Information recommendation method, apparatus, device, storage medium, and program product

Abstract

The application relates to an information recommendation method and device, equipment, a storage medium and a program product, and relates to the technical field of information recommendation. The method comprises the following steps: obtaining a first recall information set based on this-time recall information recalled in a this-time recall process; in historical recall information recalled in a historical recall process, filtering historical recall information with a historical score result higher than a first condition to obtain a second recall information set; the historical score result is used to indicate the probability of a target object account performing a target operation on the historical recall information at a historical recall moment; and based on the first recall information set and the second recall information set, providing this-time recommended recall information to the target object account. Through the above method, the historical recall result with a score meeting the condition in the previous recommendation process is reused in the this-time recall stage, the association between different information recommendation processes is realized, the utilization effect of the recall information is improved, and the information recommendation effect is improved.

Description

Technical Field

[0001] This application relates to the field of information recommendation technology, and in particular to an information recommendation method, apparatus, device, storage medium, and program product. Background Technology

[0002] The development of multimedia and big data technologies has enabled information recommendation to better meet user needs. The information recommendation process can be implemented through a recommendation system. A single recommendation process within a recommendation system includes a recall phase, a ranking phase, and a recommendation phase.

[0003] In related technologies, the goal of the recall stage is usually to sift out a small amount of information from a massive amount of information as recall information; the goal of the ranking stage is usually to rank the recall information; and the goal of the recommendation stage is usually to recommend information based on the ranking results.

[0004] The proportion of information recommended during the recommendation phase is relatively small compared to the information recalled during the recall phase. Furthermore, due to factors such as the timeliness of information, there are differences in the recalled information across different recommendation processes. This results in a lower probability that recalled information that was not recommended during the recommendation phase will be recommended later, thus affecting the effectiveness of information recommendation. Summary of the Invention

[0005] This application provides an information recommendation method, apparatus, device, storage medium, and program product, which can realize the association between different information recommendation processes, thereby improving the utilization effect of recalled information and thus improving the information recommendation effect. The technical solution is as follows:

[0006] On the one hand, an information recommendation method is provided, the method comprising:

[0007] Based on the recall information retrieved during this recall process, a first recall information set is obtained;

[0008] In the historical recall process, the historical recall information with historical score results higher than the first condition is filtered to obtain a second recall information set; the historical score results are used to indicate the probability that the target account will perform the target operation on the historical recall information at the time of historical recall.

[0009] Based on the first recall information set and the second recall information set, the recall information recommended in this instance is provided to the target account.

[0010] On the other hand, an information recommendation device is provided, the device comprising:

[0011] The first set acquisition module is used to obtain the first recall information set based on the recall information recalled in this recall process.

[0012] The second set acquisition module is used to filter the historical recall information in the historical recall process that has a historical score result higher than the first condition to obtain a second recall information set; the historical score result is used to indicate the probability that the target account will perform the target operation on the historical recall information at the time of historical recall.

[0013] The information recommendation module is used to provide the recommended recall information to the target account based on the first recall information set and the second recall information set.

[0014] In one possible implementation, the second set acquisition module is used to filter the historical recall information recalled in the historical recall process that has a historical rating result higher than the first condition and has not been recommended to the target account, thereby obtaining the second recall information set.

[0015] In one possible implementation, the second set acquisition module includes:

[0016] A subset reading module is used to read historical information subsets corresponding to n record time periods in a historical information set; the historical recall time of the historical recall information in the historical information subset is within the historical time period corresponding to the historical information subset; the historical score result of the historical recall information in the historical information subset is higher than the first condition and has not been recommended to the target account; n is a positive integer;

[0017] The set acquisition submodule is used to acquire the second recall information set from the n subsets of historical information.

[0018] In one possible implementation, the collection acquisition submodule includes:

[0019] A subset acquisition unit is used to acquire, when generating the historical information subset for each of the recorded time periods in the n historical information subsets, the historical recall information subset corresponding to each of the i recommendation processes executed within the recorded time period; the historical recall information subset contains target historical recall information; the historical rating result of the target historical recall information is among the top k in the historical recall information subset; i and k are positive integers;

[0020] The first set acquisition unit is used to acquire a set of unrecommended information, which includes unrecommended information from the target historical recall information corresponding to each of the i recommendation processes.

[0021] The subset generation unit is used to generate the historical information subset corresponding to the recorded time period based on the set of unrecommended information.

[0022] In one possible implementation, the subset generation unit is configured to, in response to the number of unrecommended information in the set of unrecommended information being greater than a first recall threshold, sort the unrecommended information based on the historical rating results of each of the unrecommended information to obtain a first sorting result.

[0023] Based on the first sorting result, the number of unrecommended information equal to the first recall threshold is obtained, and a subset of historical information corresponding to the recorded time period is formed.

[0024] In one possible implementation, the subset acquisition submodule includes:

[0025] A time acquisition unit is used to acquire the recommendation time; the recommendation time is the time corresponding to this recommendation process.

[0026] The subset acquisition unit is used to acquire m target historical information subsets from n historical information subsets based on the recommendation time; m≤n, and m is a positive integer;

[0027] The second set acquisition unit is used to acquire the set of historical recall information composed of m target historical subsets as the second recall information set.

[0028] In one possible implementation, the subset acquisition unit is configured to:

[0029] The time interval is determined based on the time interval within the recorded time period in which the recommendation time occurs during the current recommendation process;

[0030] The target time period is the recording time period that is before the recording time period in which the current recommendation process takes place and differs from the recording time period in which the current recommendation process takes place by the time interval specified.

[0031] The historical information subset corresponding to the target time period is obtained into m target historical information subsets.

[0032] In one possible implementation, the second set acquisition module is used to filter the historical recommendation information from the historical recommendation information recommended in the historical recall process that did not receive the target operation performed by the target object, and obtain the second recall information set; the target object is the object corresponding to the target object account.

[0033] In one possible implementation, the information recommendation module includes:

[0034] The scoring acquisition submodule is used to acquire the current scoring result of each recall information in the recall information set; the recall information set includes the first recall information set and the second recall information set;

[0035] The recommendation information determination submodule is used to determine recommendation information based on the current scoring results of each recall information in the recall information set;

[0036] The information recommendation submodule is used to provide the recommendation information to the target account.

[0037] In one possible implementation, the recommendation information determination submodule includes:

[0038] The sorting unit is used to sort the recall information in the recall information set based on the current scoring result of each recall information in the recall information set, and obtain a second sorting result;

[0039] The recommendation information determination unit is used to determine the recommendation information based on the second ranking result.

[0040] In one possible implementation, the scoring acquisition submodule includes:

[0041] The object feature acquisition unit is used to acquire the object features of the target object account.

[0042] The scoring acquisition unit is used to acquire the current scoring result of each recall information based on the object characteristics of the target object account.

[0043] In one possible implementation, the scoring acquisition unit is used to input the object characteristics of the target object account and the target recall information into the scoring model to obtain the current scoring result of the target recall information output by the scoring model; the target recall information is any one of various recall information;

[0044] The scoring model is obtained by training a training sample set, which includes sample object features, sample information, and scoring labels of the sample object account relative to the sample object account.

[0045] In one possible implementation, the sample information in the training sample set is historical recommendation information; the sample object feature is the object feature of the recommendation object account corresponding to the historical recommendation information when the historical recommendation information was recommended; the rating tag of the sample information relative to the sample object account is the recommendation result of the historical recommendation information relative to the recommendation object account.

[0046] The recommendation result includes one of the following: the recommended object performs the target operation, or the recommended object does not perform the target operation; the recommended object is the object corresponding to the recommended object's account.

[0047] In one possible implementation, the first recall information set includes at least one of a first sub-information set, a second sub-information set, and a third sub-information set; the first sub-information set is an information set composed of recall information obtained from the candidate recall information based on keywords determined from the historical behavior data of the target account; the second sub-information set is an information set composed of recall information obtained from the candidate recall information based on semantic vectors determined from the historical behavior data of the target account; and the third sub-information set is an information set composed of recall information obtained from the candidate recall information based on information attributes.

[0048] On the other hand, a computer device is provided, the computer device including a processor and a memory, the memory storing at least one computer program, the at least one computer program being loaded and executed by the processor to implement the above-described information recommendation method.

[0049] On the other hand, a computer-readable storage medium is provided, wherein at least one computer program is stored in the computer program, which is loaded and executed by a processor to implement the above-described information recommendation method.

[0050] On the other hand, a computer program product or computer program is provided, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the information recommendation method provided in the various alternative implementations described above.

[0051] The technical solution provided in this application may include the following beneficial effects:

[0052] By utilizing the recall information from this recall process, a first recall information set is obtained; based on the relationship with the first condition, a second recall information set is obtained; and recommended information for this recommendation process is determined from the information in the first and second recall information sets, and then recommended information is made. Through this scheme, in this recall phase, historical recall results that met the scoring conditions in previous recommendation processes are reused, realizing the correlation between different information recommendation processes, thereby improving the utilization effect of recall information and ultimately improving the information recommendation effect.

[0053] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0054] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0055] Figure 1 A schematic diagram illustrating an exemplary embodiment of the present application shows an information recommendation process;

[0056] Figure 2 This is a schematic diagram illustrating the structure of an information recommendation system according to an exemplary embodiment;

[0057] Figure 3 This is a flowchart illustrating an information recommendation method according to an exemplary embodiment;

[0058] Figure 4 This illustration shows a schematic diagram of the process for generating a subset of historical information according to an exemplary embodiment of this application;

[0059] Figure 5 This illustration shows a schematic diagram of the process of obtaining target historical recall information, as illustrated in an exemplary embodiment of this application.

[0060] Figure 6 This illustration shows a schematic diagram illustrating the acquisition of a set of unrecommended information according to an exemplary embodiment of this application;

[0061] Figure 7 A schematic diagram illustrating the historical information set generation process provided in an exemplary embodiment of this application is shown;

[0062] Figure 8 A flowchart illustrating an exemplary embodiment of this application is shown;

[0063] Figure 9 This illustration shows a schematic diagram illustrating the correspondence between time intervals and sources of target historical information subsets, as shown in an exemplary embodiment of this application.

[0064] Figure 10 This application shows a schematic diagram of the architecture of an information recommendation system illustrated in an exemplary embodiment.

[0065] Figure 11 A block diagram illustrating an information recommendation apparatus according to an exemplary embodiment of this application is shown;

[0066] Figure 12 This is a structural block diagram of a computer device according to an exemplary embodiment;

[0067] Figure 13 This is a structural block diagram of a computer device according to an exemplary embodiment. Detailed Implementation

[0068] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0069] The information recommendation method provided in this application embodiment can establish the correlation between independent recommendation processes and realize the reuse of historical recall information in historical recommendation processes. Figure 1 A schematic diagram illustrating an exemplary embodiment of this application shows an information recommendation process. For example... Figure 1 As shown, in the recall phase of the recommendation process, the computer device can recall information through two different information sources. In this embodiment, the recall path with the information source being candidate recall information is described as the first recall path 110, and the recall path with the information source being historical recall information is described as the second recall path 120. The first recall path 110 is used to filter candidate recall information that matches the recall tags from a massive amount of candidate recall information, based on the matching results of each candidate recall information with the recall tags, to form a first recall information set 130. The second recall path 120 is used to obtain historical recall information with high historical ratings but not recommended to the target object account from historical recall information retrieved and filtered from previous recommendation processes, or to obtain historical recommendation information that was recommended in a historical recommendation process but did not receive the target object's target operation, forming a second recall information set 140, where the target object is the object corresponding to the target object account. Subsequently, the computer device uses the information in the first recall information set 130 and the information in the second recall information set 140 as the recall information in this recommendation process. During the ranking stage, the recall information is ranked, and the top few recall information items are used as recommendation information to provide the recommendation information to the target account during the recommendation stage.

[0070] Figure 2 This is a schematic diagram of the structure of an information recommendation system 200 according to an exemplary embodiment. The information recommendation system 200 includes a server 220 and a plurality of terminals 240.

[0071] Server 220 includes at least one of a single server, multiple servers, a cloud computing platform, and a virtualization center. Server 220 provides backend services to terminal 240; in this embodiment, server 220 can be used to execute an information recommendation process, sending the acquired recommendation information to terminal 240 so that terminal 240 can recommend information through methods such as interface display or audio playback.

[0072] Terminal 240 can be a terminal device with information recommendation function. For example, terminal 240 can be a mobile phone, tablet computer, e-book reader, smart glasses, smartwatch, MP3 player (Moving Picture Experts Group Audio Layer III), MP4 player (Moving Picture Experts Group Audio Layer IV), laptop computer, desktop computer, etc. Terminal 240 may include an application with information push function for information push. Optionally, the above application can be an application that requires downloading and installation, or it can be an application that can be used instantly; this embodiment of the application does not limit this.

[0073] Terminal 240 and server 220 are connected via a communication network. Optionally, the communication network can be a wired network or a wireless network.

[0074] Optionally, the aforementioned wireless or wired networks use standard communication technologies and / or protocols. The network is typically the Internet, but can also be any network, including but not limited to Local Area Networks (LANs), Metropolitan Area Networks (MANs), Wide Area Networks (WANs), mobile, wired or wireless networks, private networks, or any combination of virtual private networks. In some embodiments, technologies and / or formats, including Hyper Text Markup Language (HTML), Extensible Markup Language (XML), etc., are used to represent data exchanged over the network. Furthermore, conventional encryption technologies such as Secure Socket Layer (SSL), Transport Layer Security (TLS), Virtual Private Networks (VPNs), and Internet Protocol Security (IPsec) can be used to encrypt all or some links. In other embodiments, custom and / or dedicated data communication technologies can be used to replace or supplement the aforementioned data communication technologies.

[0075] Figure 3This is a flowchart illustrating an information recommendation method according to an exemplary embodiment. The information recommendation method can be executed by a computer device, which can be implemented as a terminal or a server. Schematic, the computer device can be implemented as follows: Figure 1 The terminal 240 or server 220 shown. For example... Figure 3 As shown, this information recommendation method may include the following steps:

[0076] Step 310: Based on the recall information recalled in this recall process, obtain the first recall information set.

[0077] In this embodiment of the application, when recalling the recall information in this recall process, the computer device can filter the recall information from each candidate recall information based on the current matching result of each candidate recall information to obtain a first recall information set; the current matching result is used to indicate the matching degree between the candidate recall information and the recall label; the recall label is used to indicate the rules for filtering the recall information for the target account from each candidate recall information.

[0078] The content type of the recall information varies depending on the application scenario. For example, in a shopping recommendation scenario, the recall information is product information; in a video recommendation scenario, it is video information; in an audio recommendation scenario, it is audio information; in an image push scenario, it is image information, and so on. Therefore, the content type of the recall information can be determined based on the recommendation needs of different application scenarios, and this application does not impose any restrictions on it.

[0079] In this embodiment, the number of candidate recall information can be much larger than the number of recall information obtained based on the current matching result. That is, the process of obtaining the first recall information set can be considered as initially filtering a portion of the candidate recall information that matches the recall tag from a massive amount of candidate recall information as the current recall information, thereby reducing the amount of information to be processed in the sorting stage and improving information processing efficiency. The recall tag can be based on the target object or filtering conditions, and is illustrative. The recall tag indicates the rules for filtering the current recall information for the target object account from each candidate recall information. The target object account is the recommended object account to be recommended during the current recommendation process. The computer device can use different rules to filter the candidate recall information. For example, it can filter the candidate recall information based on the target object account, or it can filter the candidate recall information based on information attributes. Information attributes can include the vertical category (or vertical domain) to which the information belongs, the information's hot ranking, etc. Therefore, the recall tag can be a tag corresponding to object features, or it can be a tag corresponding to information attributes, etc. This application does not impose any limitations on this.

[0080] When obtaining the first set of recall information during the recall phase, the higher the matching degree between the candidate recall information and the recall tag, the higher the probability of it being selected as the recall information for this recall.

[0081] Step 320: In the historical recall information recalled during the historical recall process, filter the historical recall information whose historical score results are higher than the first condition to obtain the second recall information set; the historical score results are used to indicate the probability that the target account will perform the target operation on the historical recall information at the time of historical recall.

[0082] The first condition can be used to indicate a threshold for historical scoring results; or, the first condition can also indicate a threshold for the ranking after sorting based on historical scoring results.

[0083] The historical recall time is the time when the historical recall information was recalled in the past.

[0084] In the information recommendation process, to filter and obtain recommended information from the recalled information, computer devices need to score the recalled information during the sorting stage. This determines the probability that the target account will perform a target operation on the recalled information after it is recommended to the target account. The top-scoring entries are then selected as recommended information and presented to the target account during the recommendation stage. Therefore, for historical recalled information, each historical recalled information has a corresponding historical score. This historical score indicates the probability that the historical recalled information received a target operation from the target account at the time of the historical recall. Because in the recommendation process, recalled information... The number of historical recall messages exceeds the number of recommended messages. Therefore, there may be historical recall messages with high historical ratings that were not pushed to the target account; or there may be historical recommendation messages that were recommended to the target account but did not receive the target action from the target account. For example, if there are 10 historical recall messages, and the two historical recall messages with the highest historical ratings are determined to be pushed after being sorted by historical ratings, then the historical recall messages ranked after the second highest will not be pushed during the historical recommendation process; or, of the two historical recall messages that are determined to be pushed, one of them did not receive the target action performed by the target account. Because the candidate recall information is updated frequently, historical recall information that was not pushed during the previous recommendation process and has a high score, or historical recommendation information that was pushed but did not receive the target object's target operation, may not be recalled during the next information recall, and the probability of it being recommended to the target object is low. Although such historical recall information or historical recommendation information may also receive a high score during the ranking stage. Based on the above factors, if only the first recall path is used for recall, the following problems will occur: due to the update of the recall model or the impact of information timeliness, historical recall information or historical recommendation information that could have received a high score in the ranking stage cannot enter the ranking stage because it was not recalled in the recall stage of this recommendation process.

[0085] In this embodiment of the application, in order to effectively utilize historical recall information, the computer device can use the information set (second recall information set) composed of historical recall information obtained by filtering historical scoring results based on historical recall information as another path for obtaining recall information in the current recall process, so as to establish the correlation between the historical recommendation process and the current recommendation process, thereby realizing the reuse of historical recall information and improving the efficiency and effect of information recommendation.

[0086] In this embodiment of the application, the target operation performed by the target object corresponding to the target object account on the recall information can refer to the positive feedback operation of the target object on the recall information. For example, the positive feedback operation may include at least one of the operations such as liking, collecting, clicking, sharing, and sending gifts. The rating result can be used to indicate the probability that the target object will perform the above-mentioned target operation. For example, when the target operation performed by the target object on the recall information is a click operation, the rating result can be used to indicate the probability that the recall information will be clicked by the target object corresponding to the target object account.

[0087] In this embodiment, historical recall information can be recorded according to a recording time period. Since the probability of the same recommended object performing the target operation on the same historical recall information may differ in different recording time periods, historical scoring results have a time attribute, and the historical scoring results corresponding to the same historical recall information may be different at different historical recall times.

[0088] Since the target audience's interests change over time, to improve the relevance between the historical recall information used in this recall and the current recommendation process, in this embodiment, the computer device can filter and obtain a second set of recall information from the historical recall information retrieved in the historical recall process based on the historical recall time. Illustratively, the computer device can filter and obtain a second set of recall information from the historical recall information based on the recorded time period where the historical recall time of the historical recall information is located. This avoids invalid recalls caused by a large time difference between the historical recall time of the historical recall information obtained when randomly filtering historical recall information from the historical recall process and the current recommendation process, thus improving the recall effect of historical recall information.

[0089] Step 330: Based on the first recall information set and the second recall information set, provide the recall information for this recommendation to the target account.

[0090] In one possible implementation, the computer device can obtain recommendation information from the recall information set and recommend the recommendation information; the recall information set includes the current recall information in the first recall information set and the historical recall information in the second recall information set.

[0091] When a computer device is implemented as a server, the server can send a set of recall information to the terminal, so that the terminal can determine the recommended information from the set of recall information and make recommendations based on the recommended information.

[0092] Alternatively, the server can determine the recommended information from the recall information set and send the recommended information to the terminal so that the terminal can make recommendations based on the received information.

[0093] When a computer device is implemented as a terminal, the terminal determines the recommended information from the recall information set and recommends the recommended information.

[0094] Optionally, the computer device can recommend information through a targeted recommendation method. For example, when the computer device is implemented as a server, the server can instruct the terminal to recommend information in a targeted recommendation method. Optionally, this targeted recommendation method can include at least one of an interface display method or an audio playback method; for example, the terminal can recommend information through a notification bar; or, the terminal can recommend information through an interface displayed in an installed application; or, the terminal can recommend information through a pop-up window; or, the terminal can recommend information through audio playback, etc. This application does not limit the method by which the computer device recommends information.

[0095] In summary, the information recommendation method provided in this application obtains a first recall information set based on the recall information recalled during the current recall process; obtains a second recall information set based on the relationship between historical scoring results of historical recall information and a first condition; determines the recommended information for the current recommendation process from the information in the first and second recall information sets, and recommends the recommended information. Through this scheme, in the current recall stage, historical recall results that meet the scoring conditions in previous recommendation processes are reused, realizing the correlation between different information recommendation processes, thereby improving the utilization effect of recall information and ultimately improving the information recommendation effect.

[0096] Optionally, the second recall set includes historical recommendation information from the historical recommendation information recommended during the historical recommendation process where the target operation was not performed by the target object; or, the second recall set includes at least one of the historical recall information from the historical recall process where the historical score result is higher than the first condition and was not recommended to the target object's account. In other words, during this recommendation process, the computer device can filter historical recall information from the historical recall information retrieved during the historical recall process where the historical score result is higher than the first condition and was not recommended to the target object's account to obtain the second recall information set; and / or, filter historical recommendation information from the historical recommendation information recommended during the historical recommendation process where the target operation was not performed by the target object to obtain the second recall information set.

[0097] When the second recall information set includes historical recommendation information that was recommended in the historical recommendation process but whose target operation was not executed by the target object, the computer device can pre-determine the historical recommendation information determined in each historical recommendation process by performing statistics on the historical recommendation information, and store the historical recommendation information and the historical scoring results of the historical recommendation information to generate a historical recommendation information set containing the recommended target operation that was not executed by the target object; during the current recommendation process, the computer device can obtain the second recall information set by filtering the information in the historical recommendation information set.

[0098] When the second recall information set includes historical recall information recalled during the historical recall process, where the historical score is higher than the first condition and it was not recommended to the target account, since there is a large number of historical recall information during the historical recall process, in order to obtain the historical recall information with high historical scores but not recommended, the computer device can filter the historical recall information to generate a historical information set containing historical recall information with high historical scores but not recommended; during this recommendation process, the computer device can obtain the second recall information set by filtering the information in the historical information set.

[0099] This application uses historical recall information included in the second recall information set, which includes historical recall information recalled during the historical recall process, where the historical score result is higher than the first condition and it was not recommended to the target account, as an example to illustrate the information recommendation method provided in this application.

[0100] Optionally, when retrieving historical recall information from the historical information set, the computer device performs information retrieval and storage on a periodic basis, with each recording time period as a cycle. This historical information set contains n subsets of historical information corresponding to different recording time periods; the historical recall time of each piece of historical recall information contained in each subset falls within the recording time period corresponding to that subset, where n is a positive integer. As time changes, the computer device generates new subsets of historical information according to a certain time cycle and adds these subsets to the historical information set to update the historical information set.

[0101] The following is about the above. Figure 3 The process of generating historical information subsets for each recorded time period in the historical information set involved in the illustrated embodiment is explained; Figure 4 The illustration shows a schematic diagram of the process for generating a subset of historical information according to an exemplary embodiment of this application, such as... Figure 4 As shown, the process includes the following steps:

[0102] Step 410: When generating historical information subsets for each recording time period in the n historical information subsets, obtain the historical recall information subsets corresponding to the i recommendation processes executed within the recording time period; the historical recall information subset contains target historical recall information; the historical rating result of the target historical recall information is among the top k in the historical recall information subset; i and k are positive integers.

[0103] In other words, within a recorded time period, the computer device can execute at least one recommendation process, each of which includes a recall phase, a sorting phase, and a recommendation phase. In this embodiment, the number of recommendation processes executed within each recorded time period can be set by relevant personnel. For example, relevant personnel can control the frequency of information recommendations to the target account by setting the number of times a recommendation process is executed within a recorded time period, or by setting the recommendation time of each recommendation process within a recorded time period; alternatively, the number of recommendation processes that can be executed within each recorded time period can be determined based on the number of information recommendation operations triggered by the target object. The number of recommendation processes executed within each recorded time period varies depending on the number of information recommendation operations triggered by the target object. For example, the information recommendation operation may include at least one of the target object's operation to launch an application and the target object's operation to refresh the recommended content.

[0104] Taking the first recording time period out of n recording time periods as an example, if there are other recording time periods before the first recording time period, the historical recall information set obtained in the recall phase of any recommendation process within the first recording time period includes the historical first recall information set and the historical second recall information set; if there are no other recording time periods before the first recording time period, the historical recall information set obtained in the recall phase of any recommendation process within the first recording time period includes the historical first recall information set.

[0105] Taking an example where there are other recorded time periods before the first recorded time period, and the first recall information set contains historical recall information obtained through explicit recall branches, implicit recall branches, and functional recall branches, respectively. Figure 5 This illustration shows a schematic diagram of the process for obtaining target historical recall information, as shown in an exemplary embodiment of this application. Figure 5As shown, the recalled items (information) obtained through the first recall path (including explicit recall branches, implicit recall branches, and functional recall branches) and the second recall path (historical recall path) will all enter the sorting stage to form a recall information set 510 (this recall information set 510 can be used as the aforementioned historical recall information set); the computer device sorts each piece of recall information in the historical recall information set 510 during the sorting stage; wherein, each recall path may correspond to its own recall quantity limit, such as Figure 5 As shown, the recall quantity limit for each recall branch in the first recall path is 4, and the recall quantity limit for the second recall path (historical recall path) is also 4. The number of recall information items in the recall information set is 16. After sorting each recall information item in the recall information set 510 according to the scoring results during the sorting stage, the sorting results can be obtained. Based on the push information quantity settings, the computer device can indicate the ranking of the top N items in the scoring results (…). Figure 5 The recall information (N=2) is pushed as recommendation information to the target account, and this information is also obtained as historical recommendation information. Furthermore, considering that recall information with higher ratings needs to be pushed during information push, to reduce storage space and processing resource consumption, when obtaining the target historical recall information, the ranking result can indicate the top M rating results (N=2). Figure 5 The recall information obtained in M=4) is the target historical recall information 520 obtained during a recommendation process executed in the first recording time period, where M>N and M and N are both positive integers.

[0106] Step 420: Obtain the set of unrecommended information, which contains the unrecommended information in the target historical recall information corresponding to each of the i recommendation processes.

[0107] The computer device will compile a set of unrecommended information from the subsets of historical recall information corresponding to each recommendation process executed within the first recorded time period. This set of unrecommended information refers to historical recall information that was not recommended to the target account during the corresponding recommendation process. Specifically, during the same recommendation process, if there is overlap in historical recall information obtained from different recall branches, only one of the overlapping historical recall information will be retained. Similarly, for the unrecommended information in the target historical recall information corresponding to each recommendation process, if there is overlap in the unrecommended information, only one of the overlapping unrecommended information will be retained. Figure 6 This illustration shows a schematic diagram of obtaining a set of unrecommended information, as shown in an exemplary embodiment of this application. Figure 6As shown, taking the execution of 12 recommendation processes within the first recording time period as an example, the target historical recall information corresponding to each of the 12 recommendation processes is recorded. If there is overlap in the target historical recall information of the 12 recommendation processes, only one of them is retained. That is to say, in the above 12 recommendation processes, the target historical recall information is item1 to item16; among them, the historical recall information that has been recommended to the target account is: item1, item2, item5 to 8, item11 to 12, item15 to 16; the historical recall information that has not been recommended to the target account is: item3, item4, item9, item10, item13, item14; that is to say, the non-recommended information set 610 corresponding to the first recording time period includes the non-recommended information as: item3, item4, item9, item10, item13, item14.

[0108] Step 430: Based on the set of unrecommended information, generate a subset of historical information corresponding to the recorded time period.

[0109] In one possible implementation, the set of unrecommended information corresponding to the recorded time period can be obtained as a subset of historical information corresponding to that recorded time period.

[0110] Alternatively, in another possible implementation, to avoid wasting computer storage and subsequent data processing resources due to an excessive number of historical recall information retrieved through the historical recall path, the computer device sets a first recall threshold when generating historical information subsets corresponding to each recording time period. This threshold limits the maximum number of historical recall information that can be included in each historical information subset. In this case, in response to the number of unrecommended information in the unrecommended information set exceeding the first recall threshold, the computer device sorts the unrecommended information based on its historical rating results, obtains a first sorting result, and, based on the first sorting result, retrieves unrecommended information equal to the first recall threshold to form the historical information subset corresponding to the recording time period.

[0111] In other words, when storing historical recall information for a target, the computer device also stores the historical rating results of that historical recall information. This allows it to select the unrecommended information with higher historical ratings based on the historical rating results of each unrecommended information item when retrieving a subset of historical information, forming a subset of historical information corresponding to the recording time period. Taking the first recording time period as an example... Figure 6 As shown, if the first recall threshold is 4, the top 4 unrecommended items with higher historical ratings can be obtained: item3, item4, item10, and item14, forming a subset 620 of historical information corresponding to the first record time period.

[0112] It should be noted that the sorting method based on the historical rating results of each unrecommended information can be either ascending or descending. When the sorting method is ascending, the number of unrecommended information items counted from the end of the first sorted result to the first recall threshold is obtained to form a subset of historical information. When the sorting method is descending, the number of unrecommended information items counted from the beginning of the first sorted result to the first recall threshold is obtained to form a subset of historical information.

[0113] If the same unrecommended information is recalled more than once in at least one recommendation process within the first recording time period, then the unrecommended information will obtain multiple rating results after entering the ranking stage multiple times. In this case, the computer device can obtain the maximum value or average value of the multiple rating results of the unrecommended information within the first recording time period as the historical rating result of the unrecommended information within the first recording time period.

[0114] Based on the above Figure 4 The illustrated embodiment explains the generation process of each subset of historical information in the historical information set. Figure 7 The illustration shows a schematic diagram of the historical information set generation process provided in an exemplary embodiment of this application, such as... Figure 7 As shown, for n record time periods in the historical information set, i recommendation processes are executed in each record time period. Taking the generation process of the historical subset of record time period 1 in the n record time periods as an example: In the i recommendation processes of record time period 1, each recommendation process corresponds to a historical recall information subset 710; the computer device obtains the top M historical recall information in each historical recall information subset 710 and makes it the target historical recall information 720; then, it obtains non-recommended information 730 from the target historical recall information 720 corresponding to each recommendation process, and integrates the non-recommended information in the i recommendation processes to obtain a set of non-recommended information 740; after obtaining the non-recommended information, the computer device can filter the information in the set of non-recommended information 740 based on the first recall threshold set to limit the number of historical recall information in the historical information subset, and obtain the set of filtered non-recommended information as the historical information subset of record time period 1.

[0115] The process of generating historical information subsets for other recording time periods can refer to the process of generating historical information subsets for recording time period 1, and will not be repeated here.

[0116] After obtaining the historical information subsets corresponding to each of the n recorded time periods, the computer device obtains the set of the n historical information subsets as the historical information set 750.

[0117] The historical information subsets in the historical information set are continuously updated over time. Historical information subsets with longer time intervals than the current one have a lower probability of being recommended. Therefore, to reduce the consumption of computer storage resources, in one possible implementation, the computer device can set a historical information subset threshold to limit the maximum number of historical information subsets that the historical information set can store. Optionally, historical information subsets are stored in the historical information set in order of their recording time periods from oldest to newest. In response to a situation where the number of historical information subsets stored in the historical information set equals the historical information subset threshold, and a new historical information subset with a new recording time period is available for addition, the computer device removes the historical information subset with the oldest recording time period and adds the historical information subset with the new recording time period to the historical information set.

[0118] It should be noted that the process of obtaining the historical information set described above can be a preprocessing process, which can be performed between two information recommendation processes; or, it can be performed during the execution of an information recommendation process preceding the current recommendation process, thus eliminating the need to obtain the historical information set during the current recommendation process and improving the recall efficiency of the retrieved information. Alternatively, the process of obtaining the historical information set described above can also be performed during the current information recommendation process, and this application does not impose any restrictions on this.

[0119] After obtaining the historical information set, Figure 8 This application illustrates a flowchart of an exemplary embodiment of an information recommendation method. This method can be executed by a computer device, which can be implemented as a terminal or a server. Schematic, the computer device can be implemented as follows: Figure 1 The terminal 240 or server 220 shown. For example... Figure 8 As shown, this information recommendation method may include the following steps:

[0120] During the recall phase of this recommendation process:

[0121] Step 810: Based on the recall information recalled in this recall process, obtain the first recall information set.

[0122] In one possible implementation, the first recall information set includes at least one of a first sub-information set, a second sub-information set, and a third sub-information set; the first sub-information set is an information set composed of recall information obtained from candidate recall information based on keywords determined from the historical behavior data of the target account; the second sub-information set is an information set composed of recall information obtained from candidate recall information based on semantic vectors determined from the historical behavior data of the target account; and the third sub-information set is an information set composed of recall information obtained from candidate recall information based on information attributes.

[0123] The method of recalling information based on keywords determined by the historical behavior data of the target account can be called explicit recall.

[0124] Optionally, the computer device can determine target characteristics based on the historical behavioral data of the target account; generate keywords based on the target characteristics, which can be used to explicitly describe the target account's characteristics; illustratively, the target characteristics may include the target's region, gender, age, interests, the model of the terminal used, purchasing power, and other aspects. In explicit recall, the recall tag is the keyword determined based on the historical behavioral data of the target account.

[0125] Optionally, the object feature can be determined based on the target object's selection of object tags. These object tags can include object categories and object interests. For example, when the target object first opens the application, the terminal can provide selectable object tags. The object feature of the target object's account can be determined based on the target object's selection of these tags. For instance, when the target object first opens a video application, multiple object tags can be provided. If the target object selects tags such as "comics" or "anime," the tag content of the selected object tag can be obtained as the object feature, and the keyword determined based on this object feature could be "anime / manga." Alternatively, the object feature can also be determined based on the target object's historical activity records. For example, in a video recommendation application, based on the target object's historical browsing, clicking, and searching records, it can be determined that the target object frequently browses, clicks, and searches for financial videos. Therefore, the object feature can be identified as a preference for financial content, and the keyword determined based on this object feature could be "finance." The above two methods of determining object features to identify target object account keywords can be used separately or in combination; this application does not impose any restrictions on this.

[0126] Because the historical behavioral data of the target account changes over time or with changes in the target's interests, the recall information obtained through explicit recall will also change accordingly. To ensure the timeliness of the recall information, optionally, the historical behavioral data can be the target's behavioral data within a specified time threshold from the current time point; illustratively, the historical behavioral data can be the target's behavioral data within one week or three days from the current time point; the specified time threshold can be set by relevant personnel, and this application does not impose any restrictions on it.

[0127] In one possible implementation, the process of recalling candidate recall information through explicit recall can be implemented using a machine learning model. This application provides an explicit recall model for determining the matching degree between candidate recall information and the target account based on the keywords of the target account and the candidate recall information, and then determining whether to recall the candidate recall information based on the matching degree. This process can be implemented as follows: inputting the keywords of the target account and the candidate recall information into the explicit recall model to obtain the matching degree output by the explicit recall model; wherein, the explicit recall model can be trained based on a first sample set, which may include the keywords of the target account, the candidate recall information, and the matching degree labels corresponding to the candidate recall information; the candidate recall information includes positive samples and negative samples, where positive samples refer to candidate recall information that matches the keywords of the target account, and negative samples refer to candidate recall information that does not match the keywords of the target account.

[0128] The method of information recall based on semantic vectors determined from the historical behavior data of the target account can be called implicit recall; in the process of implicit recall, computer devices cannot explicitly describe the target's points of interest.

[0129] Implicit recall can be achieved through two methods: User-to-item (U2I) and Item-to-item (I2I). U2I is a method in recommendation systems that recalls users by calculating the similarity between them and items; I2I is a method in recommendation systems that recalls items by calculating the similarity between them. In implicit recall, the recall label is a semantic vector determined based on the historical behavior data of the target account. For U2I, the recall label is the semantic vector of the target account; for I2I, the recall label is the semantic vector of the matching information. This matching information can include historical recall information, historical information about received target actions, etc.

[0130] Implicit recall typically utilizes machine learning algorithms for information transformation and matching. Taking U2I as an example, a computer device uses machine learning algorithms to convert target objects and items into semantic vectors, then calculates the vector similarity between their respective semantic vectors, and uses this vector similarity for information recall. Illustratively, suppose the semantic vector of target object 1 after machine learning is (1, 1, 1), the semantic vector of item 1 is (1, 0.9, 1), and the semantic vector of item 2 is (1, 2, 2). Then, through vector similarity calculation, it can be determined that the vector similarity between the semantic vector of target object 1 and the semantic vector of item 1 is higher than the vector similarity between the semantic vector of target object 1 and the semantic vector of item 2. Therefore, item 1 is recalled between item 1 and item 2. The vector similarity can be calculated using one of the following formulas: cosine similarity, dot product similarity, Euclidean similarity, etc., and this application does not impose any restrictions on this.

[0131] Optionally, embodiments of this application provide an implicit recall model to perform vector transformation on the conversion object and determine the vector similarity between different conversion objects; wherein, for U2I, the conversion objects are the target object account and the candidate recall information; for I2I, the conversion objects are the matching information and the candidate recall information. This implicit recall model can be trained based on a second sample set, which may include the sample object account, the sample candidate recall information, and the vector similarity labels corresponding to the sample candidate recall information.

[0132] Because the candidate recall information in the recommendation system is updated frequently, and the target object account and the object characteristics of the same target object account will change over time, in order to ensure the accuracy of the recall model (including explicit recall model and implicit recall model) in recalling information, the recall model can be updated according to the target training cycle.

[0133] Indicatively, the candidate recall information in each of the above sample sets (including the first sample set and the second sample set) can be historical candidate recall information, and the sample object accounts in each of the above sample sets can be historical recommendation object accounts corresponding to historical candidate recall information. The matching degree label in the first sample set or the vector similarity label in the second sample set can be used to indicate whether the historical candidate recall information was obtained as historical recall information under the historical recommendation object account. For example, taking a target training frequency of 1 day as an example, the update process of the explicit recall model can be as follows: based on today's candidate recall information, the keywords of the recommendation object account, and the recall information, the first sample set is determined. The explicit recall model is trained and updated using today's first sample set, so that the matching degree of the candidate recall information and the keywords of the recommendation object account obtained tomorrow can be predicted by the updated explicit recall model, thereby determining the recall information obtained tomorrow through explicit recall.

[0134] Information recall based on information attributes can be called functional recall. Functional recall can include recall methods such as hotspot recall and vertical recall. Hotspot recall refers to recalling information based on the cumulative results (information attributes) of target operations received by each candidate recall information within a certain period of time, in order to recall the current hot information. For example, hotspot information can be obtained from candidate information based on the CTR (Click-Through-Rate) and recalled. Vertical recall refers to recalling information based on the information type (information attributes) of each candidate recall information, in order to recall information within a specific domain. Functional recall ensures that target users have a chance to receive the latest information content, or a chance to receive information content within a specific domain; in this case, the recall tag is the information attribute.

[0135] In one possible implementation, to reduce the information processing pressure on the computer device during the sorting stage, the computer device can limit the number of current recall information items that can be included in the first recall information set. That is, the first recall information set contains a first number of current recall information items. In response to the first recall information set containing a first sub-information set, a second recall information set, and a third sub-information set, the number of current recall information items included in each sub-information set can be set respectively to limit the number of current recall information items included in the first recall information set. The number of current recall information items in each sub-information set can be the same or different, but the sum of the number of current recall information items in the sub-information sets is less than or equal to the first number.

[0136] Step 820: Read the historical information subsets corresponding to each of the n recorded time periods in the historical information set; the historical recall time of the historical recall information in the historical information subset is within the historical time period corresponding to the historical information subset; the historical score result of the historical recall information in the historical information subset is higher than the first condition, and it has not been recommended to the target account.

[0137] The process by which computer devices acquire historical information sets can be referenced. Figure 4 Corresponding embodiments or Figure 7 The relevant details of the corresponding embodiments will not be repeated here.

[0138] Step 830: Obtain the second recall information set from n subsets of historical information.

[0139] Optionally, the process of obtaining the second recall information set from the n historical information subsets can be implemented as steps S8301 to S8303:

[0140] S8301, retrieve the recommendation time corresponding to this recommendation process.

[0141] The recommendation moment can refer to the moment when the target object triggers the recommendation process, or it can be the moment when relevant personnel pre-set the time to proactively recommend information to the target object.

[0142] S8302, based on the recommendation time, obtain m target historical information subsets from n historical information subsets; m≤n, and m is a positive integer.

[0143] In one possible implementation, within the recording time period corresponding to the recommendation time of this recommendation process, the target historical information subsets for each recommendation process are the same or the same set of historical information subsets. For example, the m target historical information subsets for each recommendation process are all historical information subsets of the previous recording time period (m=1); or, the m target historical information subsets for each recommendation process are a set of historical information subsets composed of historical information subsets of the m recording times before the current recording time period (m>1). Illustratively, the recording time period is in days. Therefore, the target historical information subset for a recommendation process performed at any time today is the historical information subset obtained yesterday (m=1); or, it is a set of historical information subsets composed of the historical information subsets of each of the 3 days prior to today (m=3).

[0144] In another possible implementation, to fully utilize the various subsets of historical information in the historical information set, a correspondence can be established between the time interval of the recommendation time and the recording time period of the historical information subset. The process of obtaining the target historical information subset described above can be implemented as follows:

[0145] Based on the time interval within the recorded time period in which the recommendation time occurs during this recommendation process, the time interval is obtained;

[0146] The target time period is the recording time period that is before the current recommendation process and differs from the current recommendation process by the specified time interval.

[0147] The historical information subset corresponding to the target time period is obtained into m target historical information subsets.

[0148] Taking the recommendation time being set by relevant personnel as an example, these personnel can set the recommendation time to every hour on the hour. This means the computer device will push information at the top of the hour every day. Each hour on the hour can have a corresponding (m=1) subset of target historical information. For example, if the recording time period is set in days, the correspondence between the recommended time period and the recording time period can be set as follows: when the recommended time is between 0:00 and 2:00, the time interval between the target time period and the current recording time period is 8 days; when the recommended time is between 3:00 and 5:00, the time interval is 7 days; when the recommended time is between 6:00 and 8:00, the time interval is 6 days, and so on, until the recommended time is between 21:00 and 23:00, where the time interval between the target time period and the current recording time period is 1 day.

[0149] When the recommendation time is the point in time when the information recommendation operation is triggered by the target object, the source of the target's historical information subset can be determined based on the time interval in which the recommendation time falls. Figure 9 This illustration shows a schematic diagram illustrating the correspondence between time intervals and the sources of target historical information subsets, as shown in an exemplary embodiment of this application. Figure 9 As shown, in the process of obtaining a subset of target historical information, taking one (m=1) record time period corresponding to each time interval as an example, when the recommended time is within the time interval of 0 to 3 o'clock, the time interval between the target time period and the current record time period is 8 days; when the recommended time is within the time interval of 3 to 6 o'clock, the time interval between the target time period and the current record time period is 7 days; when the recommended time is within the time interval of 6 to 9 o'clock, the time interval between the target time period and the current record time period is 6 days, and so on.

[0150] Alternatively, in one possible implementation, a correspondence can be set between time intervals and multiple (m>1) recording time periods; illustratively, when the recommended time is within the time interval of 0 to 3 o'clock, the time interval between the recording time period in the target time period and the current recording time period is 16 days and 15 days, respectively; when the recommended time is within the time interval of 3 to 6 o'clock, the time interval between the recording time period in the target time period and the current recording time period is 14 days and 13 days, respectively, and so on.

[0151] It should be noted that the correspondence between the time interval in which the above-mentioned recommended time is located and the target time interval can be set by relevant personnel. In other words, the setting of the recommended time, the setting of the time interval, the source of the target historical subset corresponding to different time intervals, and the number of target historical subsets corresponding to different time intervals can all be changed based on actual needs, and this application does not impose any restrictions on this.

[0152] S8303: Obtain the set of historical recall information from m target historical subsets as the second recall information set.

[0153] Optionally, when m=1, the target history subset is obtained as the second recall information set.

[0154] When m>1, the set of historical recall information ranked in the top x of the historical scores in the m target historical subsets can be obtained as the second recall information set; where the value of x is equal to the maximum number of historical recall information that the second recall information set can contain, and x is a positive integer.

[0155] Optionally, since the recall model is updated according to the target training period, and the target history subset is obtained according to the recording time period; in response to the fact that the target training period of the recall model is the same as the period of the recording time period, the set of historical recall information composed of m target history subsets can be obtained as the second recall information set.

[0156] Alternatively, in response to the difference between the target training period of the recall model and the period of the recording time, after obtaining m target history subsets, these m target history subsets can be further filtered according to the target training period of the recall model to obtain a second recall information set from the target history subset corresponding to the recording time period before the most recent update of the recall model. This process can be as follows:

[0157] The set of historical recall information from the m target historical subsets that occurred before the most recent recall model update is obtained as the second recall information set. For example, assuming the recording period is 1 day and the target training period of the recall model is 2 days, after obtaining the target historical subsets corresponding to each of the 3 recording periods, if the 3rd recording period (yesterday) and the current recording period (today) belong to the same target training period (meaning the recall model used yesterday and today is the same), then the recall results of yesterday and today may be relatively similar. Therefore, the target historical subset of the 3rd recording period can be excluded, and the set of historical recall information from the remaining 2 target historical subsets is obtained as the second recall information set.

[0158] During the scoring phase of this recommendation process:

[0159] Step 840: Obtain the current scoring results of each recall information in the recall information set; the recall information set includes the first recall information set and the second recall information set.

[0160] In this embodiment of the application, the recall information in the first recall information set and the historical recall information in the second recall information set are jointly obtained as the recall result of the recall stage of the current recommendation process, forming a recall information set; by obtaining the current scoring result of each recall information in the recall information set, the recommendation information in the current recommendation process is determined.

[0161] It should be noted that the rating results of each recall message are related to the recommended target account. The same recall message may have different rating results for different recommended target accounts; that is, different target accounts may respond differently to the same recall message. Therefore, when rating each recall message in the recall message set, the rating can be combined with the recommended target account to determine the recommendation information for the current recommended target account (target target account). Thus, when obtaining the current rating result of each recall message, the target target account's characteristics can be obtained first; based on the target target account's characteristics, the current rating result of each recall message can be obtained separately.

[0162] To improve the accuracy of the scoring results for each recall information, in this embodiment of the application, the computer device can obtain the scoring results for each recall information through machine learning methods. Illustratively, the process of obtaining the current scoring results for each recall information can be implemented as follows: inputting the object characteristics of the target account and the target recall information into the scoring model, and obtaining the current scoring result of the target recall information output by the scoring model; the target recall information is any one of the various recall information;

[0163] The scoring model is trained on a training sample set, which includes the sample object features, sample information, and the scoring labels of the sample object account relative to the sample object account.

[0164] The scoring model can be at least one of LR (Logistic Regression), FM (Factor Machine), and DNN (Deep Neural Network).

[0165] During model training, the sample object features and sample information of the sample object account are input into the scoring model to obtain the predicted score result of the sample information relative to the sample object account. The scoring model is trained based on the predicted score result and the score label corresponding to the sample information. Based on different combinations of sample information and sample object features of different sample object accounts, the above process is repeated to iteratively train the scoring model until the training completion condition is met, and a trained scoring model is obtained. The trained scoring model is then used to predict the score result of the target recall information relative to the target object account.

[0166] In the above process, the value of the loss function can be calculated based on the predicted rating result of the sample information relative to the sample object account, and the rating label of the sample information relative to the sample object account, so as to train the rating model based on the value of the loss function.

[0167] The training completion conditions may include the scoring accuracy of the trained scoring model reaching an accuracy threshold, the trained scoring model converging, or the number of iterations reaching a number threshold.

[0168] In one possible implementation, to improve the training effect of the rating model, the historical information recommendation results can be obtained as the training sample set; that is, the sample information in the training sample set is the historical recommendation information; the sample object feature is the object feature of the recommendation object account corresponding to the historical recommendation information when the historical recommendation information was recommended; the rating label of the sample information relative to the sample object account is the recommendation result of the historical recommendation information relative to the recommendation object account.

[0169] The recommendation result includes either the recommended object performing the target operation or the recommended object not performing the target operation; for example, if the recommendation result is that the recommended object performed the target operation, the rating label corresponding to this sample information can be 100, and if the recommendation result is that the recommended object did not perform the target operation, the rating label corresponding to this sample information is 0; where the recommended object is the object corresponding to the recommended object account.

[0170] Step 850: Determine recommended information based on the current scoring results of each recall message in the recall information set.

[0171] In one possible implementation, when determining recommended information based on the current rating results of each recalled information in the recall information set, the recalled information in the recall information set can be sorted based on the current rating results of each recalled information in the recall information set to obtain a second sorting result; and the recommended information can be determined based on the second sorting result.

[0172] Specifically, a second ranking result can be obtained by sorting the recall information in the recall information set in ascending or descending order. Since the second ranking result is obtained by sorting each recall information in the recall information set in descending order based on the current scoring result, the top N recall information indicated by the second ranking result can be obtained as the recommended information, where N is a positive integer. At the same time, the top M recall information indicated by the second ranking result is obtained as the target historical recall information in the current recommendation process. The target historical recall information and the scoring results of the target historical recall information in the current recommendation process are stored. Combined with the target historical recall information obtained by other recommendation processes in the current recording time period, a subset of historical information for the current recording time period is generated for the historical recall path of subsequent recommendation times to obtain the second recall information set.

[0173] During the recommendation phase of this recommendation process:

[0174] Step 860: Provide recommendation information to the target account.

[0175] The computer device retrieves the top N recall information items from the second ranking result as recommended information and makes recommendations. Illustratively, when the computer device is a server, it sends the determined recommended information to the terminal, enabling the terminal to recommend the information through its own screen display or sound playback functions.

[0176] In summary, the information recommendation method provided in this application obtains a first recall information set based on the recall information recalled during the current recall process; obtains a second recall information set based on the relationship between historical scoring results of historical recall information and a first condition; determines the recommended information for the current recommendation process from the information in the first and second recall information sets, and recommends the recommended information. Through this scheme, in the current recall stage, historical recall results that meet the scoring conditions in previous recommendation processes are reused, realizing the correlation between different information recommendation processes, thereby improving the utilization effect of recall information and ultimately improving the information recommendation effect.

[0177] When acquiring historical recall information, obtaining a second information set from the historical information set based on the historical recall time of the information can make the historical recall time of the acquired historical recall information have a certain temporal correlation with the current recommendation process, thereby improving the acquisition effect of historical recall information.

[0178] When making information recommendations, a correspondence is set between different time intervals and the target time period for obtaining historical information subsets, so that each historical information subset can be used at the corresponding time, thereby improving the information utilization rate of historical recall information.

[0179] In one possible implementation, when obtaining recommendation information, recommendation information is retrieved from both a first recall information set and a second recall information set. In this case, during the ranking phase, when ranking the recall information in the recall information sets, the current rating result of the current recall information in the first recall information set can be obtained; based on the current rating result of the current recall information, the current recall information is ranked to obtain a third ranking result; based on the third ranking result, the current recall information that meets the first recommendation threshold is considered recommended information; during the ranking phase, based on the historical rating results of historical recall information in the second recall information set, historical recall information is ranked to obtain a fourth ranking result; based on the fourth ranking result, historical recall information that meets the second recommendation threshold is considered recommended information. The values ​​of the first recommendation threshold and the second recommendation threshold can be the same or different, and the first and second recommendation thresholds can be set by relevant personnel according to actual needs. To illustrate, assuming that both the first recommendation threshold and the second recommendation threshold are 2, after obtaining the first recall information set and the second recall information set, during the sorting stage, the current recall information of the top two candidates with higher current ratings in the first recall information set, and the historical recall information of the top two candidates with higher historical ratings in the second recall information set are combined to form the recommendation information for this recommendation process, and the recommendation information is provided to the target account.

[0180] Figure 10 This illustration shows a schematic diagram of the architecture of an information recommendation system according to an exemplary embodiment of this application, used to implement the information recommendation method provided in this application, such as... Figure 10As shown, the recommendation process of this information recommendation system can be divided into a recall stage 1010, a ranking stage 1020, and a recommendation stage 1030. To implement the information recommendation method provided in this application and to reuse historical recall information, this embodiment adds multiple components to the information recommendation system for statistical analysis and filtering of historical recall information. These components include a new recall path (historical recall path) as a second recall path 1040, which, together with the explicit recall branch, the implicit recall branch, and the functional recall branch, forms the first recall path 1050, jointly realizing information recall in the recall stage. These multiple components include a historical recall result recording component 1041, a historical recall result statistics component 1042, and a historical recall component 1043.

[0181] Among them, the historical recall result recording component 1041 is used to obtain and store the top recommendation results reported by the online recommendation system, combined with... Figure 3 In the illustrated embodiment, the historical recall result recording component is used to store the top N recall information (recommendation information) and the top M recall information; where M>N. The historical recall result statistics component 1042 is used to perform statistical summarization according to the period of the recording time period, obtain the historical recall subset for each recording time period, and store it; that is, the historical recall result statistics component 1042 is used to count the non-recommended information in each recommendation process executed within each recording time period, and generate the historical recall subset corresponding to each recording time period based on the set of non-recommended information in each segment within each recording time period. The historical recall component 1043 is used to reuse historical recall information by pulling the stored historical recall subset back to the sorting stage of a new round of recommendation process as a new recall path.

[0182] The top recommendation results reported by the online recommendation system obtained by the historical recall result recording component 1041, and the historical recall subsets for each recording time period obtained by the historical recall result statistics component 1042, can be stored in an offline storage system; or, to ensure information security, the above information can also be stored in a blockchain.

[0183] In one possible implementation, to ensure the security and accuracy of the object information and recommendation information of the recommended object during the information recommendation process, the historical behavior data, object feature data, and historical recall information data of the recommended object involved in the information recommendation method disclosed in this application can all be stored on the blockchain.

[0184] It is understood that in the specific implementation of this application, data related to object characteristics and historical behavior data of recommended objects are involved. When the above implementation of this application is applied to specific products or technologies, user permission or consent is required, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of relevant countries and regions.

[0185] This application improves the effectiveness and efficiency of information recommendation by adding historical recall paths to the recommendation system, while using less storage resources, such as 200G of Redis storage and a small amount of Spark computing resources.

[0186] The information recommendation method provided in this application can be applied to any scenario that requires information recommendation. The following provides two possible information recommendation scenarios:

[0187] 1. Scenarios where applications installed on a terminal proactively push information to target objects;

[0188] In some cases, applications installed on a terminal need to proactively push information to target users. For example, even when the application is not running on the terminal, it may push recommended information to the target user at a predetermined time via pop-ups, notifications, etc., so that the target user can access the information content based on the recommendations. In this process, at the predetermined time, the application sends a push notification command to its server via the terminal. Upon receiving the push notification command, the server uses the push notification method provided in this application to retrieve information from both candidate and historical sources, sorts the retrieved information, obtains the push information, and returns it to the application on the terminal. After obtaining the recommended information, the application pushes it to the target user via a notification bar or pop-up on the terminal.

[0189] 2. Use cases for recommendation features such as "You May Also Like" provided in the application;

[0190] Applications typically offer information recommendation features. For example, in video applications, upon receiving an action from a target user to open the application or refresh the application's recommendation page, the application can recommend information to the target user based on their historical behavior data or real-time trending data. During this process, the application can send an information push command to its server when the target user opens the application or refreshes the application's page. This command retrieves recommendation information from the server based on the information push method provided in this application and displays that recommendation information on the application's interface.

[0191] Figure 11 A block diagram of an information recommendation apparatus illustrated in an exemplary embodiment of this application is shown, such as Figure 11 As shown, the information recommendation device includes:

[0192] The first set acquisition module 1110 is used to obtain the first recall information set based on the recall information recalled in this recall process.

[0193] The second set acquisition module 1120 is used to filter the historical recall information in the historical recall process that has a historical score result higher than the first condition to obtain a second recall information set; the historical score result is used to indicate the probability that the target account will perform the target operation on the historical recall information at the time of historical recall.

[0194] The information recommendation module 1130 is used to provide the recommended recall information to the target account based on the first recall information set and the second recall information set.

[0195] In one possible implementation, the second set acquisition module 1120 is used to filter the historical recall information recalled in the historical recall process that has a historical rating result higher than the first condition and has not been recommended to the target account, so as to obtain the second recall information set.

[0196] In one possible implementation, the second set acquisition module 1120 includes:

[0197] A subset reading module is used to read historical information subsets corresponding to n record time periods in a historical information set; the historical recall time of the historical recall information in the historical information subset is within the historical time period corresponding to the historical information subset; the historical score result of the historical recall information in the historical information subset is higher than the first condition and has not been recommended to the target account; n is a positive integer;

[0198] The set acquisition submodule is used to acquire the second recall information set from the n subsets of historical information.

[0199] In one possible implementation, the collection acquisition submodule includes:

[0200] A subset acquisition unit is used to acquire, when generating the historical information subset for each of the recorded time periods in the n historical information subsets, the historical recall information subset corresponding to each of the i recommendation processes executed within the recorded time period; the historical recall information subset contains target historical recall information; the historical rating result of the target historical recall information is among the top k in the historical recall information subset; i and k are positive integers;

[0201] The first set acquisition unit is used to acquire a set of unrecommended information, which includes unrecommended information from the target historical recall information corresponding to each of the i recommendation processes.

[0202] The subset generation unit is used to generate the historical information subset corresponding to the recorded time period based on the set of unrecommended information.

[0203] In one possible implementation, the subset generation unit is configured to, in response to the number of unrecommended information in the set of unrecommended information being greater than a first recall threshold, sort the unrecommended information based on the historical rating results of each of the unrecommended information to obtain a first sorting result.

[0204] Based on the first sorting result, the number of unrecommended information equal to the first recall threshold is obtained, and a subset of historical information corresponding to the recorded time period is formed.

[0205] In one possible implementation, the subset acquisition submodule includes:

[0206] A time acquisition unit is used to acquire the recommendation time; the recommendation time is the time corresponding to this recommendation process.

[0207] The subset acquisition unit is used to acquire m target historical information subsets from n historical information subsets based on the recommendation time; m≤n, and m is a positive integer;

[0208] The second set acquisition unit is used to acquire the set of historical recall information composed of m target historical subsets as the second recall information set.

[0209] In one possible implementation, the subset acquisition unit is configured to:

[0210] The time interval is determined based on the time interval within the recorded time period in which the recommendation time occurs during the current recommendation process;

[0211] The target time period is the recording time period that is before the recording time period in which the current recommendation process takes place and differs from the recording time period in which the current recommendation process takes place by the time interval specified.

[0212] The historical information subset corresponding to the target time period is obtained into m target historical information subsets.

[0213] In one possible implementation, the second set acquisition module 1120 is used to filter the historical recommendation information in the historical recommendation information recommended in the historical recall process that did not receive the target operation performed by the target object, and obtain the second recall information set; the target object is the object corresponding to the target object account.

[0214] In one possible implementation, the information recommendation module 1130 includes:

[0215] The scoring acquisition submodule is used to acquire the current scoring result of each recall information in the recall information set; the recall information set includes the first recall information set and the second recall information set;

[0216] The recommendation information determination submodule is used to determine recommendation information based on the current scoring results of each recall information in the recall information set;

[0217] The information recommendation submodule is used to provide the recommendation information to the target account.

[0218] In one possible implementation, the recommendation information determination submodule includes:

[0219] The sorting unit is used to sort the recall information in the recall information set based on the current scoring result of each recall information in the recall information set, and obtain a second sorting result;

[0220] The recommendation information determination unit is used to determine the recommendation information based on the second ranking result.

[0221] In one possible implementation, the scoring acquisition submodule includes:

[0222] The object feature acquisition unit is used to acquire the object features of the target object account.

[0223] The scoring acquisition unit is used to acquire the current scoring result of each recall information based on the object characteristics of the target object account.

[0224] In one possible implementation, the scoring acquisition unit is used to input the object characteristics of the target object account and the target recall information into the scoring model to obtain the current scoring result of the target recall information output by the scoring model; the target recall information is any one of various recall information;

[0225] The scoring model is obtained by training a training sample set, which includes sample object features, sample information, and scoring labels of the sample object account relative to the sample object account.

[0226] In one possible implementation, the sample information in the training sample set is historical recommendation information; the sample object feature is the object feature of the recommendation object account corresponding to the historical recommendation information when the historical recommendation information was recommended; the rating tag of the sample information relative to the sample object account is the recommendation result of the historical recommendation information relative to the recommendation object account.

[0227] The recommendation result includes one of the following: the recommended object performs the target operation, or the recommended object does not perform the target operation; the recommended object is the object corresponding to the recommended object's account.

[0228] In one possible implementation, the first recall information set includes at least one of a first sub-information set, a second sub-information set, and a third sub-information set; the first sub-information set is an information set composed of recall information obtained from the candidate recall information based on keywords determined from the historical behavior data of the recommended object account; the second sub-information set is an information set composed of recall information obtained from the candidate recall information based on semantic vectors determined from the historical behavior data of the recommended object account; and the third sub-information set is an information set composed of recall information obtained from the candidate recall information based on information attributes.

[0229] In summary, the information recommendation device provided in this application obtains a first set of recall information based on the recall information recalled during the current recall process; obtains a second set of recall information based on the relationship between historical scoring results of historical recall information and a first condition; determines the recommended information for the current recommendation process from the information in the first and second sets of recall information, and recommends the recommended information. Through this scheme, in the current recall stage, historical recall results that meet the scoring conditions in previous recommendation processes are reused, realizing the correlation between different information recommendation processes, thereby improving the utilization effect of recall information and ultimately improving the information recommendation effect.

[0230] Figure 12 A structural block diagram of a computer device 1200 illustrating an exemplary embodiment of this application is shown. This computer device can be implemented as a server as described above in this application. The computer device 1200 includes a Central Processing Unit (CPU) 1201, a system memory 1204 including Random Access Memory (RAM) 1202 and Read-Only Memory (ROM) 1203, and a system bus 1205 connecting the system memory 1204 and the CPU 1201. The computer device 1200 also includes a mass storage device 1206 for storing an operating system 1209, application programs 1210, and other program modules 1211.

[0231] The mass storage device 1206 is connected to the central processing unit 1201 via a mass storage controller (not shown) connected to the system bus 1205. The mass storage device 1206 and its associated computer-readable media provide non-volatile storage for the computer device 1200. That is, the mass storage device 1206 may include computer-readable media (not shown) such as a hard disk or a compact disc read-only memory (CD-ROM) drive.

[0232] Without loss of generality, the computer-readable medium may include computer storage media and communication media. Computer storage media include volatile and non-volatile, removable and non-removable media implemented using any method or technology for storing information such as computer-readable instructions, data structures, program modules, or other data. Computer storage media include RAM, ROM, erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other solid-state storage technologies, CD-ROM, digital versatile disc (DVD) or other optical storage, magnetic tape cassettes, magnetic tape, disk storage, or other magnetic storage devices. Of course, those skilled in the art will recognize that the computer storage media are not limited to the above-mentioned types. The system memory 1204 and mass storage device 1206 described above can be collectively referred to as memory.

[0233] According to various embodiments of this disclosure, the computer device 1200 can also be connected to a remote computer on a network, such as the Internet. That is, the computer device 1200 can be connected to the network 1208 via a network interface unit 1207 connected to the system bus 1205, or the network interface unit 1207 can be used to connect to other types of networks or remote computer systems (not shown).

[0234] The memory further includes at least one instruction, at least one program, code set, or instruction set, which are stored in the memory. The central processing unit 1201 executes the at least one instruction, at least one program, code set, or instruction set to implement all or part of the steps in the information recommendation method shown in the above embodiments.

[0235] Figure 13 A structural block diagram of a computer device 1300 illustrating an exemplary embodiment of this application is shown. The computer device 1300 can be implemented as the aforementioned terminal, such as a smartphone, tablet computer, laptop computer, or desktop computer. The computer device 1300 may also be referred to as a terminal device, portable terminal, laptop terminal, desktop terminal, or other names.

[0236] Typically, computer device 1300 includes a processor 1301 and a memory 1302.

[0237] In some embodiments, the computer device 1300 may also optionally include a peripheral device interface 1303 and at least one peripheral device. The processor 1301, memory 1302, and peripheral device interface 1303 can be connected via a bus or signal line. Each peripheral device can be connected to the peripheral device interface 1303 via a bus, signal line, or circuit board. Specifically, the peripheral device includes at least one of the following: a radio frequency circuit 1304, a display screen 1305, a camera assembly 1306, an audio circuit 1307, and a power supply 1308.

[0238] In some embodiments, the computer device 1300 further includes one or more sensors 1309. The one or more sensors 1309 include, but are not limited to, an accelerometer 1310, a gyroscope 1311, a pressure sensor 1312, an optical sensor 1313, and a proximity sensor 1314.

[0239] Those skilled in the art will understand that Figure 13 The structure shown does not constitute a limitation on the computer device 1300, and may include more or fewer components than shown, or combine certain components, or use different component arrangements.

[0240] In one exemplary embodiment, a computer-readable storage medium is also provided for storing at least one computer program, which is loaded and executed by a processor to implement all or part of the steps in the information push method described above. For example, the computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a compact disc read-only memory (CD-ROM), magnetic tape, floppy disk, or optical data storage device, etc.

[0241] In one exemplary embodiment, a computer program product or computer program is also provided, the computer program product including at least one computer program, the computer program being loaded and executed by a processor as described above. Figure 3 , Figure 4 or Figure 8 All or part of the steps of the method shown in any embodiment.

[0242] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this application are indicated by the following claims.

[0243] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.

Claims

1. An information recommendation method, characterized in that, The method includes: Based on the recall information retrieved during this recall process, a first recall information set is obtained; In the historical recall process, historical recall information is filtered based on the recorded time period in which the historical recall time is located. Historical recall information with historical score results higher than the first condition and not recommended to the target account is filtered. The historical score results are used to indicate the probability that the target account corresponding to the target account will perform the target operation on the historical recall information at the historical recall time. In the historical recall information recalled during the historical recall process, historical recommendation information that has been recommended to the target account but has not been received by the target account to perform the target operation is filtered based on the recorded time period in which the historical recall time is located. Based on the filtered historical recall information and the filtered historical recommendation information, a second set of recall information is obtained; Based on the first recall information set and the second recall information set, the recall information recommended in this instance is provided to the target account.

2. The method according to claim 1, characterized in that, The historical recall information recalled during the historical recall process, which filters historical recall information based on the recorded time period in which the historical recall time occurred, includes historical recall information with historical rating results higher than the first condition and not recommended to the target account. Read the historical information subsets corresponding to each of the n recorded time periods in the historical information set; the historical recall time of the historical recall information in the historical information subset is within the historical time period corresponding to the historical information subset; the historical rating result of the historical recall information in the historical information subset is higher than the first condition, and it has not been recommended to the target account; n is a positive integer; Obtain the filtered historical recall information from the n subsets of historical information.

3. The method according to claim 2, characterized in that, The method further includes: When generating the historical information subset for each of the recorded time periods in the n historical information subsets, the historical recall information subset corresponding to each of the i recommendation processes executed within the recorded time period is obtained; the historical recall information subset contains target historical recall information; the historical rating result of the target historical recall information is among the top k in the historical recall information subset; i and k are positive integers; Obtain a set of unrecommended information, which contains unrecommended information from the target historical recall information corresponding to each of the i recommendation processes; Based on the set of unrecommended information, a subset of historical information corresponding to the recorded time period is generated.

4. The method according to claim 3, characterized in that, The step of generating the subset of historical information corresponding to the recorded time period based on the set of unrecommended information includes: In response to the fact that the number of unrecommended information in the set of unrecommended information is greater than a first recall threshold, the information is sorted based on the historical rating results of each unrecommended information to obtain a first sorting result; Based on the first sorting result, the number of unrecommended information equal to the first recall threshold is obtained, and a subset of historical information corresponding to the recorded time period is formed.

5. The method according to claim 2, characterized in that, The step of obtaining the second recall information set from the n subsets of historical information includes: Obtain the recommendation time corresponding to this recommendation process; Based on the recommended time, m target historical information subsets are obtained from n historical information subsets; m≤n, and m is a positive integer; The set of historical recall information from the m subsets of the target historical information is obtained as the second recall information set.

6. The method according to claim 5, characterized in that, Based on the recommendation time, obtaining m target historical information subsets from n historical information subsets includes: The time interval is determined based on the time interval within the recorded time period in which the recommendation time occurs during the current recommendation process; The target time period is the recording time period that is before the recording time period in which the current recommendation process takes place and differs from the recording time period in which the current recommendation process takes place by the time interval specified. The historical information subset corresponding to the target time period is obtained into m target historical information subsets.

7. The method according to any one of claims 1 to 6, characterized in that, The step of providing the recommended recall information to the target account based on the first recall information set and the second recall information set includes: Obtain the current scoring result of each recall information in the recall information set; the recall information set includes the first recall information set and the second recall information set; Based on the current scoring results of each recall information in the recall information set, recommended information is determined; The recommendation information is provided to the target account.

8. The method according to claim 7, characterized in that, The determination of recommendation information based on the current scoring results of each recall information in the recall information set includes: Based on the current scoring results of each recall information in the recall information set, the recall information in the recall information set is sorted to obtain a second sorting result; Based on the second ranking result, the recommended information is determined.

9. The method according to claim 7, characterized in that, The current scoring results for each recall information in the recall information set include: Obtain the object characteristics of the target account; Based on the object characteristics of the target account, the current scoring results of each recall information are obtained respectively.

10. The method according to claim 9, characterized in that, The scoring results for each recall information are obtained based on the object characteristics of the target account, including: The target characteristics of the target account and the target recall information are input into the scoring model to obtain the current scoring result of the target recall information output by the scoring model; the target recall information is any one of the various recall information; The scoring model is obtained by training a training sample set, which includes sample object features, sample information, and scoring labels of the sample object account relative to the sample object account.

11. The method according to claim 10, characterized in that, The sample information in the training sample set is historical recommendation information; the sample object feature is the object feature of the recommendation object account corresponding to the historical recommendation information when the historical recommendation information was recommended. The rating tag of the sample information relative to the sample object account is the recommendation result of the historical recommendation information relative to the recommendation object account; The recommendation result includes one of the following: the recommended object performs the target operation, or the recommended object does not perform the target operation; the recommended object is the object corresponding to the recommended object's account.

12. The method according to any one of claims 1 to 6, characterized in that, The first recall information set includes at least one of a first sub-information set, a second sub-information set, and a third sub-information set; the first sub-information set is an information set composed of recall information obtained from candidate recall information based on keywords determined by the historical behavior data of the target account; the second sub-information set is an information set composed of recall information obtained from the candidate recall information based on semantic vectors determined by the historical behavior data of the target account; and the third sub-information set is an information set composed of recall information obtained from the candidate recall information based on information attributes.

13. An information recommendation device, characterized in that, The device includes: The first set acquisition module is used to obtain the first recall information set based on the recall information recalled in this recall process. The second set acquisition module is used to filter historical recall information retrieved during the historical recall process based on the recording time period in which the historical recall time occurred, for historical recall information with historical score results higher than the first condition and not recommended to the target account. The historical score result is used to indicate the probability that the target account corresponding to the target account will perform the target operation on the historical recall information at the historical recall time. The module also filters historical recommendation information retrieved during the historical recall process based on the recording time period in which the historical recall time occurred, for historical recommendation information that has been recommended to the target account but has not received the target account's execution of the target operation. Based on the filtered historical recall information and the filtered historical recommendation information, a second recall information set is obtained. The information recommendation module is used to provide the recommended recall information to the target account based on the first recall information set and the second recall information set.

14. The apparatus according to claim 13, characterized in that, The second set acquisition module includes: The subset reading module is used to read the historical information subsets corresponding to each of the n recorded time periods in the historical information set; the historical recall time of the historical recall information in the historical information subset is within the historical time period corresponding to the historical information subset; the historical score result of the historical recall information in the historical information subset is higher than the first condition and has not been recommended to the target account; n is a positive integer; The set acquisition submodule is used to acquire the filtered historical recall information from the n subsets of historical information.

15. The apparatus according to claim 14, characterized in that, The collection acquisition submodule includes: A subset acquisition unit is used to acquire, when generating the historical information subset for each of the recorded time periods in the n historical information subsets, the historical recall information subset corresponding to each of the i recommendation processes executed within the recorded time period; the historical recall information subset contains target historical recall information; the historical rating result of the target historical recall information is among the top k in the historical recall information subset; i and k are positive integers; The first set acquisition unit is used to acquire a set of unrecommended information, which includes unrecommended information from the target historical recall information corresponding to each of the i recommendation processes. The subset generation unit is used to generate the historical information subset corresponding to the recorded time period based on the set of unrecommended information.

16. The apparatus according to claim 15, characterized in that, The subset generation unit is configured to, in response to the number of unrecommended information in the set of unrecommended information being greater than a first recall threshold, sort the unrecommended information based on the historical rating results of each unrecommended information to obtain a first sorting result. Based on the first sorting result, the number of unrecommended information equal to the first recall threshold is obtained, and a subset of historical information corresponding to the recorded time period is formed.

17. The apparatus according to claim 14, characterized in that, The subset acquisition submodule includes: The time acquisition unit is used to acquire the recommendation time corresponding to this recommendation process; The subset acquisition unit is used to acquire m target historical information subsets from n historical information subsets based on the recommendation time; m≤n, and m is a positive integer; The second set acquisition unit is used to acquire the set of historical recall information composed of m subsets of the target historical information as the second recall information set.

18. The apparatus according to claim 17, characterized in that, The subset acquisition unit is used for: The time interval is determined based on the time interval within the recorded time period in which the recommendation time occurs during the current recommendation process; The target time period is the recording time period that is before the recording time period in which the current recommendation process takes place and differs from the recording time period in which the current recommendation process takes place by the time interval specified. The historical information subset corresponding to the target time period is obtained into m target historical information subsets.

19. The apparatus according to any one of claims 13 to 18, characterized in that, The information recommendation module includes: The scoring acquisition submodule is used to acquire the current scoring result of each recall information in the recall information set; the recall information set includes the first recall information set and the second recall information set; The recommendation information determination submodule is used to determine recommendation information based on the current scoring results of each recall information in the recall information set; The information recommendation submodule is used to provide the recommendation information to the target account.

20. The apparatus according to claim 19, characterized in that, The recommendation information determination submodule includes: The sorting unit is used to sort the recall information in the recall information set based on the current scoring result of each recall information in the recall information set, and obtain a second sorting result; The recommendation information determination unit is used to determine the recommendation information based on the second ranking result.

21. The apparatus according to claim 19, characterized in that, The scoring acquisition submodule includes: An object feature acquisition unit is used to acquire the object features of the target object account; The scoring acquisition unit is used to acquire the current scoring result of each recall information based on the object characteristics of the target object account.

22. The apparatus according to claim 21, characterized in that, The scoring acquisition unit is used to input the object characteristics of the target object account and the target recall information into the scoring model to obtain the current scoring result of the target recall information output by the scoring model; the target recall information is any one of the various recall information; The scoring model is obtained by training a training sample set, which includes sample object features, sample information, and scoring labels of the sample object account relative to the sample object account.

23. The apparatus according to claim 22, characterized in that, The sample information in the training sample set is historical recommendation information; the sample object feature is the object feature of the recommendation object account corresponding to the historical recommendation information when the historical recommendation information was recommended. The rating tag of the sample information relative to the sample object account is the recommendation result of the historical recommendation information relative to the recommendation object account; The recommendation result includes one of the following: the recommended object performs the target operation, or the recommended object does not perform the target operation; the recommended object is the object corresponding to the recommended object's account.

24. The apparatus according to any one of claims 13 to 18, characterized in that, The first recall information set includes at least one of a first sub-information set, a second sub-information set, and a third sub-information set; the first sub-information set is an information set composed of recall information obtained from candidate recall information based on keywords determined by the historical behavior data of the target account; the second sub-information set is an information set composed of recall information obtained from the candidate recall information based on semantic vectors determined by the historical behavior data of the target account; and the third sub-information set is an information set composed of recall information obtained from the candidate recall information based on information attributes.

25. A computer device, characterized in that, The computer device includes a processor and a memory, the memory storing at least one computer program, the at least one computer program being loaded and executed by the processor to implement the information recommendation method as described in any one of claims 1 to 12.

26. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores at least one computer program, which is loaded and executed by a processor to implement the information recommendation method as described in any one of claims 1 to 12.

27. A computer program product, characterized in that, The computer program product includes at least one computer program, which is loaded and executed by a processor to implement the information recommendation method as described in any one of claims 1 to 12.

Images