaning, data conversion, and merging long and short series. The offline long-term series construction process is divided into hourly and daily levels.

[0117] 1) The daily-level construction process extracts useful interaction data sequences from historical object interaction data and merges them offline with existing long-term offline sequences to form new long-term offline sequences. Spark (a distributed big data processing framework) is used for daily-level construction and outputs the corresponding long-term sequences to form offline long-term sequences. Daily-level data is stored in plain text on disk to facilitate rapid case troubleshooting (case troubleshooting refers to technical analysis and diagnostic methods used to identify and resolve problems).

[0118] 2) The hourly offline build process reads daily and 5-minute data sets, merges them (i.e., combines long and short sequences), and converts the data into petabyte format data to provide hourly offline long-term sequences for generating the target object description set. After the hourly offline build is complete, the hourly offline data set must be flushed to disk.

[0119] The above-mentioned method for constructing sub-interaction data sequences identifies three types of sub-interaction data sequences based on different time granularities, and describes the corresponding methods for constructing these sub-interaction data sequences. Different methods are used to generate different sub-interaction data sequences for different types of sub-interaction data sequences. The above-mentioned construction method enables the construction of sub-interaction data sequences with different time granularities, so that the interaction data sequences corresponding to each interaction behavior type include sub-interaction data sequences with various time granularities. Because basic data sequences must be selected from the interaction data sequences, sub-interaction data sequences with different time granularities can improve the real-time performance and accuracy of the basic data sequences used to generate the description set.

[0120] The aforementioned S203 mentions "selecting corresponding basic data sequences from the N interactive data sequences based on the data requirement parameters of the description set carried by the data acquisition request". It can be seen from the previous description that the interactive data sequence includes interactive data sequences corresponding to each interactive behavior type, and the interactive data sequence corresponding to a single interactive behavior type can be composed of sub-interaction data sequences of different time granularities. Therefore, when selecting a basic data sequence from an interactive data sequence, the selection can be made in combination with the time granularity. When the data requirement parameters of the description set include a time range requirement, the sub-interaction data sequences can also be spliced ​​in combination with the time range requirement to meet the time range requirement. In a possible implementation, the method for selecting a basic data sequence can be: first, based on the data requirement parameters of the description set carried by the data acquisition request, determine the target interactive behavior type and time range involved in the data requirement parameters. Then, when the target interactive behavior type is a target type among the N interactive behavior types, a data sequence that matches the time range is spliced ​​out from the M types of sub-interaction data sequences corresponding to the target type as the basic data sequence.

[0121] The target interaction behavior type refers to the interaction behavior type involved in the description set of the target object, and the time range refers to the time interval involved in the description set of the target object.

[0122] When a server receives a data acquisition request for a target object's description set from a recommendation server, it first determines the target interaction behavior type and time range involved in generating the target object's description set based on the data acquisition request. It then determines whether the target interaction behavior type is one of the N interaction behavior types. The target interaction behavior type can include one or more target types.

[0123] The aforementioned interaction data sequence corresponding to a single interaction behavior type includes sub-interaction data sequences of different time granularities. The time range involved in generating the description set of the target object is obtained, and the sub-interaction data sequences corresponding to the target type can be spliced ​​according to the time range so that the spliced ​​data sequence meets the time range requirements in the data acquisition request.

[0124] For example, assume that the data requirement parameter carried in the data acquisition request indicates that the target interactive behavior type involved is a playback type, and the time range involved is 10 days, 5 hours, and 20 minutes. It is also assumed that the target interaction sequence corresponding to the playback type includes 3 (i.e., M=3) types of sub-interaction data sequences, namely, a first-type sub-interaction data sequence, a second-type sub-interaction data sequence, and a third-type sub-interaction data sequence. The time granularity corresponding to the first-type sub-interaction data sequence is minutes, the time granularity corresponding to the second-type sub-interaction data sequence is hours, and the time granularity corresponding to the third-type sub-interaction data sequence is days. At this time, based on the time range of 10 days, 5 hours, and 20 minutes involved in the data requirement parameter, it is necessary to splice the first-type sub-interaction data sequence, the second-type sub-interaction data sequence, and the third-type sub-interaction data sequence so that the time range of the spliced ​​data sequence design meets 10 days, 5 hours, and 20 minutes. The splicing process requires 20 first-type sub-interaction data sequences, 5 second-type sub-interaction data sequences, and 10 third-type sub-interaction data sequences.

[0125] When the data request parameter carries a target interaction behavior type of playback or exposure, it is necessary to obtain both the target interaction sequence corresponding to the playback type and the target interaction sequence corresponding to the exposure type. Then, the sub-interaction data sequences are concatenated based on the time range specified in the data request parameter. The specific concatenation process is similar to that described above and will not be repeated here.

[0126] The aforementioned method for selecting a base data sequence allows us to select the target type based on the interaction behavior type and time range required for the target object's description set. The corresponding sub-interaction data sequences of that target type are then concatenated according to the time range requirements, generating a data sequence that meets the requirements for generating the description set as the base data sequence. Because the sub-interaction data sequences are divided based on time granularity, they are able to closely match the time range requirements of the description set. The resulting base data sequence will meet the requirements for generating the description set, resulting in a higher level of accuracy and precision for the resulting description set.

[0127] The aforementioned S201 mentions “obtaining object interaction data of the target object”. In a possible implementation, a specific method for obtaining the object interaction data may be: obtaining the object interaction data provided by the terminal device from the second message queue.

[0128] The terminal device is the terminal device that the target object performs interactive behavior on the content. That is, in this embodiment of the application, the server can obtain the object interaction data provided by the terminal device from the message queue, that is, the direct source of the object interaction data is the terminal device that performs interactive behavior on the recommended content.

[0129] By sending object interaction data to a second message queue, which is then retrieved by the server, the second message queue serves as a transit point for object interaction data. If the server is unavailable when sending object interaction data, the second message queue retains the message until the data can be successfully delivered to the server. This ensures both reliable and flexible object interaction data delivery.

[0130] It should be noted that the "first", "second" and "third" mentioned in the embodiments of the present application are only for the purpose of dividing message queues and sub-interaction data sequences, and are not used to express priority, importance and other meanings.

[0131] The method for obtaining object interaction data described above introduces a second message queue as a container for storing object interaction data during the process of generating a target object description set. This allows for the server to deliver object interaction data at any time based on its data needs, ensuring efficient generation of the target object description set. Furthermore, when the server no longer needs the object interaction data, it can be retained, ensuring the security of the object interaction data to a certain extent, preventing loss and enabling "access and use" of the data.

[0132] The above describes how to obtain object interaction data. The following describes the source of the object interaction data. The object interaction data can be collected by the target application in the terminal device and cached locally on the terminal device; or the object interaction data can be recorded in the log information of the terminal device.

[0133] When object interaction data is directly collected and cached locally on the terminal device, it is easier for the server to obtain the object interaction data than when the object interaction data is recorded in the log information of the terminal device. The reason is that although obtaining object interaction data from the log information of the terminal device is feasible, it is more complicated to implement. A series of processing is required on the log information to obtain the object interaction data. If the object interaction data is directly collected and cached locally on the terminal device through the target application in the terminal device, then when the object interaction data needs to be obtained, it can be directly obtained without the need for additional processing. The acquisition difficulty is lower and the real-time transmission of the object interaction data to the server can be guaranteed.

[0134] By using the method of determining the storage location of object interaction data provided above, when object interaction data is collected by the target application in the terminal device and cached locally on the terminal device, the difficulty of obtaining the object interaction data is greatly reduced, and the object interaction data can be obtained in real time when the server needs to obtain it, thereby ensuring the timeliness of the description set of the target object generated subsequently.

[0135] The above also mentioned a way to store object interaction data: recording it in the log information of the terminal device. Figure 5 A data acquisition diagram based on log storage is provided in an embodiment of the present application, such as Figure 5 As shown in the figure, the terminal device includes a logging system, which includes a log upload SDK, a log receiving service, a data transmission pipeline, and data distribution. The terminal device relies on the logging system to collect logs, obtaining various behavior logs of the target object, such as exposure logs, playback logs, and interaction logs. These logs record the target object's various behaviors within the product or service, including browsing, clicks, and comments. When acquiring object interaction data, log access is required, as shown in the figure with Log 1 and Log 2. Log processing is then required because logs contain other irrelevant data in addition to object interaction data, which needs to be extracted. The specific log processing process includes data cleansing and business logic processing. Data cleansing removes invalid or erroneous data, such as deduplication, handling missing values, and correcting format errors. Business logic processing filters, transforms, and performs calculations on the data based on specific business needs to extract useful information (i.e., object interaction data). The object interaction data finally extracted needs to form M types of sub-interaction data sequences, which may include long-term sub-interaction data sequences and short-term sub-interaction data sequences. The long-term and short-term can be understood as the large or small time range involved in the sub-interaction data sequence. The long-term sub-interaction data sequence is usually stored in an external KV storage system, such as Redis (key-value storage database). The short-term sub-interaction data sequence can be understood as the intermediate result obtained after processing the log. The long-term sub-interaction data sequence and the short-term sub-interaction data sequence can be merged and spliced ​​to obtain the basic data sequence. It should be noted that the external storage component used in the embodiment of the present application can be replaced.

[0136] The present application also provides a data processing method, which is executed by a server. Figure 6 A flowchart of another data processing method provided in an embodiment of the present application, the method comprising:

[0137] S601: When a recommendation service is started, a data acquisition request for a description set of a target object is sent to a data server according to the description set required by the recommendation service.

[0138] The recommendation service refers to the process of recommending content to a target user. When the recommendation service is activated, a data acquisition request is sent to the data server based on the target user's description set required by the recommendation service. This data acquisition request carries data requirement parameters for the target user's description set, including information such as the target interaction behavior type and time range covered by the description set.

[0139] S602: Acquire a basic data sequence from the data server.

[0140] The above-mentioned basic data sequence is used to generate a description set of the target object, wherein, for the target object, the basic data sequence is selected from N interaction data sequences based on data requirement parameters, and the basic data sequence is at least one of the N interaction data sequences. The N interaction data sequences are generated based on the object interaction data of the target object and N interaction behavior types. The interaction data sequences and interaction behavior types correspond one-to-one. The N interaction behavior types are determined based on the number of interactions. Different interaction behavior types correspond to different numbers of interactions. The number of interactions is used to identify the number of interaction behaviors of the target object on the same content, and N>1.

[0141] When a server sends a data acquisition request to a data server, the data server determines a basic data sequence based on the data requirement parameters carried in the data acquisition request and returns it to the server. The basic data sequence is selected from the interactive data sequence based on the data requirement parameters, and there can be one or more basic data sequences.

[0142] The process of generating interaction data sequences has been described in detail in the aforementioned embodiments. The interaction data sequences are generated by the data server based on the acquired target object's object interaction data and interaction behavior types, and there is a corresponding relationship between the interaction behavior sequences and interaction behavior types. The interaction behavior types are determined based on the number of interactions the target object has with the recommended content. That is, the interaction behavior types are divided based on the number of interactions the target object has with the recommended content, thereby generating N corresponding interaction data sequences. Based on the data requirement parameters in the data acquisition request, a basic data sequence can be determined from the N interaction data sequences, and then the generated basic data sequence can be obtained from the data server.

[0143] S603: Construct the description set according to the basic data sequence.

[0144] When the server obtains the basic data sequence sent by the data server, it constructs a description set of the target object based on the basic data sequence. The description set is used to provide content recommendation services for the target object.

[0145] Through the data processing method provided above and executed by the server, when the recommendation service is started, a data acquisition request for the description set of the target object is sent to the data server, and then the basic data sequence is obtained from the data server, and finally the description set of the target object is constructed based on the obtained basic data sequence. Through the data acquisition request, the basic data sequence that meets the data requirement parameters such as the target interaction behavior type and time range can be obtained from the data server in a targeted manner. Since the basic data sequence (that is, at least one of the N interaction data sequences) is obtained after processing the object interaction data from the dimension of the number of interactions, it can adapt to the generation requirements of different description sets and has universality. At the same time, since the basic data sequence obtained can meet the requirements of the data requirement parameters, the description set finally constructed can well fit the interest needs of the target object and improve the efficiency of the content recommendation service.

[0146] The above describes in detail the method by which the server constructs a description set of a target object. This method is applicable to the scenario of initially constructing a description set of a target object. After the construction of the description set of a target object is completed, the subsequent process requires updating the description set of the target object generated based on the acquired object interaction data to adapt to changes in the target object's content interaction interests and improve the efficiency of content recommendation. Therefore, in one possible approach, after starting the recommendation service (i.e., constructing the description set), the following operations need to be performed:

[0147] C1: Obtaining a content recommendation request for the target object from a terminal device.

[0148] The above-mentioned content recommendation request includes the object interaction data of the target object, and the object interaction data is collected by the target application in the terminal device and cached locally in the terminal device. It can be seen from the description of C1 that the terminal device will send a content recommendation request for the target object to the server, and the object interaction data of the target object is carried in the content recommendation request. In other words, when the target object corresponding to the terminal device needs to make content recommendations, the data of the interaction behavior between the target object and the content, such as the data of the interaction behaviors such as browsing, playing, liking, and commenting on each content, will be sent to the server at the same time. The object interaction data can be real-time data or historical data, which is not limited here. In addition to the object interaction data, the content recommendation request can also include the identification information of the target object and the current context environment (which can be understood as the recommendation scenario in which the target object is located when making content recommendations).

[0149] In the recommendation scenario, the terminal device collects the target object's object interaction data and stores the collected object interaction information locally on the terminal device, such as in a database on the terminal device. The database on the terminal device is a local, lightweight database used to persist the target object's object interaction data. When storing the object interaction data, key information such as the target object's identifier, the type of interaction behavior, the object of the interaction behavior (i.e., the behavior object), and timestamp information are also recorded.

[0150] C2: updating the description set of the target object according to the object interaction data, and sending the object interaction data to a data server.

[0151] In the previous section, we described how to construct a target object description set for the initial construction scenario. Once the recommendation service is activated, the previously constructed target object description set needs to be updated. Specifically, upon receiving a content recommendation request for a target object, the server parses the object interaction data and processes it based on the recommendation scenario to update the target object description set.

[0152] At the same time, the server also sends the object interaction data to the data server, which processes it and ultimately obtains the basic data sequence used to generate the target object description set and returns it to the server. In one possible implementation, data transmission between the server and the data server can be implemented based on a second message queue. When the server sends object interaction data to the data server, the data is stored in the second message queue and retrieved from the second message queue when the data server needs it.

[0153] C3: Returning corresponding content recommendation results to the terminal device according to the content recommendation request and the updated description set of the target object.

[0154] After the server completes updating the description set of the target object, it returns the corresponding content recommendation result to the terminal device based on the content recommendation request and the updated description set of the target object.

[0155] Figure 7 This is a schematic diagram of a recommendation process in an application scenario provided by an embodiment of the present application, such as Figure 7 As shown, after starting the recommendation service, the terminal device sends a content recommendation request for the target object to the server through the application backend. This content recommendation request includes the target object's object interaction data. The server updates the target object's description set based on the object interaction data and sends the object interaction data to the data server. Based on the received content recommendation request and the updated target object description set, the content recommendation results are returned to the terminal device through the application backend.

[0156] Figure 8 A schematic diagram of the structure of a server provided in an embodiment of the present application is shown in FIG. Figure 8 As shown, the server includes an access layer, an algorithm module, and a description set construction module. The algorithm module includes a recall model, a coarse ranking model, a fine ranking model, and a re-ranking model. The description set construction module is used to generate basic data sequences, generate description sets, and store these description sets in a database. The model in the algorithm module recommends content for target objects based on the description sets of the target objects constructed in the description set construction module.

[0157] In the aforementioned method for generating content recommendation results, object interaction data is stored in the local database of the terminal device. After the recommendation service is activated, the server can directly obtain the object interaction data from the terminal device. Compared with using the terminal device's log system to collect object interaction data, this can avoid complex data processing. When object interaction data needs to be transmitted, it is directly sent to the server, which can improve the real-time data transmission and, to a certain extent, ensure the timely generation of content recommendation results.

[0158] The aforementioned C1 mentions "obtaining a content recommendation request for the target object from the terminal device." After the server receives the content recommendation request, in order to avoid data loss during transmission, it can optionally return the data transmission status to the terminal device so that the terminal device can determine whether the content recommendation request has been completely transmitted to the server. Therefore, in one possible implementation, after the server obtains the content recommendation request for the target object (including object interaction data), it returns the archiving timestamp to the terminal device.

[0159] The archiving timestamp is used to identify the latest timestamp of the acquired object interaction data. It also instructs the terminal device to locally delete any object interaction data generated before the archiving timestamp. In other words, the archiving timestamp is generated after the server receives a content recommendation request and identifies the latest timestamp of the acquired object interaction data.

[0160] For example, if a terminal device sends a content recommendation request to a server that includes multiple pieces of object interaction data, each of which includes timestamp information, the latest timestamp among the object interaction data will be used as the archiving timestamp. By returning this archiving timestamp to the terminal device, the terminal device can obtain the latest object interaction data received by the server based on the archiving timestamp, making it easier to determine whether the sent object interaction data is consistent with the object interaction data received by the server.

[0161] In addition, when the terminal device obtains the archiving timestamp, it can determine the object interaction data that has been successfully sent to the server based on the archiving timestamp, and thus delete the object interaction data generated before the object interaction data corresponding to the archiving timestamp. In other words, the object interaction data generated before the archiving timestamp will be deleted.

[0162] By returning the archiving timestamp to the terminal device after the server receives a content recommendation request for the target object, the terminal device can be informed of the latest object interaction data successfully sent to the server. This allows the terminal device to determine whether any object interaction data was lost during the transmission process and promptly resend the data based on the loss, thus ensuring the accuracy of the data received by the server to a certain extent. At the same time, the archiving timestamp can be used to instruct the terminal device to locally delete object interaction data generated before the archiving timestamp, avoiding waste of the terminal device's local storage resources and reducing local storage pressure on the terminal device.

[0163] In the aforementioned description, when constructing a description set for a target object, it is necessary to do so based on a base data sequence. In addition to constructing the description set for the target object based on the base data sequence, the description set for the target object can also be constructed in conjunction with a forward-ranked index. In one possible implementation, the method for constructing the description set for the target object can be: constructing the description set based on the base data sequence and the forward-ranked index.

[0164] Get the forward index required to build the description set. The forward index uses the content identifier as the keyword and includes the descriptive features of the corresponding content. Specifically, content can refer to videos, pictures, and product links. When the content is a video, the identifier of the video can be used as the keyword. Specifically, the identifier of the video can be the video ID. At this time, the forward index corresponding to the video refers to the video ID as the keyword, and also includes the descriptive features corresponding to the video, such as: video duration, video type, and video clarity. When the video ID is obtained using the forward index, you can quickly find information such as the video duration, video type, and video clarity of the video through the video ID.

[0165] By combining a basic data sequence and a forward-ranked index to construct a description set for the target object, the playback server can determine the occurrence of interactions between the target object and the content using the basic data sequence. Combined with the forward-ranked index, the playback server can determine the descriptive features corresponding to the content that interacted with the target object. This method for constructing a description set for the target object combines both the occurrence of interactions between the target object and the content and the descriptive features corresponding to the content as data sources, resulting in a more accurate and detailed description set for the target object.

[0166] For example, it is assumed that the basic data sequence is a play sequence and the content is a video. The play sequence includes the occurrence of the interactive behavior of the target object and the video. It is assumed that the play sequence includes 10 interactive data of the interactive behavior of the target object and the video (wherein the videos are different). At this time, the description features corresponding to the corresponding 10 videos can be determined by the positive index (the description features include the video type), such as: 5 videos are life types, 2 videos are campus types, and 3 videos are suspense types. At this time, combined with the description features of the content obtained by the play sequence and the positive index, it can be roughly determined that the description set of the target object is more inclined to be interested in life types of videos, so that the determination accuracy of the description set can be improved, and then the efficiency of the server in providing content recommendation services for the target object can be improved.

[0167] As can be seen from the above description, it is necessary to obtain basic data sequences and positive indexes in the process of constructing the description set on the server. In practical applications, the basic data sequences may include but are not limited to light sequences, playback sequences, and interactive sequences. Each type of basic data sequence is composed of sub-interaction data sequences of different time granularities, such as the aforementioned real-time short-term sequences (i.e., the first type of sub-interaction data sequences), near-line short-term sequences (i.e., the second type of sub-interaction data sequences), and offline long-term sequences (i.e., the third type of sub-interaction data sequences). Among them, the real-time short-term sequence is generally sent to the second message queue (such as Kafka, stream processing platform), and the offline long-term sequence is usually stored in an offline storage system (such as HDFS (Hadoop Distributed File System, distributed file system)). Since only three basic data sequences are stored, storage space and cost can be saved without the need to store a large amount of redundant data. The following information is recorded in each basic data sequence:

[0168] Each sequence records the following key information:

[0169] (1) Target object identification: used to identify different target objects.

[0170] (2) Behavior object identifier: such as video ID, used to identify different content.

[0171] (3) Timestamp information: identifies the specific time when the interaction occurs.

[0172] (4) The scenario where the interactive behavior occurs: such as web page ID, module ID, etc., used to identify the specific scenario where the interactive behavior occurs.

[0173] (5) Playback duration: In the playback sequence, record the playback duration of the video.

[0174] (6) Interaction type: In the interaction sequence, record the type of interaction behavior (such as likes, comments, etc.).

[0175] By using the above-mentioned method for constructing a description set of a target object, a description set is constructed in combination with a basic data sequence and a positive index. While determining the occurrence of the interactive behavior between the target object and the content, the specific circumstances of the content that interacts with the target object can be determined in combination with the positive index, thereby more accurately constructing a description set of the target object and improving the quality of the content recommendation service provided to the target object.

[0176] Figure 9 This is an interactive diagram of a data processing method provided in an embodiment of the present application. As shown in the figure, three computer devices are involved in the data processing process: a terminal device, a recommendation server, and a data server. The method is applied to the scenario of content recommendation for a target object, and the method is specifically as follows:

[0177] S11: Initiate a content recommendation request.

[0178] The terminal device sends a content recommendation request to the recommendation server, wherein the content recommendation request includes object interaction data of a target object collected by a target application in the terminal device and cached locally in the terminal device.

[0179] S12: Send a data acquisition request.

[0180] After the recommendation server receives the content recommendation request for the target object from the terminal device, it will update the description set of the target object based on the object interaction data in the content recommendation request, and send a data acquisition request and object interaction data to the data server, including the data requirement parameters of the description set in the data acquisition request.

[0181] S13: Generate basic data sequence.

[0182] The data server generates a basic data sequence based on the acquired object interaction parameters. When receiving a data acquisition request from the recommendation server, the basic data sequence is determined from the interaction data sequence in combination with the data requirement parameters in the data acquisition request.

[0183] S14: Obtain basic data sequence.

[0184] The recommendation server obtains the basic data sequence used to generate the description set from the data server.

[0185] S15: Build a description set.

[0186] After the recommendation server obtains the basic data sequence, it will construct a description set of the target object based on the basic data sequence.

[0187] S16: Perform content recommendation.

[0188] The recommendation server will recommend content to the target object using the terminal device based on the constructed description set of the target object.

[0189] In the aforementioned Figure 1-9 Based on the corresponding embodiments, Figure 10 A schematic diagram of a data processing device provided in an embodiment of the present application is shown. The data processing device 1000 includes: an acquisition module 1001, a generation module 1002, a selection module 1003, and a return module 1004.

[0190] The acquisition module 1001 is used to acquire object interaction data of a target object, where the object interaction data is used to identify the interaction behavior between the target object and the content;

[0191] The generating module 1002 is configured to generate N interaction data sequences of the target object based on the object interaction data and N interaction behavior types, wherein the interaction data sequences correspond to the interaction behavior types in a one-to-one manner, and the N interaction behavior types are determined based on the number of interactions, with different interaction behavior types corresponding to different numbers of interactions. The number of interactions is used to identify the number of interactions of the target object with respect to the same content, where N>1.

[0192] The selection module 1003 is configured to, when receiving a data acquisition request for the description set of the target object from the recommendation server, select a corresponding basic data sequence from the N interaction data sequences based on the data requirement parameter of the description set carried in the data acquisition request, where the basic data sequence is at least one of the N interaction data sequences;

[0193] The returning module 1004 is configured to return the basic data sequence used to generate the description set to the recommendation server.

[0194] In a possible implementation, the generating module 1002 is specifically configured to:

[0195] According to the object interaction data and the N interaction behavior types, M types of sub-interaction data sequences are constructed for each interaction behavior type through the timestamp information carried by the object interaction data. The M types of sub-interaction data sequences serve as the interaction data sequences of the corresponding interaction behavior types. The time granularity of sub-interaction data sequences of different types is different, and M>1.

[0196] In a possible implementation, when M=3, the M types of sub-interaction data sequences include, based on time granularity from small to large, a first type of sub-interaction data sequence, a second type of sub-interaction data sequence, and a third type of sub-interaction data sequence, and the generation module 1002 is specifically configured to:

[0197] Constructing the first type of sub-interaction data sequence in real time according to the timestamp information carried by the object interaction data, and storing the sequence in the first message queue;

[0198] Reading a plurality of first-category sub-interaction data sequences from the first message queue according to the time granularity of the second-category sub-interaction data sequence to generate the second-category sub-interaction data sequence;

[0199] The third type of sub-interaction data sequence is obtained by splicing the timestamp information of the object interaction data stored locally offline and the second type of sub-interaction data sequence.

[0200] In a possible implementation, the selection module 1003 is specifically configured to:

[0201] Determining, according to the data requirement parameters of the description set carried in the data acquisition request, a target interactive behavior type and a time range involved in the data requirement parameters;

[0202] When the target interaction behavior type is a target type among the N interaction behavior types, a data sequence that matches the time range is spliced ​​out from the M types of sub-interaction data sequences corresponding to the target type as the basic data sequence.

[0203] In a possible implementation, the acquisition module 1001 is specifically configured to:

[0204] The object interaction data provided by a terminal device is acquired from a second message queue, where the terminal device is a terminal device through which the target object performs the interaction behavior on the content.

[0205] In a possible implementation, the apparatus is configured to: collect the object interaction data through a target application in the terminal device and cache the data locally in the terminal device; or

[0206] The object interaction data is recorded in the log information of the terminal device.

[0207] The data processing device provided above divides object interaction data, which identifies interactions between a target object and content, into N interaction data sequences based on N interaction behavior types. The interaction behavior type is determined based on the number of interactions, which identifies the number of interactions a target object has had with the same content. Therefore, N different interaction behavior types can be divided based on the number of interactions. The interaction data sequences, corresponding to each interaction behavior type, identify which content the target object is willing to interact with more or less frequently. When constructing a description set for content recommendation for a target object, regardless of the content recommendation requirement, the target object is expected to have a greater or lesser number of interactions with the recommended content. Therefore, when constructing data requirement parameters for the description set for the recommendation server, the interaction data sequence that meets the requirements can be selected from the N interaction data sequences and returned as the base data sequence. Thus, by processing the object interaction data based on the number of interactions, the N interaction data sequences generated can adapt to the generation requirements of different description sets, providing universality. The data server only needs to store N interaction data sequences, effectively conserving storage resources.

[0208] Figure 11 This is a schematic diagram of another data processing device provided in an embodiment of the present application. The data processing device 1100 includes: a sending module 1101, a generating module 1102, and a constructing module 1103;

[0209] The sending module 1101 is configured to send a data acquisition request for a description set of a target object to a data server according to a description set required by the recommendation service when the recommendation service is started, wherein the data acquisition request carries data requirement parameters of the description set;

[0210] The generating module 1102 is configured to obtain a basic data sequence from the data server, the basic data sequence being used to generate a description set of the target object, wherein for the target object, the basic data sequence is selected from N interaction data sequences according to the data requirement parameters, the basic data sequence being at least one of the N interaction data sequences, the N interaction data sequences being generated according to the object interaction data of the target object and N interaction behavior types, the interaction data sequences corresponding to the interaction behavior types being one-to-one, the N interaction behavior types being determined according to the number of interactions, with different interaction behavior types corresponding to different numbers of interactions, the number of interactions being used to identify the number of interactions of the target object with respect to the same content, where N>1;

[0211] The construction module 1103 is used to construct the description set according to the basic data sequence, and the description set is used to provide content recommendation service for the target object.

[0212] In a possible implementation, after starting the recommendation service, the apparatus is specifically configured to:

[0213] Obtaining a content recommendation request for the target object from a terminal device, the content recommendation request including object interaction data of the target object, the object interaction data being collected by a target application in the terminal device and cached locally on the terminal device;

[0214] updating the description set of the target object according to the object interaction data, and sending the object interaction data to a data server;

[0215] According to the content recommendation request and the updated description set of the target object, a corresponding content recommendation result is returned to the terminal device.

[0216] In a possible implementation manner, after acquiring the basic data sequence from the data server, the apparatus is specifically configured to:

[0217] An archiving timestamp is returned to the terminal device, where the archiving timestamp is used to identify the latest timestamp of the acquired object interaction data, and the archiving timestamp is used to instruct the terminal device to locally delete the object interaction data whose generation time is earlier than the archiving timestamp.

[0218] In a possible implementation, the device is specifically configured to:

[0219] Obtaining a forward index required to construct the description set, wherein the forward index uses the content identifier as a keyword and includes description features of the corresponding content;

[0220] The construction module 1103 is specifically used to:

[0221] The description set is constructed according to the basic data sequence and the forward index.

[0222] The data processing device provided above sends a data acquisition request for a description set of a target object to a data server when starting a recommendation service, then acquires a basic data sequence from the data server, and finally constructs a description set of the target object based on the acquired basic data sequence. Through the data acquisition request, a basic data sequence that meets data requirement parameters such as the target interaction behavior type and time range can be acquired from the data server in a targeted manner. Since the basic data sequence (i.e., at least one of the N interaction data sequences) is obtained after processing the object interaction data from the dimension of the number of interactions, it can adapt to the generation requirements of different description sets and has universality. At the same time, since the acquired basic data sequence can meet the requirements of the data requirement parameters, the description set finally constructed can well fit the interest needs of the target object and improve the efficiency of the content recommendation service.

[0223] The embodiment of the present application further provides a computer device, including a terminal device or a server. The computer device is described below with reference to the accompanying drawings.

[0224] If the computer device is a terminal device, see Figure 12 As shown, the embodiment of the present application provides a terminal device, taking a mobile phone as an example:

[0225] Figure 12 The block diagram shows a partial structure of the mobile phone provided by the embodiment of the present application. Figure 12 The mobile phone includes components such as a radio frequency (RF) circuit 1410, a memory 1420, an input unit 1430, a display unit 1440, a sensor 1450, an audio circuit 1460, a wireless fidelity (WiFi) module 1470, a processor 1480, and a power supply 1490. Those skilled in the art will understand that Figure 12 The mobile phone structure shown in the figure does not constitute a limitation to the mobile phone, and may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently.

[0226] The following combination Figure 12 A detailed introduction to the various components of a mobile phone:

[0227] The RF circuit 1410 may be used for receiving and sending signals during information transmission or calls. In particular, after receiving downlink information from the base station, it is sent to the processor 1480 for processing. In addition, the designed uplink data is sent to the base station.

[0228] Memory 1420 can be used to store software programs and modules. Processor 1480 executes the various functional applications and data processing of the mobile phone by running the software programs and modules stored in memory 1420. Memory 1420 may mainly include a program storage area and a data storage area. The program storage area may store an operating system and at least one application required for a function (such as a sound playback function, an image playback function, etc.); the data storage area may store data created based on the use of the mobile phone (such as audio data, a phone book, etc.). In addition, memory 1420 may include high-speed random access memory and non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0229] The input unit 1430 may be configured to receive input digital or character information and generate key signal input related to user settings and function control of the mobile phone. Specifically, the input unit 1430 may include a touch panel 1431 and other input devices 1432 .

[0230] The display unit 1440 may be configured to display information input by the user or information provided to the user, as well as various menus of the mobile phone. The display unit 1440 may include a display panel 1441 .

[0231] The mobile phone may also include at least one sensor 1450, such as a light sensor, a motion sensor, and other sensors.

[0232] The audio circuit 1460 , the speaker 1461 , and the microphone 1462 can provide an audio interface between the user and the mobile phone.

[0233] WiFi is a short-range wireless transmission technology. The mobile phone can help users send and receive emails, browse web pages, and access streaming media through the WiFi module 1470, providing users with wireless broadband Internet access.

[0234] The processor 1480 is the control center of the mobile phone. It uses various interfaces and lines to connect various parts of the entire mobile phone. It executes various functions of the mobile phone and processes data by running or executing software programs and / or modules stored in the memory 1420 and calling data stored in the memory 1420.

[0235] The mobile phone also includes a power supply 1490 (such as a battery) for supplying power to various components.

[0236] In this embodiment, the processor 1480 included in the terminal device is also used to execute the steps in the methods of each embodiment of the present application.

[0237] If the computer device is a server, this embodiment of the application also provides a server, see Figure 13 As shown, Figure 13 The structural diagram of the server 1500 provided in the embodiment of the present application, the server 1500 may have relatively large differences due to different configurations or performances, and may include one or more central processing units (CPUs) 1522 (for example, one or more processors) and a memory 1532, and one or more storage media 1530 (for example, one or more mass storage devices) for storing application programs 1542 or data 1544. Among them, the memory 1532 and the storage medium 1530 can be temporary storage or permanent storage. The program stored in the storage medium 1530 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations on the server. Furthermore, the central processing unit 1522 can be configured to communicate with the storage medium 1530 to execute a series of instruction operations in the storage medium 1530 on the server 1500.

[0238] The server 1500 may also include one or more power supplies 1526, one or more wired or wireless network interfaces 1550, one or more input and output interfaces 1558, and / or one or more operating systems 1541, such as Windows Server 2003. TM , Mac OS X TM , Unix TM ,Linux TM , FreeBSD TM etc.

[0239] The steps performed by the server in the above embodiment can be based on Figure 13 The server structure shown.

[0240] In addition, an embodiment of the present application further provides a storage medium, which is used to store a computer program, and the computer program is used to execute the method provided by the above embodiment.

[0241] An embodiment of the present application further provides a computer program product including a computer program, which, when executed on a computer device, enables the computer device to execute the method provided in the above embodiment.

[0242] Those skilled in the art will understand that all or part of the steps of implementing the above-mentioned method embodiment can be completed by hardware related to program instructions, and the above-mentioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above-mentioned method embodiment; and the above-mentioned storage medium can be at least one of the following media: read-only memory (English: Read-only Memory, abbreviated: ROM), RAM, magnetic disk or optical disk, etc., various media that can store computer programs.

[0243] In the embodiments of the present application, the term "module" or "unit" refers to a computer program or a part of a computer program that has a predetermined function and works together with other related parts to achieve a predetermined goal, and can be implemented in whole or in part by using software, hardware (such as processing circuits or memories) or a combination thereof. Similarly, a processor (or multiple processors or memories) can be used to implement one or more modules or units. In addition, each module or unit can be part of an overall module or unit that includes the function of the module or unit.

[0244] It should be noted that the various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device and system embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiments. The device and system embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the scheme of this embodiment. A person of ordinary skill in the art can understand and implement it without expending creative work.

[0245] The above is only one specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the scope of protection of the present application. Moreover, based on the implementation methods provided in the above aspects, the present application can also be further combined to provide more implementation methods. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A data processing method, characterized in that: The method comprises: Acquire object interaction data of a target object, where the object interaction data is used to identify an interaction behavior between the target object and content; Generate N interaction data sequences for the target object based on the object interaction data and N interaction behavior types, wherein the interaction data sequences correspond to the interaction behavior types in a one-to-one manner, and the N interaction behavior types are determined based on the number of interactions, with different interaction behavior types corresponding to different numbers of interactions. The number of interactions is used to identify the number of interactions of the target object with respect to the same content, where N>1. When a data acquisition request for the description set of the target object is obtained from the recommendation server, a corresponding basic data sequence is selected from the N interaction data sequences according to the data requirement parameter of the description set carried in the data acquisition request, where the basic data sequence is at least one of the N interaction data sequences; The basic data sequence used to generate the description set is returned to the recommendation server.

2. The method according to claim 1, characterized in that Generating N interaction data sequences of the target object according to the object interaction data and N interaction behavior types includes: According to the object interaction data and the N interaction behavior types, M types of sub-interaction data sequences are constructed for each interaction behavior type through the timestamp information carried by the object interaction data. The M types of sub-interaction data sequences serve as the interaction data sequences of the corresponding interaction behavior types. The time granularity of sub-interaction data sequences of different types is different, and M>1.

3. The method according to claim 2, characterized in that When M=3, the M types of sub-interaction data sequences include, based on time granularity from small to large, a first type of sub-interaction data sequence, a second type of sub-interaction data sequence, and a third type of sub-interaction data sequence. The timestamp information carried by the object interaction data is used to construct M types of sub-interaction data sequences for each interaction behavior type, including: Constructing the first type of sub-interaction data sequence in real time according to the timestamp information carried by the object interaction data, and storing the sequence in the first message queue; Reading a plurality of first-category sub-interaction data sequences from the first message queue according to the time granularity of the second-category sub-interaction data sequence to generate the second-category sub-interaction data sequence; The third type of sub-interaction data sequence is obtained by splicing the timestamp information of the object interaction data stored locally offline and the second type of sub-interaction data sequence.

4. The method according to claim 2, characterized in that The selecting a corresponding basic data sequence from the N interactive data sequences according to the data requirement parameter of the description set carried in the data acquisition request includes: Determining, according to the data requirement parameters of the description set carried in the data acquisition request, a target interactive behavior type and a time range involved in the data requirement parameters; When the target interaction behavior type is a target type among the N interaction behavior types, a data sequence that matches the time range is spliced ​​out from the M types of sub-interaction data sequences corresponding to the target type as the basic data sequence.

5. The method according to claim 1, wherein The acquiring the object interaction data of the target object includes: The object interaction data provided by a terminal device is acquired from a second message queue, where the terminal device is a terminal device through which the target object performs the interaction behavior on the content.

6. The method according to claim 5, characterized in that The object interaction data is collected by the target application in the terminal device and cached locally in the terminal device; or, The object interaction data is recorded in the log information of the terminal device.

7. A data processing method, characterized in that: The method comprises: When starting the recommendation service, according to the description set required by the recommendation service, a data acquisition request for the description set of the target object is sent to the data server, wherein the data acquisition request carries the data requirement parameters of the description set; Obtaining a basic data sequence from the data server, the basic data sequence being used to generate a description set of the target object, wherein, for the target object, the basic data sequence is selected from N interaction data sequences according to the data requirement parameters, the basic data sequence being at least one of the N interaction data sequences, the N interaction data sequences being generated according to object interaction data of the target object and N interaction behavior types, the interaction data sequences corresponding to the interaction behavior types being one-to-one, the N interaction behavior types being determined according to a number of interactions, with different interaction behavior types corresponding to different numbers of interactions, the number of interactions being used to identify the number of interaction behaviors of the target object with respect to the same content, where N>1; The description set is constructed according to the basic data sequence, and the description set is used to provide content recommendation service for the target object.

8. The method according to claim 7, characterized in that After starting the recommendation service, the method further includes: Obtaining a content recommendation request for the target object from a terminal device, the content recommendation request including object interaction data of the target object, the object interaction data being collected by a target application in the terminal device and cached locally on the terminal device; updating the description set of the target object according to the object interaction data, and sending the object interaction data to a data server; According to the content recommendation request and the updated description set of the target object, a corresponding content recommendation result is returned to the terminal device.

9. The method according to claim 8, characterized in that After acquiring the basic data sequence from the data server, the method further includes: An archiving timestamp is returned to the terminal device, where the archiving timestamp is used to identify the latest timestamp of the acquired object interaction data, and the archiving timestamp is used to instruct the terminal device to locally delete the object interaction data whose generation time is earlier than the archiving timestamp.

10. The method according to claim 7, characterized in that The method further comprises: Obtaining a forward index required to construct the description set, wherein the forward index uses the content identifier as a keyword and includes description features of the corresponding content; The constructing the description set according to the basic data sequence includes: The description set is constructed according to the basic data sequence and the forward index.

11. A data processing device, characterized in that: The device includes: an acquisition module, a generation module, a selection module and a return module; The acquisition module is used to acquire object interaction data of the target object, where the object interaction data is used to identify the interaction behavior between the target object and the content; The generating module is configured to generate N interaction data sequences of the target object based on the object interaction data and N interaction behavior types, wherein the interaction data sequences correspond to the interaction behavior types in a one-to-one manner, and the N interaction behavior types are determined based on the number of interactions, with different interaction behavior types corresponding to different numbers of interactions. The number of interactions is used to identify the number of interactions of the target object with respect to the same content, where N>1; The selection module is configured to, when receiving a data acquisition request for the description set of the target object from the recommendation server, select a corresponding basic data sequence from the N interaction data sequences based on the data requirement parameter of the description set carried in the data acquisition request, where the basic data sequence is at least one of the N interaction data sequences; The returning module is configured to return the basic data sequence used to generate the description set to the recommendation server.

12. A data processing device, characterized in that: The device comprises: a sending module, a generating module and a building module; The sending module is configured to send a data acquisition request for the description set of the target object to the data server according to the description set required by the recommendation service when the recommendation service is started, wherein the data acquisition request carries the data requirement parameters of the description set; The generation module is configured to obtain a basic data sequence from the data server, the basic data sequence being used to generate a description set of the target object, wherein, for the target object, the basic data sequence is selected from N interaction data sequences based on the data requirement parameters, the basic data sequence being at least one of the N interaction data sequences, the N interaction data sequences being generated based on the object interaction data of the target object and N interaction behavior types, the interaction data sequences corresponding to the interaction behavior types being one-to-one, the N interaction behavior types being determined based on the number of interactions, with different interaction behavior types corresponding to different numbers of interactions, the number of interactions being used to identify the number of interactions of the target object with respect to the same content, where N>1; The construction module is used to construct the description set according to the basic data sequence, and the description set is used to provide content recommendation service for the target object.

13. A computer device, characterized in that: The computer device includes a processor and a memory: The memory is used to store computer programs; The processor is configured to execute the method according to any one of claims 1 to 10 according to the computer program.

14. A computer-readable storage medium, characterized in that The computer-readable storage medium is used to store a computer program, and when the computer program is executed by a computer device, the computer program implements the method according to any one of claims 1 to 10.

15. A computer program product comprising a computer program, which, when run on a computer device, causes the computer device to perform the method according to any one of claims 1 to 10.

Images