Information processing method and device, electronic equipment, computer readable storage medium and computer program product

Abstract

The invention provides an information processing method and device, electronic equipment, a computer readable storage medium and a computer program product. The information processing method and device are applied to various information recommendation scenes such as cloud technology, artificial intelligence, news, games and intelligent transportation. The information processing method comprises the steps of predicting M current interest values of a first recommendation object sample for M pieces of recommendation information based on a historical action data sequence executed by the first recommendation object sample for the recommendation information; selecting N maximum current interest values from the M current interest values; selecting L recommendation information negative samples corresponding to the N maximum current interest values from the M recommendation information; combining the first recommendation object sample and the L recommendation information negative samples to construct a negative sample data set; and training the first to-be-trained model based on the negative sample data set to obtain an information recommendation model, the information recommendation model being used for predicting a conversion rate of the to-be-recommended object for the to-be-recommended information. According to the invention, the accuracy of model training can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of artificial intelligence, and particularly relates to an information processing method and device, an electronic device, a computer readable storage medium and a computer program product. BACKGROUND

[0002] Information recommendation refers to a process of recommending recommendation information to a recommendation object, and the information recommendation can be implemented by using a neural network model. When training the neural network model in the field of information recommendation, a first to-be-trained model is often supervised trained by using a positive and negative sample data set. However, the negative sample data set is usually randomly selected during model training, which affects the quality of the negative sample data set and further affects the accuracy of model training. SUMMARY

[0003] Embodiments of the present application provide an information processing method and device, an electronic device, a computer readable storage medium and a computer program product, which can prompt the accuracy of model training.

[0004] The technical solution of the embodiments of the present application is implemented as follows:

[0005] An information processing method is provided in the embodiments of the present application, and the method comprises the following steps:

[0006] Obtaining a historical action data sequence of a first recommendation object sample on recommendation information;

[0007] Based on the historical action data sequence, predicting M current interest values of the first recommendation object sample on M recommendation information, M>1, and M is an integer;

[0008] From the M current interest values, selecting N maximum current interest values, N is a positive integer;

[0009] From the M recommendation information, selecting L recommendation information negative samples corresponding to the N maximum current interest values, L is a positive integer, and L≤N;

[0010] Combining the first recommendation object sample and the L recommendation information negative samples, constructing a negative sample data set;

[0011] Training a first to-be-trained model based on the negative sample data set to obtain an information recommendation model, the information recommendation model being used to predict a conversion rate of a to-be-recommended object on to-be-recommended information.

[0012] An information processing device is provided in the embodiments of the present application, and the information processing device comprises:

[0013] A data collection module is configured to obtain a historical action data sequence of a first recommendation object sample on recommendation information;

[0014] an interest prediction module, configured to predict M current interest values of the first recommendation object sample on M recommendation information based on the historical action data sequence, M>1, and M is an integer;

[0015] a sample construction module, configured to select N maximum current interest values from the M current interest values, N is a positive integer;

[0016] The sample construction module is further configured to select L recommendation information negative samples corresponding to the N maximum current interest values from the M recommendation information, L is a positive integer, and L≤N.

[0017] The sample construction module is further configured to construct a negative sample data set in combination with the first recommendation object sample and the L recommendation information negative samples.

[0018] a model training module, configured to train a first to-be-trained model based on the negative sample data set to obtain an information recommendation model, the information recommendation model being used to predict a conversion rate of a to-be-recommended object on to-be-recommended information.

[0019] In an embodiment of the present application, the data collection module is further configured to, for each historical moment in the historical moment sequence, perform the following processing: obtaining historical action data of the first recommendation object sample on the recommendation information at the historical moment; obtaining the historical action data sequence corresponding to the historical moment sequence from the historical action data corresponding to each historical moment.

[0020] In an embodiment of the present application, the interest prediction module is further configured to predict the M current interest values of the first recommendation object sample on the M recommendation information based on data features corresponding to each historical action data in the historical action data sequence.

[0021] In an embodiment of the present application, the sample construction module is further configured to select N recommendation information corresponding to the N maximum current interest values from the M recommendation information; select P recommendation information in which the first recommendation object sample has not been converted from the M recommendation information, P is a positive integer, and the not being converted refers to not having occurred conversion; obtain L common recommendation information between the N recommendation information and the P recommendation information; and determine the L common recommendation information as the L recommendation information negative samples.

[0022] In the embodiment of the present application, the sample construction module is further configured to: obtain object association information of the first recommended object sample; combine the object association information with L negative sample pieces of recommendation information respectively to obtain L negative sample information pieces, wherein the negative sample information piece includes the object association information, the negative sample piece of recommendation information, and an unconverted label of the first recommended object sample for the negative sample piece of recommendation information; and determine the L negative sample information pieces as the negative sample data set.

[0023] In the embodiment of the present application, the model training module is further configured to: select Q positive sample pieces of recommendation information in which the first recommended object sample has been converted from M pieces of recommendation information, Q being a positive integer; construct a positive sample data set in combination with the first recommended object sample and the Q positive sample pieces of recommendation information; and train the first to-be-trained model in combination with the positive sample data set and the negative sample data set to obtain the information recommendation model.

[0024] In the embodiment of the present application, the model training module is further configured to: combine object association information of the first recommended object sample with Q positive sample pieces of recommendation information respectively to obtain Q positive sample information pieces, wherein the positive sample information piece includes the object association information, the positive sample piece of recommendation information, and a converted label of the first recommended object sample for the positive sample piece of recommendation information; and determine the Q positive sample information pieces as the positive sample data set.

[0025] In the embodiment of the present application, the model training module is further configured to, for each negative sample information piece in the negative sample data set: use the first to-be-trained model to predict the object association information and the negative sample piece of recommendation information in the negative sample information piece to obtain a first estimated conversion rate of the first recommended object sample for the negative sample piece of recommendation information; calculate a first loss function value based on a difference between the first estimated conversion rate and an unconverted label in the negative sample information piece; and train the first to-be-trained model based on the first loss function value to obtain the information recommendation model.

[0026] In the embodiment of the present application, the model training module is further configured to, for each positive sample information piece in the positive sample data set: use the first to-be-trained model to predict the object association information and the positive sample piece of recommendation information in the positive sample information piece to obtain a second estimated conversion rate of the first recommended object sample for the positive sample piece of recommendation information.

[0027] In the embodiment of the present application, the model training module is further configured to calculate the first loss function value based on a difference between the first estimated conversion rate and the unconverted label in the negative sample information piece, and a difference between the second estimated conversion rate and a converted label in the positive sample information piece.

[0028] In the embodiment of the present application, the information processing apparatus further comprises an information recommendation module configured to, in response to an information recommendation request, obtain the to-be-recommended object and the to-be-recommended information; perform feature extraction on the to-be-recommended object to obtain to-be-recommended object features; perform feature extraction on the to-be-recommended information to obtain to-be-recommended information features; use the information recommendation model to predict the to-be-recommended object features and the to-be-recommended information features to obtain a target conversion rate; and determine a target recommendation result based on the target conversion rate, the target recommendation result indicating whether to recommend the to-be-recommended information to the to-be-recommended object.

[0029] In the embodiment of the present application, the prediction of the M current interest values of the first recommendation object sample for the M recommendation information based on the historical action data sequence is achieved by using an interest prediction model; the model training module is further configured to obtain a recommendation information sample sequence of an action performed by a second recommendation object sample and a recommendation information label of an action performed by the second recommendation object sample at a next action execution time, the next action execution time being a time next to a last action execution time corresponding to the recommendation information sample sequence; use a second to-be-trained model to predict the recommendation information sample sequence to obtain an estimated recommendation information of the action performed by the second recommendation object sample at the next action execution time; calculate a second loss function value based on a difference between the estimated recommendation information and the recommendation information label; and train the second to-be-trained model based on the second loss function value to obtain the interest prediction model.

[0030] In the embodiment of the present application, the model training module is further configured to obtain a recommendation information initial sequence of an action performed by the second recommendation object sample and a number K of recommendation information of the recommendation information initial sequence, K being an integer greater than 1; start iteration from a lowest sequence number k, and perform the following processing through iteration k, k≤K-1: from a first recommendation information in the recommendation information initial sequence, sequentially select k recommendation information; construct the selected k recommendation information into the recommendation information sample sequence; and determine a k+1 recommendation information in the recommendation information initial sequence as the recommendation information label of the action performed by the second recommendation object sample at the next action execution time.

[0031] In the embodiment of the present application, the number of model parameters of the interest prediction model is greater than the number of model parameters of the information recommendation model.

[0032] The embodiment of the present application provides an electronic device for information processing, which comprises:

[0033] a memory configured to store computer executable instructions or computer programs;

[0034] The processor is configured to execute the computer-executable instructions or the computer program stored in the memory to implement the information processing method provided in the embodiments of the present application.

[0035] The embodiments of the present application provide a computer readable storage medium storing computer-executable instructions or a computer program, and the computer-executable instructions or the computer program are configured to be executed by a processor to implement the information processing method provided in the embodiments of the present application.

[0036] The embodiments of the present application provide a computer program product comprising computer-executable instructions or a computer program, and the computer-executable instructions or the computer program are configured to be executed by a processor to implement the information processing method provided in the embodiments of the present application.

[0037] The embodiments of the present application have at least the following beneficial effects: in constructing the negative sample data set based on the first recommended object sample, M current interest values of M recommended information are predicted based on the historical action data sequence of the first recommended object sample, and L recommended information negative samples are determined from the M recommended information based on N maximum current interest values, so as to construct the negative sample data set by combining the first recommended object sample and the L recommended information negative samples; in this way, the recommended information negative samples in the constructed negative sample data set are the recommended information negative samples that the first recommended object sample is interested in, the quality of the negative sample data set is improved, and when model training is performed based on the negative sample data set, the accuracy of the model training can be improved. BRIEF DESCRIPTION OF DRAWINGS

[0038] Figure 1 FIG. 1 is a schematic diagram of an architecture of an information processing system provided in the embodiments of the present application;

[0039] Figure 2 FIG. 2 is a schematic diagram of a structure of a server in the information processing system provided in the embodiments of the present application; Figure 1

[0040] Figure 3 FIG. 3 is a flowchart of an information processing method provided in the embodiments of the present application; Figure 1

[0041] Figure 4 FIG. 4 is a flowchart of an information processing method provided in the embodiments of the present application; Figure 2

[0042] Figure 5 FIG. 5 is a flowchart of an information processing method provided in the embodiments of the present application; Figure 3

[0043] Figure 6 FIG. 6 is a flowchart of an information processing method provided in the embodiments of the present application; Figure 4 FIG. 7 is a flowchart of an information processing method provided in the embodiments of the present application.​​​​

[0044] Figure 7 is a flowchart of an information processing method provided by an embodiment of the present application Figure 5 ;

[0045] Figure 8 is an exemplary schematic diagram of training a business model provided by an embodiment of the present application;

[0046] Figure 9 is an exemplary schematic diagram of training a generative model provided by an embodiment of the present application;

[0047] Figure 10 is an exemplary schematic diagram of recommending game props provided by an embodiment of the present application. DETAILED DESCRIPTION

[0048] In order to make the purposes, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the drawings, and the described embodiments should not be regarded as limiting the present application. All other embodiments obtained by those of ordinary skill in the art without making creative labor fall within the scope of protection of the present application.

[0049] In the following description, "some embodiments" are described, which describe a subset of all possible embodiments, but it can be understood that "some embodiments" can be the same subset or different subsets of all possible embodiments, and can be combined with each other without conflict.

[0050] In the following description, the terms "first\second" are used to distinguish similar objects, and do not represent a specific order of the objects. It can be understood that "first\second" can be interchanged with a specific order or sequence as allowed, so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein.

[0051] In the embodiments of the present application, the term "module" or "unit" refers to a computer program or a part of a computer program with a predetermined function, and works with other related parts to achieve a predetermined target, and can be implemented entirely or partially by using software, hardware (such as a processing circuit or a memory) or a combination thereof. Similarly, one processor (or multiple processors or memories) can be used to implement one or more modules or units. In addition, each module or unit can be a part of an overall module or unit that includes the functions of the module or unit.

[0052] Unless otherwise defined, all technical and scientific terms used in the embodiments of the present application have the same meanings as those commonly understood by those skilled in the art. The terms used in the embodiments of the present application are only for the purpose of describing the embodiments of the present application and are not intended to limit the present application.

[0053] The related data collection and processing in the embodiments of the present application should strictly comply with the requirements of relevant national laws and regulations, obtain the informed consent or separate consent of the personal information subject, and carry out subsequent data use and processing within the scope of authorization of the laws and regulations and the personal information subject.

[0054] Before the embodiments of the present application are further described in detail, the terms and phrases involved in the embodiments of the present application are explained, and the terms and phrases involved in the embodiments of the present application are applicable to the following explanations.

[0055] 1) Artificial Intelligence (AI) is the use of digital computers or computer-controlled machines to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results. That is, artificial intelligence is a comprehensive technology of computer science, which is used to understand the essence of intelligence and produce a new intelligent machine that can react in a similar way to human intelligence. Artificial intelligence studies the design principles and implementation methods of various intelligent machines, enabling machines to have perception, reasoning and decision-making functions.

[0056] It should be noted that artificial intelligence technology is a comprehensive discipline, involving a wide range of fields, both hardware and software technologies. Artificial intelligence basic technologies generally include sensors, special artificial intelligence chips, cloud computing, distributed storage, big data processing technology, pre-training model technology, operation / interaction system, mechatronics, etc. Among them, the pre-training model is also called a large model, a basic model; the pre-training model can be widely applied to downstream tasks in various directions of artificial intelligence after fine-tuning. Artificial intelligence software technologies include computer vision technology, speech processing technology, natural language processing technology, and machine learning / deep learning, etc. In the embodiments of the present application, information recommendation can be realized through artificial intelligence technology.

[0057] 2) Machine Learning (ML) is a multi-disciplinary subject that involves probability theory, statistics, approximation theory, convex analysis, and algorithm complexity theory. It is used to study the computer simulation or implementation of human learning behavior to acquire new knowledge or skills; reorganize the existing knowledge structure to continuously improve the performance. Machine learning is the core of artificial intelligence and the fundamental approach to making computers intelligent. Machine learning is applied in various fields of artificial intelligence. Machine learning / deep learning usually includes artificial neural networks, belief networks, reinforcement learning, transfer learning and inductive learning, etc. Large models are the latest development of machine learning / deep learning, which integrate the above technologies. In the embodiments of the present application, information recommendation can be realized by combining machine learning / deep learning.

[0058] 3) Artificial neural network, which is a mathematical model that simulates the structure and function of biological neural networks. In the embodiments of the present application, exemplary structures of artificial neural networks include Graph Convolutional Network (GCN, a neural network for processing graph-structured data), Deep Neural Networks (DNN), Convolutional Neural Network (CNN) and Recurrent Neural Network (RNN), Neural State Machine (NSM) and Phase-Functioned Neural Network (PFNN), etc. In the embodiments of the present application, information recommendation can be realized through a network model of artificial neural network (referred to as neural network model). In addition, the first to-be-trained model, the second to-be-trained model, the information recommendation model and the interest prediction model provided in the embodiments of the present application are all neural network models. The second to-be-trained model and the interest prediction model can be large models, the second to-be-trained model is a large model to be fine-tuned, and the interest prediction model is a large model after fine-tuning.

[0059] 4) Recommendation object, which refers to the object of recommended information, and can be a single object or a group of objects. For example, game players, instant messaging accounts and forum accounts in the game field.

[0060] 5) Recommended information, which refers to the information recommended to the recommendation object, and can be a single information or multiple information. For example, game props, game applications, news, videos, etc. in the game field.

[0061] 6) Conversion, which refers to the specified processing of recommended information by the recommendation object, such as resource exchange, account registration, service reservation, information download, information browsing, information playing, information sharing, information commenting, etc. Conversion rate refers to the probability of conversion, indicating the possibility of conversion.

[0062] It should be noted that in order to train the neural network model in the field of information recommendation, the first to-be-trained model is often supervised trained by using positive and negative sample data sets. However, in the model training, the negative sample data set is usually randomly selected, which affects the quality of the negative sample data set, and further affects the accuracy of the model training.

[0063] Exemplarily, when the recommendation information is a game prop, the game prop that has been transformed (for example, has been exchanged for resources) is used to construct the positive sample data set, and the game prop that has not been transformed is used to construct the negative sample data set; since the number of game props that have not been transformed is greater than the prop quantity threshold, the game prop that has not been transformed is often selected in a random manner, which increases the distance between the positive sample data set and the negative sample data set, causes the game prop recommendation model to converge in advance in the training process, and further affects the accuracy of the game prop recommendation model.

[0064] Based on this, the embodiment of the present application provides an information processing method and device, electronic equipment, computer readable storage medium and computer program product, which can improve the accuracy of model training. The following describes an exemplary application of an electronic device for information processing (hereinafter referred to as an information processing device) provided by the embodiment of the present application. The information processing device provided by the embodiment of the present application can be implemented as a robot, a smart phone, a smart watch, a notebook computer, a tablet computer, a desktop computer, a smart home appliance, a set-top box, a smart car device, a portable music player, a personal digital assistant, a dedicated message device, a smart voice interaction device, a portable game device, and a smart speaker, etc. Various types of terminals, and can also be implemented as a server. The following describes an exemplary application when the information processing device is implemented as a server.

[0065] Referring to Figure 1 , Figure 1 is an architecture diagram of an information processing system provided by the embodiment of the present application; as Figure 1 indicated, to support an information processing application, in the information processing system 100, the terminal 200 (exemplarily shows the terminal 200-1 and the terminal 200-2) connects the server 400 (referred to as an information processing device) through the network 300. The network 300 can be a wide area network or a local area network, or a combination of the two. In addition, the information processing system 100 also includes a database 500 for providing data support to the server 400; and Figure 1 indicated in the figure is a case where the database 500 is independent of the server 400. In addition, the database 500 can also be integrated in the server 400, which is not limited by the embodiment of the present application.

[0066] The terminal 200 is configured to send an information recommendation request to the server 400 through the network 300, and receive a target recommendation result sent by the server 400 through the network 300. When the target recommendation result indicates that the to-be-recommended information is recommended to the to-be-recommended object, the to-be-recommended information (exemplarily shows a graphical interface 210-1 for recommending a game prop and a graphical interface 210-2 for recommending a video) is displayed.

[0067] Server 400 is configured to obtain a historical action data sequence of a first recommended object sample on recommended information; based on the historical action data sequence, predict M current interest values ​​of the first recommended object sample on M recommended information, where M>1 and M is an integer; select N maximum current interest values ​​from the M current interest values, where N is a positive integer; select L negative samples of recommended information corresponding to the N maximum current interest values ​​from the M recommended information, where L is a positive integer and L≤N; construct a negative sample data set by combining the first recommended object sample and the L negative samples of recommended information; and train a first to-be-trained model based on the negative sample data set. An information recommendation model is obtained, and the information recommendation model is used to predict the conversion rate of the object to be recommended to the information to be recommended; the network 300 receives the information recommendation request sent by the terminal 200, and obtains the object to be recommended and the information to be recommended in response to the information recommendation request; features are extracted from the object to be recommended to obtain features of the object to be recommended; features are extracted from the information to be recommended to obtain features of the information to be recommended; the information recommendation model is used to predict the features of the object to be recommended and the features of the information to be recommended to obtain a target conversion rate; based on the target conversion rate, a target recommendation result is determined; and the target recommendation result is sent to the terminal 200 via the network 300.

[0068] In some embodiments, server 400 may be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms. The terminal and the server may be connected directly or indirectly via wired or wireless communication, which is not limited in the embodiments of the present application.

[0069] See also Figure 2 , Figure 2 This embodiment of the present application provides a Figure 1 The structural diagram of the server in Figure 2 As shown, the server 400 includes: at least one processor 410, a memory 450 and at least one network interface 420. The various components in the server 400 are coupled together via a bus system 440. It is understood that the bus system 440 is used to achieve connection and communication between these components. In addition to the data bus, the bus system 440 also includes a power bus, a control bus and a status signal bus. However, for the sake of clarity, Figure 2 Various buses are labeled as bus system 440 .

[0070] The processor 410 can be an integrated circuit chip having a processing capability of signals, such as a general purpose processor, a digital signal processor (DSP), or other programmable logic device, discrete gate or transistor logic device, discrete hardware component, etc., wherein the general purpose processor can be a microprocessor or any conventional processor.

[0071] The memory 450 can be removable, non-removable, or a combination thereof. Exemplary hardware devices include solid-state memory, hard disk drives, optical drives, etc. The memory 450 optionally includes one or more storage devices remotely located from the processor 410 in physical location.

[0072] The memory 450 includes volatile memory or non-volatile memory, and can also include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), and the volatile memory can be random access memory (RAM). The memory 450 described in the embodiments of the present application is intended to include any suitable type of memory.

[0073] In some embodiments, the memory 450 is capable of storing data to support various operations, examples of which include programs, modules, and data structures or subsets or supersets thereof, which are exemplarily illustrated below.

[0074] The operating system 451 includes system programs for processing various basic system services and performing hardware-related tasks, such as a framework layer, a core library layer, a driver layer, etc., for implementing various basic services and processing hardware-based tasks.

[0075] The network communication module 452 is used to communicate with other electronic devices via one or more (wired or wireless) network interfaces 420, and exemplary network interfaces 420 include Bluetooth, wireless compatibility certification (Wi-Fi), and universal serial bus (USB), etc.

[0076] In some embodiments, the information processing apparatus provided by the embodiments of the present application can be realized in a software manner, Figure 2 An information processing apparatus 455 stored in the memory 450 is shown, which can be software in the form of programs and plug-ins, etc., including the following software modules: a data acquisition module 4551, an interest prediction module 4552, a sample construction module 4553, a model training module 4554, and an information recommendation module 4555, which are logical, and thus can be combined or further split according to the implemented functions. The functions of each module will be described below.

[0077] In some embodiments, the information processing apparatus provided by the embodiments of the present application can be implemented in a hardware manner. For example, the information processing apparatus provided by the embodiments of the present application can be a processor in the form of a hardware decoding processor, which is programmed to perform the information processing method provided by the embodiments of the present application. For example, the processor in the form of a hardware decoding processor can be implemented by using one or more Application Specific Integrated Circuits (ASICs), DSPs, Programmable Logic Devices (PLDs), Complex Programmable Logic Devices (CPLDs), Field-Programmable Gate Arrays (FPGAs), or other electronic elements.

[0078] In some embodiments, the terminal or server can implement the information processing method provided by the embodiments of the present application by running various computer-executable instructions or computer programs. For example, the computer-executable instructions can be microprogram-level commands, machine instructions, or software instructions. The computer program can be a native program or a software module in an operating system; can be a native application (APP), i.e., a program that needs to be installed in an operating system to run, such as a live APP or a game APP; or can be a small program that can be embedded into any APP, i.e., a program that only needs to be downloaded into a browser environment to run. In summary, the above computer-executable instructions can be any form of instructions, and the above computer programs can be any form of application programs, modules, or plug-ins.

[0079] In the following, the information processing method provided by the embodiments of the present application will be described in conjunction with an exemplary application and implementation of the information processing apparatus provided by the embodiments of the present application. In addition, the information processing method provided by the embodiments of the present application is applied to various information recommendation scenarios such as cloud technology, artificial intelligence, news, games, and intelligent transportation.

[0080] Referring to Figure 3 , Figure 3 is a flowchart of the information processing method provided by the embodiments of the present application Figure 1 , wherein Figure 3 The execution subject of each step in Figure 3 will be described in conjunction with the steps shown in

[0081] Step 101, obtaining a historical action data sequence performed by a first recommended object sample on recommended information.

[0082] In the embodiment of the present application, when the information processing device determines to construct the training sample based on the first recommended object sample, the information processing device determines a historical time sequence before the current action execution time (i.e., the time when the first recommended object sample performs the action on the recommendation information, referred to as the current action execution time), and acquires a historical action data sequence of the first recommended object sample performed on the recommendation information at the historical time sequence.

[0083] It should be noted that the first recommended object sample refers to a recommended object used to train the first to-be-trained model; and the first to-be-trained model refers to a neural network model to be trained for predicting the conversion rate of the recommended object on the recommendation information, which can be an original neural network model constructed, or a pre-trained neural network model, etc., which is not limited in the embodiment of the present application. In addition, each constructed training sample corresponds to an action execution time, which represents the time when the recommended object in the training sample performs the action on the recommendation information; therefore, the historical time sequence refers to a sequence composed of at least one action execution time before the current action execution time; therefore, one historical time in the historical time sequence is one action execution time before the current action execution time, and one historical action data in the historical action data sequence corresponds to one historical time in the historical time sequence; wherein the historical action data refers to the data of the action performed by the first recommended object sample on the recommendation information at the historical time. In addition, the historical time refers to the time when the first recommended object sample performs the action on the recommendation information before the current action execution time.

[0084] In the embodiment of the present application, the information processing device acquires the historical action data sequence of the first recommended object sample performed on the recommendation information, comprising: the information processing device performs the following processing for each historical time in the historical time sequence: the information processing device first acquires the historical action data of the first recommended object sample performed on the recommendation information at the historical time; and then obtains the historical action data sequence corresponding to the historical time sequence from the historical action data corresponding to each historical time.

[0085] It should be noted that the historical action data can include the first recommended object sample, recommended information of the executed action, and the executed action; wherein the executed action belongs to the first conversion, the first conversion is a kind of conversion, which refers to the full amount type of action executed by the recommended object on the recommended information, and the executed action includes one of the following: resource exchange, registration, click, browse, stay, download, share, like, comment, appointment, application and play, etc. Here, since the information processing device obtains a corresponding historical action data for each historical moment in the historical moment sequence, at least one historical action data corresponding to at least one historical moment in the historical moment sequence can be obtained, and at least one historical action data is combined into a sequence based on the historical moment sequence, so as to obtain the historical action data sequence.

[0086] Step 102, based on the historical action data sequence, predicting M current interest values of the first recommended object sample to M recommended information.

[0087] In the embodiment of the present application, the information processing device analyzes the historical action data sequence to predict the possibility (also known as probability) of the first recommended object sample executing actions on the M recommended information at the current action execution moment, that is, to obtain the M current interest values; therefore, the current interest value refers to the probability of the first recommended object sample executing actions on the recommended information at the current action execution moment. Here, the information processing device can directly predict the M current interest values based on the corresponding features of the historical action data sequence, and can also use a neural network model to process the historical action data sequence to predict the M current interest values, which is not limited in the embodiment of the present application.

[0088] It should be noted that the M recommended information and the M current interest value are one-to-one corresponding; wherein M>1, and M is an integer, that is, the M recommended information is multiple recommended information, for example, all the game props in a specified game.

[0089] In the embodiment of the present application, when the historical action data sequence is composed of the historical action data corresponding to each historical moment in the historical moment sequence, correspondingly, in step 102, the information processing device predicts M current interest values of the first recommended object sample to M recommended information based on the historical action data sequence, including: the information processing device predicts M current interest values of the first recommended object sample to M recommended information based on the data features corresponding to each historical action data in the historical action data sequence.

[0090] It should be noted that the data features corresponding to the historical action data refer to the embedding representation corresponding to the historical action data.

[0091] Step 103, selecting N maximum current interest values from the M current interest values.

[0092] In the embodiment of the present application, the information processing device ranks the M current interest values to obtain a current interest value sequence, selects N largest current interest values from the current interest value sequence, and takes the selected N largest current interest values as the N largest current interest values; that is, the largest current interest value is one of the first N largest current interest values among the M current interest values. Of course, the information processing device can also obtain the N largest current interest values by comparing the M current interest values; the embodiment of the present application does not limit this. Wherein, N is a positive integer.

[0093] It should be noted that the N largest current interest values are the N current interest values with the largest current interest values among the M current interest values. In addition, the ranking of the M current interest values by the information processing device can be a positive sequence ranking, or can be a reverse sequence ranking, and the embodiment of the present application does not limit this. When the ranking of the M current interest values is a positive sequence ranking, the N current interest values are selected from the tail of the current interest value sequence to obtain the N largest current interest values, at this time, the N largest current interest values are values with a specified percentage of the rear in the current interest value sequence. When the ranking of the M current interest values is a reverse sequence ranking, the N current interest values are selected from the head of the current interest value sequence to obtain the N largest current interest values, at this time, the N largest current interest values are values with a specified percentage of the front in the current interest value sequence.

[0094] Step 104, selecting L negative samples of recommendation information corresponding to the N largest current interest values from the M recommendation information.

[0095] In the embodiment of the present application, since each current interest value corresponds to one of the M recommendation information, and the largest current interest value is also a current interest value, which also corresponds to one of the M recommendation information; therefore, the N largest current interest values correspond to N recommendation information in the M recommendation information, and the information processing device selects L recommendation information which is not converted by the first recommendation object sample from the N recommendation information, and the selected L recommendation information which is not converted by the first recommendation object sample is called L negative samples of recommendation information.

[0096] It should be noted that the negative sample of the recommended information refers to the recommended information which is not converted by the first recommended object sample and whose current interest value is greater than or equal to a specified interest value, wherein the specified interest value refers to the smallest one of N maximum current interest values; it is known that the specified interest value is negatively correlated with N; that is, the N current interest values greater than or equal to the specified interest value in the M current interest values are the N maximum current interest values. In addition, L is a positive integer, and L≤N; when L is equal to N, it indicates that the first recommended object sample is not converted to the N recommended information corresponding to the N maximum current interest values; and when L is less than N, it indicates that the first recommended object sample is not converted to part of the N recommended information. Here, not converted means not converted, converted is a target action performed, referred to as second conversion; that is, the second conversion is a specified action or multiple specified actions in the first conversion.

[0097] In step 104 of the embodiment of the present application, the information processing device selects L negative samples of recommended information corresponding to the N maximum current interest values from the M recommended information, including: the information processing device selects N recommended information corresponding to the N maximum current interest values from the M recommended information; then it can select P recommended information which is not converted by the first recommended object sample from the M recommended information; then it obtains L common recommended information between the N recommended information and the P recommended information; finally, it determines the L common recommended information as the L negative samples of recommended information. Here, after the information processing device selects N recommended information corresponding to the N maximum current interest values from the M recommended information, it can also select L recommended information which is not converted by the first recommended object sample from the N recommended information, that is, it obtains L negative samples of recommended information.

[0098] It should be noted that the common recommended information refers to the recommended information belonging to both the N recommended information and the P recommended information, wherein belonging to the N recommended information means that the corresponding current interest value is greater than or equal to the specified interest value, and belonging to the P recommended information means that the first recommended object sample does not convert the recommended information. Here, P is a positive integer, and L≤P.

[0099] Step 105, combining the first recommended object sample and the L negative samples of recommended information to construct a negative sample data set.

[0100] In the embodiment of the present application, the information processing device combines the first recommended object sample and each of the L negative samples of recommended information to construct a negative sample information; thus, the information processing device can obtain L negative sample information corresponding to the L negative samples of recommended information; here, the information processing device obtains the negative sample data set based on the L negative sample information.

[0101] It should be noted that the negative sample data set refers to a set of negative sample information used to train the first to-be-trained model; the negative sample data set at least includes L negative sample information, and of course, can also include negative sample training data corresponding to other recommended object samples, can also include training negative training data corresponding to the first recommended object sample at other action execution moments, can also include a combination of the two, etc., which is not limited by the embodiments of the present application.

[0102] Referring to Figure 4 , Figure 4 is a flowchart of an information processing method provided by the embodiments of the present application Figure 2 , wherein Figure 4 The execution subject of each step in the method is an information processing device; as Figure 4 shown, in the embodiments of the present application, step 105 can be implemented by steps 1051 to 1053; that is, the information processing device combines the first recommended object sample and L negative samples of recommendation information to construct a negative sample data set, including steps 1051 to 1053, which will be described below.

[0103] Step 1051, obtaining object association information of the first recommended object sample.

[0104] In the embodiments of the present application, the information processing device obtains the related information of the first recommended object sample, that is, the object association information; thus, the object association information refers to the information related to the first recommended object sample. Here, the information processing device can obtain the object association information of the first recommended object sample based on a specified dimension, which includes one or more of the following: conversion time dimension, action type dimension, frequency dimension, and conversion amount dimension; wherein the conversion time dimension refers to the to-be-counted time length of the conversion of the first recommended object sample to the recommendation information, such as the conversion of the recommendation information in the last 7 days, the last 14 days, the last 1 month, the last 3 months, or the last half year; the action type dimension refers to the type of the action specified by the first recommended object sample to the recommendation information, such as registration, download, resource exchange, application, etc.; the frequency dimension refers to the frequency level, such as greater than a first specified frequency (such as 9 times) being referred to as the most frequent, less than or equal to the first specified frequency and greater than a second specified frequency (such as 3 times) being referred to as the second frequent, and less than or equal to the second specified frequency being referred to as the low frequency; the conversion amount dimension can refer to the conversion quantity of the recommendation information, can also refer to the resource quantity consumed by the conversion of the recommendation information, and can also refer to a combination of the two, etc. At this time, the object association information refers to the information related to the first recommended object sample in the specified dimension.

[0105] Step 1052, combining the object association information with the L negative samples of recommendation information respectively to obtain L negative sample information.

[0106] It should be noted that when the information processing device combines the object association information and the L negative sample of recommendation information, since the negative sample of recommendation information represents the recommendation information in which the first recommended object sample does not convert, the conversion label of the first recommended object sample to the negative sample of recommendation information is the non-conversion label, indicating that the first recommended object sample does not convert to the negative sample of recommendation information; further, the information processing device combines the object association information, each of the L negative samples of recommendation information, and the corresponding non-conversion label into one negative sample information, also known as negative sample training data. It is known that the negative sample information includes the object association information, the negative sample of recommendation information, and the non-conversion label of the first recommended object sample to the negative sample of recommendation information.

[0107] Step 1053, determine the L negative sample information as a negative sample data set.

[0108] In the embodiment of the present application, the information processing device can directly determine the L negative sample information as the negative sample data set, and can also determine the L negative sample information and other non-conversion sample information as the negative sample data set, which is not limited in the embodiment of the present application.

[0109] Step 106, training the first to-be-trained model based on the negative sample data set to obtain an information recommendation model.

[0110] In the embodiment of the present application, after the information processing device obtains the negative sample data set, the negative sample data set can be directly used for training the first to-be-trained model, and the negative sample data set can also be combined with other sample data sets for training the first to-be-trained model, which is not limited in the embodiment of the present application.

[0111] It should be noted that the trained first to-be-trained model is the information recommendation model, and the information recommendation model is used to predict the target conversion rate of the to-be-recommended object to the to-be-recommended information. In addition, the loss function value calculated by the data processing device based on the negative sample data set is back propagated in the first to-be-trained model to adjust the model parameters in the first to-be-trained model; in addition, the training of the first to-be-trained model can be iteratively performed, and when the iterative training ends, the first to-be-trained model trained in the current iteration is the information recommendation model. Here, when the data processing device determines that the iterative training meets the first training end condition, the iterative training is determined to end; otherwise, the iterative training is continued. The first training end condition can be that the first accuracy index threshold is reached, or the first iteration number threshold is reached, or the first iteration time threshold is reached, or a combination of the above, etc., which is not limited in the embodiment of the present application. The to-be-recommended object refers to the recommended object to which information recommendation is performed, and the to-be-recommended information refers to the recommended information to which information recommendation is performed to the to-be-recommended object.

[0112] Referring to Figure 5 ,Figure 5 is a flowchart of an information processing method provided by an embodiment of the present application Figure 3 , wherein Figure 5 The execution subject of each step in the method is an information processing device; for example Figure 5 As shown in the figure, in the embodiment of the present application, step 106 can be implemented by steps 1061A to 1063A; that is, the information processing device trains the first to-be-trained model based on the negative sample data set to obtain the information recommendation model, which includes steps 1061A to 1063A, which will be described below.

[0113] Step 1061A, from the M recommendation information, select Q recommendation information positive samples of the first recommendation object sample that have been converted at the current action execution time.

[0114] It should be noted that the information processing device selects the recommendation information of the first recommendation object sample that has been converted at the current action execution time from the M recommendation information, thereby obtaining Q recommendation information positive samples; wherein each recommendation information positive sample represents the recommendation information of the first recommendation object sample that has been converted at the current action execution time; and the relationship between the total number of recommendation information M, the number of recommendation information positive samples Q and the number of unconverted recommendation information P can be described by formula (1), which is as follows.

[0115] M-P=Q(1);

[0116] Step 1062A, combine the first recommendation object sample and the Q recommendation information positive samples to construct a positive sample data set.

[0117] In the embodiment of the present application, the information processing device combines the first recommendation object sample with each of the Q recommendation information positive samples to construct a positive sample information; thereby, the information processing device can obtain Q positive sample information corresponding to the Q recommendation information negative samples; here, the information processing device obtains the positive sample data set based on the Q positive sample information.

[0118] It should be noted that the positive sample data set refers to a positive sample set used to train the first to-be-trained model; the positive sample data set at least includes Q positive sample information, of course, it can also include other recommendation object sample corresponding positive sample training data, it can also include the first recommendation object sample corresponding positive sample training data at other action execution time, it can also include the combination of the two, etc., which is not limited by the embodiment of the present application.

[0119] In the embodiment of the present application, the information processing device combines the first recommended object sample and the Q recommended information positive samples to construct the positive sample dataset, including: the information processing device combines the object association information of the first recommended object sample with the Q recommended information positive samples respectively to obtain Q positive sample information, the positive sample information including the object association information, the recommended information positive sample, and the converted label of the first recommended object sample for the recommended information positive sample; and the information processing device determines the Q positive sample information as the positive sample dataset.

[0120] It should be noted that, when the information processing device combines the object association information with the Q recommended information positive samples, since the recommended information positive sample represents the converted recommended information of the first recommended object sample, the converted label of the first recommended object sample for the recommended information positive sample is the converted label, indicating that the first recommended object sample has converted the recommended information positive sample; and then, the information processing device combines the object association information, each of the Q recommended information positive samples, and the corresponding converted label as one positive sample information. It is known that the positive sample information includes the object association information, the recommended information positive sample, and the converted label of the first recommended object sample for the recommended information positive sample. Here, the information processing device can directly determine the L positive sample information as the positive sample dataset, and can also determine the L positive sample information and other converted sample information as the positive sample dataset, and the embodiment of the present application does not limit this.

[0121] Step 1063A, training the first to-be-trained model in combination with the positive sample dataset and the negative sample dataset to obtain the information recommendation model.

[0122] In the embodiment of the present application, the information processing device takes the positive sample dataset and the negative sample dataset as the training data of the first to-be-trained model, to train the first to-be-trained model in combination with the positive sample dataset and the negative sample dataset.

[0123] Referring to Figure 6 , Figure 6 is the flowchart of the information processing method provided in the embodiment of the present application Figure 4 , wherein Figure 6 The execution subject of each step in the embodiment of the present application is the information processing device; as shown in Figure 6 In the embodiment of the present application, step 106 can be implemented by steps 1061B to 1063B; that is, the information processing device trains the first to-be-trained model based on the negative sample dataset to obtain the information recommendation model, including steps 1061B to 1063B, which will be described below.

[0124] In the embodiment of the present application, the information processing device performs the following processing (i.e., step 1061B) for each negative sample information in the negative sample dataset.

[0125] Step 1061B, predicting the object association information and the recommendation information negative sample in the negative sample information by using the first to-be-trained model to obtain a first estimated conversion rate of the first recommended object sample to the recommendation information negative sample.

[0126] It should be noted that the information processing device first extracts features of the object association information and the recommendation information negative sample in the negative sample information by using the first to-be-trained model to obtain first object information features corresponding to the object association information and first recommendation information features corresponding to the recommendation information negative sample; and then predicts the conversion rate of the first recommended object sample to the recommendation information negative sample by combining the first object information features and the first recommendation information features, and the predicted conversion rate is referred to as a first estimated conversion rate.

[0127] Step 1062B, calculating a first loss function value based on a difference between the first estimated conversion rate and the non-conversion label in the negative sample information.

[0128] In the embodiments of the present application, the information processing device calculates the loss function value based on the difference between the first estimated conversion rate and the non-conversion label in the negative sample information, and the calculated loss function value is referred to as a first loss function.

[0129] In the embodiments of the present application, when the information processing device trains the first to-be-trained model by combining the positive sample data set and the negative sample data set, the information processing device calculates the first loss function value based on the difference between the first estimated conversion rate and the non-conversion label in the negative sample information, including: the information processing device combines the difference between the first estimated conversion rate and the non-conversion label in the negative sample information, and the difference between the second estimated conversion rate and the conversion label in the positive sample information, to calculate the first loss function value. Therefore, before the information processing device calculates the first loss function value based on the difference between the first estimated conversion rate and the non-conversion label in the negative sample information, the information processing method further includes: the information processing device performs the following processing for each positive sample information in the positive sample data set: predicting the object association information and the recommendation information positive sample in the positive sample information by using the first to-be-trained model to obtain a second estimated conversion rate of the first recommended object sample to the recommendation information positive sample.

[0130] It should be noted that the information processing device first extracts features of the object association information and the recommendation information negative sample in the negative sample information by using the first to-be-trained model to obtain first object information features corresponding to the object association information and first recommendation information features corresponding to the recommendation information negative sample; and then predicts the conversion rate of the first recommended object sample to the recommendation information negative sample by combining the first object information features and the first recommendation information features, and the predicted conversion rate is referred to as a first estimated conversion rate.

[0131] Step 1063B, training the first to-be-trained model based on the first loss function value to obtain an information recommendation model.

[0132] It should be noted that the first loss function value represents the prediction effect of the first to-be-trained model, so that the information processing device performs back propagation in the first to-be-trained model based on the first loss function value to adjust the model parameters of the first to-be-trained model, and the first to-be-trained model after completing the model parameter training is the information recommendation model.

[0133] It can be understood that when the negative sample data set is constructed based on the first recommendation object sample, M current interest values of M recommendation information are first predicted based on the historical action data sequence of the first recommendation object sample, and then L recommendation information negative samples are determined from the M recommendation information based on the N maximum current interest values, so as to combine the first recommendation object sample and the L recommendation information negative samples to construct the negative sample data set; in this way, the recommendation information negative samples in the constructed negative sample data set are the recommendation information negative samples that the first recommendation object sample is interested in, the quality of the negative sample data set is improved, and then when the model is trained based on the negative sample data set, the accuracy of the model training can be improved.

[0134] Referring to Figure 7 , Figure 7 is a flowchart of an information processing method provided by an embodiment of the present application Figure 5 , wherein Figure 7 The execution subject of each step in Figure 7 As shown in the figure, in the embodiment of the present application, after step 106, steps 107 to 111 are further included; that is, after the information processing device trains the first to-be-trained model based on the negative sample data set to obtain the information recommendation model, the information processing method further includes steps 107 to 111, which will be described below.

[0135] Step 107, in response to an information recommendation request, obtaining a to-be-recommended object and to-be-recommended information.

[0136] In the embodiment of the present application, when it is requested to determine whether to recommend to-be-recommended information to a to-be-recommended object, the information processing device also receives an information recommendation request; at this time, the information processing device can obtain the to-be-recommended object and the to-be-recommended information in response to the information recommendation request.

[0137] Step 108, performing feature extraction on the to-be-recommended object to obtain to-be-recommended object features.

[0138] It should be noted that the information processing device can perform feature extraction on the to-be-recommended object, that is, perform feature extraction on the related data of the to-be-recommended object, and the extracted features are referred to as to-be-recommended object features. The data types of the related data of the to-be-recommended object are consistent with the data types of the object relationship information, which will not be described again in the embodiment of the present application.

[0139] Step 109, feature extraction is performed on the to-be-recommended information to obtain to-be-recommended information features.

[0140] It should be noted that the information processing device performs feature extraction on the to-be-recommended information, which can be feature extraction on related data of the to-be-recommended information, and the extracted features are referred to as to-be-recommended information features. The data types of the related data of the to-be-recommended information are consistent with the data types of the positive samples of the recommended information (or the negative samples of the recommended information), and include one or more of the following: information name, text description, voice description, video description, image description, application effect description, and the like.

[0141] Step 110, the information processing device uses the information recommendation model to predict the to-be-recommended object features and the to-be-recommended information features to obtain a target conversion rate.

[0142] It should be noted that the process of the information processing device using the information recommendation model to predict the to-be-recommended object features and the to-be-recommended information features is similar to the process of using the first training model to predict the negative samples of the recommended information (or the positive samples of the recommended information) by the first recommended object sample, and the embodiments of the present application will not be repeated here. Here, the prediction result of the information recommendation model obtained by the information processing device on the to-be-recommended object features and the to-be-recommended information features is the target conversion rate, and the target conversion rate represents the possibility of the to-be-recommended object converting the to-be-recommended information.

[0143] Step 111, determining a target recommendation result based on the target conversion rate.

[0144] In the embodiments of the present application, the information processing device compares the target conversion rate with a conversion rate threshold, and determines the target recommendation result of whether to recommend the to-be-recommended information to the to-be-recommended object based on the comparison result.

[0145] It should be noted that the target recommendation result represents whether to recommend the to-be-recommended information to the to-be-recommended object. The conversion rate threshold represents the minimum conversion rate of recommending the recommended information to the recommended object, such as 0.85; so when the comparison result indicates that the target conversion rate is greater than the conversion rate threshold, the target recommendation result indicates that the to-be-recommended information is recommended to the to-be-recommended object; and when the comparison result indicates that the target conversion rate is less than or equal to the conversion rate threshold, the target recommendation result indicates that the to-be-recommended information is not recommended to the to-be-recommended object.

[0146] In the embodiments of the present application, the information processing device determines the target recommendation result based on the target conversion rate, and can also sort the target conversion rates corresponding to the plurality of to-be-recommended information by the information processing device, to select the to-be-recommended information corresponding to the maximum target conversion rate of the specified recommendation quantity, and obtain the target recommendation result.

[0147] In the embodiment of the present application, step 102 can be implemented by a neural network model, which is an interest prediction model used to predict the degree of interest of the recommended object in the next moment of the current action execution moment for the recommended information. That is, the information processing device predicts the M current interest values of the first recommended object sample for the M recommended information based on the historical action data sequence, which is achieved by the interest prediction model. The interest prediction model is obtained by training the following steps: the information processing device first obtains the recommended information sample sequence of the action performed by the second recommended object sample and the recommended information label of the action performed by the second recommended object sample in the next action execution moment; the second to-be-trained model is used to predict the recommended information sample sequence to obtain the estimated recommended information of the action performed by the second recommended object sample in the next action execution moment; based on the difference between the estimated recommended information and the recommended information label, a second loss function value is calculated; and the second to-be-trained model is trained based on the second loss function value to obtain the interest prediction model.

[0148] It should be noted that the second to-be-trained model refers to a neural network model to be trained for predicting the interest value of the recommended object in the next moment of the current action execution moment for the recommended information, which can be an original neural network model constructed, or a pre-trained neural network model, etc., which is not limited in the embodiment of the present application. The next action execution moment refers to the next moment of the last action execution moment corresponding to the recommended information sample sequence. Here, the recommended information sample sequence is used to obtain the data sequence corresponding to the historical action data sequence, and then the recommended information of the action performed by the second recommended object sample in the next action execution moment is predicted based on the obtained data sequence, which is called estimated recommended information, and then the loss function value of the second to-be-trained model is calculated based on the difference between the estimated recommended information and the recommended information label, that is, the second loss function value is obtained. The interest prediction model refers to the second to-be-trained model after training.

[0149] It should also be noted that the training of the second to-be-trained model can be iterative, and when the iterative training ends, the second to-be-trained model trained in the current iteration is the interest prediction model. Here, when the data processing device determines that the iterative training meets the second training end condition, the iterative training is determined to end; otherwise, the iterative training is continued. The second training end condition can be that the second accuracy index threshold is reached, or the second iteration number threshold is reached, or the second iteration time threshold is reached, or a combination of the above, etc., which is not limited in the embodiment of the present application.

[0150] In the embodiment of the present application, the model parameter amount of the interest prediction model is greater than the model parameter amount of the information recommendation model.

[0151] It can be understood that the negative sample data set is constructed based on the interest prediction model to train the information recommendation model with less model parameter quantity; and the accuracy of the information recommendation model can be improved while ensuring the prediction efficiency of the information recommendation model.

[0152] In the embodiment of the application, the information processing device obtains a recommendation information sample sequence of an action performed by a second recommendation object sample and a recommendation information label of an action performed by the second recommendation object sample at a next action execution moment, including: an information acquisition device first obtains a recommendation information initial sequence of an action performed by the second recommendation object sample and a recommendation information quantity K of the recommendation information initial sequence; then, starting from a lowest sequence quantity, iterates k, and performs the following processing by iterating k: starting from a first recommendation information in the recommendation information initial sequence, sequentially selects k recommendation information; and constructs the selected k recommendation information into a recommendation information sample sequence; and determines a k+1 recommendation information in the recommendation information initial sequence as the recommendation information label of the action performed by the second recommendation object sample at the next action execution moment.

[0153] It should be noted that K is an integer greater than 1, and k≤K-1. The lowest sequence quantity refers to the number of recommendation information in the recommendation information sample sequence, such as 1, 3, etc., so that the initial value of k is the lowest sequence quantity.

[0154] Exemplarily, when the recommendation information initial sequence includes recommendation information 1 to recommendation information 10, if the maximum sequence quantity is 1, the sequentially obtained recommendation information sample sequences are {recommendation information 1} (the corresponding recommendation information label is recommendation information 2), {recommendation information 1, recommendation information 2} (the corresponding recommendation information label is recommendation information 3), {recommendation information 1, recommendation information 2, recommendation information 3} (the corresponding recommendation information label is recommendation information 4)……, and the like.

[0155] Next, an exemplary application of the embodiment of the application in an actual application scenario will be described. The exemplary application describes the process of selecting negative samples for model training in a game prop recommendation scenario; it is known that the information processing method provided by the embodiment of the application is applicable to various information recommendation scenarios, such as video information recommendation, news recommendation, and item recommendation, etc. Here, the recommendation of game props is taken as an example for description. Among them, the game prop recommendation refers to the processing of recommending corresponding game props for different game accounts.

[0156] Exemplarily, referring to Figure 8 , Figure 8 is an exemplary schematic diagram of training a business model provided by the embodiment of the application; as Figure 8As shown, the generative model 8-2 (referred to as a second to-be-trained model) is trained by the data set 8-1 (referred to as a recommended information sample sequence and a recommended information label). After the training of the generative model 8-2 is completed, the trained generative model 8-2 (referred to as an interest prediction model) is used to predict the preference probability (referred to as a current interest value) of each game prop corresponding to the game account at the next moment, and then a game prop sequence 8-3 arranged in a reverse order of the preference probability is obtained; and the first N to-be-recommended game props (referred to as recommended information) are selected from the game prop sequence 8-3, and a negative sample data set 8-4 is constructed based on the selected N to-be-recommended game props, so as to construct a data set 8-6 for training the business model 8-5 (referred to as a first to-be-trained model).

[0157] It should be noted that the data set 8-1 includes the input data of the generative model 8-2, denoted as D GPT U ] and U represents the total number of game accounts, and s u (u∈[1,U]) represents the historical action data sequence of the u-th game account on the time sequence for the game prop, wherein the action for the game prop is, for example, game prop use, game prop click, game prop acquisition through resource exchange, game prop sharing, etc.

[0158] The generative model 8-2 is denoted as F GPT :[s u ]→Y u , Y u represents the game prop sequence arranged in a reverse order of the preference probability of the u-th game account at the next moment, that is, in the game prop sequence, the game prop with a larger preference probability is more interested in the game account.

[0159] The data set 8-6 is denoted as D & ' ( * , wherein f2 represents the feature corresponding to the v-th game account, for example, the virtual prop acquired through resource exchange in the last seven days and the amount of resource exchange in the last seven days of the v-th game account; item0 represents the i-th game prop; represents the conversion label of whether the v-th game account acquires the i-th game prop through resource exchange, for example, 0 or 1. When , d2 is a positive sample; and when , d2 is a negative sample, and at this time, item0 is the game prop selected from the game prop sequence and ranked in the first N. ​​

[0160] Business Model 8-5, denoted as F & ' ( :[f2,item0]→y′ v,i , y′ v,i It represents the probability that the vth game account obtains the ith game item through resource exchange; thus, based on the difference between this probability and the conversion label, the business model 8-5 can be trained.

[0161] The following describes the training process of the generative model.

[0162] See also Figure 9 , Figure 9 is a schematic diagram of an exemplary generative model training provided in an embodiment of the present application; Figure 9 As shown, the embedding layer (Embedding Lookup) 9-1 is used to interact with the prop sequence 9-2 (recorded as item _1 ,item _2 ,item _3 ,…,item _J；1 , J-1 represents the sequence length, item _J (j∈[1, J-1]) represents the game props of the action performed by the game account at the jth moment) and then uses the generative model to be trained 9-3 to process the embedding result to predict the interactive prop sequence 9-4 (recorded as item′ _2 ,item′ _3 ,…,item′ _J-1 ,item′ _J , item _J (j∈[2, J]) represents the game props of the game account at the jth moment). Among them, item′ _2 By item _1 Predicted item _3 By item _1 ,item _2 Predicted; ...; item ' _J By item _1 ,item _2 ,item _3 ,…,item _J-1 Predicted.

[0163] Here, the cross entropy loss function is used to calculate the loss function value Loss of the generative model, as shown in formula (2).

[0164]

[0165] It should be noted that the cross-entropy loss is used to compare the difference between the predicted result item' _J and the label item _J , and the smaller the value is, the closer the two are, and vice versa.

[0166] Continuing to refer to Figure 9 , the generated model 9-3 to be trained is trained based on the loss function value 9-5 (referred to as a second loss function value) corresponding to the interactive prop sequence 9-2 and the interactive prop sequence 9-4.

[0167] Referring to Figure 10 , Figure 10 is an exemplary game prop recommendation schematic diagram provided by an embodiment of the present application; as shown in Figure 10 , a game account corresponding game prop sequence 10-2 arranged in a descending order based on preference probability is predicted by using the generated model 10-1, the first N game props not obtained by using the resource exchange mode are selected from the game prop sequence 10-2 to construct a negative sample data set 10-31 with the game account; and the game prop obtained by using the resource exchange mode is selected to construct a positive sample data set 10-32 with the game account; the business model 10-4 is trained in combination with the negative sample data set 10-31 and the positive sample data set 10-32. When the training of the business model 10-4 is completed, in response to a game prop recommendation request 10-5, the probability of recommending a game prop 10-62 to a game account 10-61 is predicted to determine a recommendation result 10-7 of whether to recommend the game prop 10-62 to the game account 10-61.

[0168] Here, the data accumulated for half a year in the game prop recommendation scenario is used as a training set, and the data for the next 7 days is used as a test set. In the conversion rate estimation task of obtaining game props by using the resource exchange mode, the prediction effects of the business model (referred to as the model of the present application) provided by an embodiment of the present application and the baseline model are compared, wherein the baseline model is trained by using a random negative sample selection mode. As shown in Table 1, the model effect is measured by using an effect measurement index (for example, Area Under Curve, AUC) on the test set.

[0169] Table 1

[0170] Effectiveness metrics Baseline model 0.9513 Applicant model 0.9556

[0171] As can be seen from Table 1, the model of the present application improves the effect measurement index by 0.45% compared with the baseline model.

[0172] It can be understood that the embodiments of the present application apply the autoregressive method to model the historical action data of the game account, generate a list of game props that the game account is likely to obtain in the next moment by resource exchange and a corresponding probability distribution, and use the probability as additional information for selecting game prop negative samples based on the probability to select game prop negative samples for training the business model. Since the probability in the game prop list represents the preference degree of the game account for different game props, the game prop with higher ranking has a larger probability, indicating that the game account has greater interest in the game prop. Therefore, when the negative sample is constructed based on the probability to select game props greater than the specified percentage of the ranking, the distance between the negative sample and the positive sample can be reduced, and the quality of the negative sample can be improved. In this way, the prediction quality of the business model can be improved, and the performance of the business model can be improved.

[0173] The following continues to illustrate an exemplary structure of the information processing apparatus 455 provided by the embodiments of the present application implemented as a software module. In some embodiments, as shown in Figure 2 The software modules stored in the information processing apparatus 455 of the memory 450 can include:

[0174] The data collection module 4551 is configured to obtain a sequence of historical action data performed by the first recommended object sample on the recommendation information.

[0175] The interest prediction module 4552 is configured to predict M current interest values of the first recommended object sample on M recommendation information based on the sequence of historical action data, M > 1, and M is an integer.

[0176] The sample construction module 4553 is configured to select N maximum current interest values from the M current interest values, N being a positive integer.

[0177] The sample construction module 4553 is further configured to select L negative sample of recommendation information corresponding to the N maximum current interest values from the M recommendation information, L being a positive integer, L ≤ N.

[0178] The sample construction module 4553 is further configured to construct a negative sample data set in combination with the first recommended object sample and the L negative sample of recommendation information.

[0179] The model training module 4554 is configured to train the first to-be-trained model based on the negative sample data set to obtain an information recommendation model, the information recommendation model being configured to predict a conversion rate of a to-be-recommended object on to-be-recommended information.

[0180] In the embodiment of the present application, the data collection module 4551 is further configured to, for each historical time in the sequence of historical times, perform the following processing: acquire historical action data of the first recommended object sample on the recommendation information at the historical time; and obtain the sequence of historical action data corresponding to the sequence of historical times from the historical action data corresponding to each historical time.

[0181] In the embodiment of the present application, the interest prediction module 4552 is further configured to predict M current interest values of the first recommended object sample on M recommendation information based on data features corresponding to each historical action data in the sequence of historical action data.

[0182] In the embodiment of the present application, the sample construction module 4553 is further configured to: select N recommendation information corresponding to N maximum current interest values from M recommendation information; select P recommendation information that has not been converted by the first recommended object sample from M recommendation information, P being a positive integer, the not being converted referring to not being converted; acquire L common recommendation information between N recommendation information and P recommendation information; and determine L common recommendation information as L recommendation information negative samples.

[0183] In the embodiment of the present application, the sample construction module 4553 is further configured to: acquire object association information of the first recommended object sample; combine the object association information with L recommendation information negative samples respectively to obtain L negative sample information, the negative sample information including the object association information, the recommendation information negative sample, and a not converted label of the first recommended object sample on the recommendation information negative sample; and determine L negative sample information as the negative sample data set.

[0184] In the embodiment of the present application, the model training module 4554 is further configured to: select Q recommendation information positive samples that have been converted by the first recommended object sample from M recommendation information, Q being a positive integer; construct a positive sample data set in combination with the first recommended object sample and Q recommendation information positive samples; and train the first to-be-trained model in combination with the positive sample data set and the negative sample data set to obtain the information recommendation model.

[0185] In the embodiment of the present application, the model training module 4554 is further configured to: combine the object association information of the first recommended object sample with Q recommendation information positive samples respectively to obtain Q positive sample information, the positive sample information including the object association information, the recommendation information positive sample, and a converted label of the first recommended object sample on the recommendation information positive sample; and determine Q positive sample information as the positive sample data set.

[0186] In the embodiment of the present application, the model training module 4554 is further configured to, for each negative sample information in the negative sample data set, perform the following processing: using the first to-be-trained model to predict the object association information and the recommendation information negative sample in the negative sample information, to obtain a first estimated conversion rate of the first recommended object sample for the recommendation information negative sample; calculating a first loss function value based on the difference between the first estimated conversion rate and the non-conversion label in the negative sample information; training the first to-be-trained model based on the first loss function value to obtain the information recommendation model.

[0187] In the embodiment of the present application, the model training module 4554 is further configured to, for each positive sample information in the positive sample data set, perform the following processing: using the first to-be-trained model to predict the object association information and the recommendation information positive sample in the positive sample information, to obtain a second estimated conversion rate of the first recommended object sample for the recommendation information positive sample.

[0188] In the embodiment of the present application, the model training module 4554 is further configured to calculate the first loss function value in combination with the difference between the first estimated conversion rate and the non-conversion label in the negative sample information, and the difference between the second estimated conversion rate and the conversion label in the positive sample information.

[0189] In the embodiment of the present application, the information processing apparatus 455 further comprises an information recommendation module 4555, configured to, in response to an information recommendation request, obtain the to-be-recommended object and the to-be-recommended information; perform feature extraction on the to-be-recommended object to obtain to-be-recommended object features; perform feature extraction on the to-be-recommended information to obtain to-be-recommended information features; use the information recommendation model to predict the to-be-recommended object features and the to-be-recommended information features, to obtain a target conversion rate; and determine a target recommendation result based on the target conversion rate, the target recommendation result indicating whether to recommend the to-be-recommended information to the to-be-recommended object.

[0190] In the embodiment of the present application, the prediction of the M current interest values of the first recommended object sample for the M recommended information based on the historical action data sequence is achieved by an interest prediction model; the model training module 4554 is further configured to obtain a recommended information sample sequence of an action performed by a second recommended object sample and a recommended information label of an action performed by the second recommended object sample at a next action execution time, wherein the next action execution time refers to a next time of a last action execution time corresponding to the recommended information sample sequence; a second to-be-trained model is used to predict the recommended information sample sequence to obtain an estimated recommended information of the action performed by the second recommended object sample at the next action execution time; a second loss function value is calculated based on a difference between the estimated recommended information and the recommended information label; and the second to-be-trained model is trained based on the second loss function value to obtain the interest prediction model.

[0191] In the embodiment of the present application, the model training module 4554 is further configured to obtain a recommended information initial sequence of an action performed by the second recommended object sample and a recommended information quantity K of the recommended information initial sequence, wherein K is an integer greater than 1; starting from a lowest sequence quantity, iterate k, and perform the following processing by iteration k, k≤K-1: starting from a first recommended information in the recommended information initial sequence, sequentially select k recommended information; construct the selected k recommended information into the recommended information sample sequence; and determine a k+1 recommended information in the recommended information initial sequence as the recommended information label of the action performed by the second recommended object sample at the next action execution time.

[0192] In the embodiment of the present application, a model parameter quantity of the interest prediction model is greater than a model parameter quantity of the information recommendation model.

[0193] An embodiment of the present application provides a computer program product, which includes computer executable instructions or computer programs stored in a computer readable storage medium. A processor of an information processing device reads the computer executable instructions or computer programs from the computer readable storage medium, and the processor executes the computer executable instructions or computer programs, so that the information processing device executes the information processing method provided in the embodiments of the present application.

[0194] An embodiment of the present application provides a computer readable storage medium, which stores computer executable instructions or computer programs. When the computer executable instructions or computer programs are executed by a processor, the processor will execute the information processing method provided in the embodiments of the present application, for example, the information processing method shown in the above embodiment. Figure 3

[0195] ​In some embodiments, the computer-readable storage media can be a memory such as FRAM, ROM, flash memory, magnetic surface memory, optical disc, or CD-ROM, etc.; or various devices including one or any combination of the above memories.

[0196] In some embodiments, the computer-executable instructions can take the form of programs, software, software modules, scripts, or code, written in any form of programming language, including compiled or interpreted languages, or declarative or procedural languages, and can be deployed in any form, including as a stand-alone program or as a module, component, subroutine, or other unit suitable for use in a computing environment.

[0197] By way of example, a computer-executable instruction can, but need not, correspond to a file in a file system. A computer-executable instruction can be stored in a portion of a file that holds other programs or data, e.g., one or more scripts stored in a web page; in a single file dedicated to the program in question; or in multiple coordinated files, e.g., files that store one or more modules, sub programs, or code portions.

[0198] By way of example, a computer-executable instruction can be deployed to be executed on one electronic device (in which case the one electronic device is the information processing device), or on multiple electronic devices located at one site (in which case the multiple electronic devices located at one site are the information processing device), or on multiple electronic devices distributed across multiple sites and interconnected through a communication network (in which case the multiple electronic devices distributed across multiple sites and interconnected through a communication network are the information processing device).

[0199] It can be understood that, in the embodiments of the present application, related data such as historical action data and conversion labels are involved. When the embodiments of the present application are applied to specific products or technologies, the user's permission or consent needs to be obtained, and the collection, use, and processing of related data need to comply with relevant laws, regulations, and standards of countries and regions. In the present application, the collection and processing of related data should strictly comply with the requirements of relevant national laws and regulations, obtain the informed consent or separate consent of the principal of personal information, and within the scope of authorization of laws and regulations and the principal of personal information, carry out subsequent data use and processing behavior. In the present application, the data crawling technical solution involved is implemented. When the above embodiments of the present application are applied to specific products or technologies, the related data collection, use, and processing process should comply with the requirements of national laws and regulations, meet the principles of legality, legitimacy, and necessity, and do not involve obtaining data types prohibited or restricted by laws and regulations, and will not hinder the normal operation of the target website.

[0200] To sum up, in the embodiment of the present application, when constructing the negative sample dataset based on the first recommended object sample, first, the M current interest values of the M recommended information are predicted based on the historical action data sequence of the first recommended object sample, and then L recommended information negative samples are determined from the M recommended information based on the N maximum current interest values, so as to construct the negative sample dataset by combining the first recommended object sample and the L recommended information negative samples. In this way, the recommended information negative samples in the constructed negative sample dataset are the recommended information negative samples that the first recommended object sample is interested in, which improves the quality of the negative sample dataset, and further improves the accuracy of model training when the model is trained based on the negative sample dataset.

[0201] The above is only an embodiment of the present application, and is not used to limit the protection scope of the present application. Any modification, equivalent replacement and improvement made within the spirit and scope of the present application are included in the protection scope of the present application.

Claims

1. An information processing method, characterized in that: The method comprises: Obtaining a historical action data sequence of the first recommended object sample performed on the recommended information; Based on the historical action data sequence, predict M current interest values ​​of the first recommended object sample for the M recommendation information, where M>1 and M is an integer; Selecting N maximum current interest values ​​from the M current interest values, where N is a positive integer; Selecting L negative samples of recommended information corresponding to the N maximum current interest values ​​from the M recommended information, where L is a positive integer and L≤N; Combining the first recommended object sample and the L negative samples of the recommendation information to construct a negative sample data set; A first to-be-trained model is trained based on the negative sample data set to obtain an information recommendation model, wherein the information recommendation model is used to predict a conversion rate of the to-be-recommended object with respect to the to-be-recommended information.

2. The method according to claim 1, characterized in that The step of obtaining a historical action data sequence of the first recommended object sample in response to the recommended information includes: For each historical moment in the historical moment sequence, perform the following processing: Obtaining historical action data of the first recommended object sample performed on the recommendation information at the historical moment; Obtaining the historical action data sequence corresponding to the historical moment sequence from the historical action data corresponding to each historical moment; The predicting, based on the historical action data sequence, M current interest values ​​of the first recommended object sample for the M pieces of recommended information includes: Based on the data features corresponding to each of the historical action data in the historical action data sequence, the M current interest values ​​of the first recommended object sample for the M recommendation information are predicted.

3. The method according to claim 1, characterized in that The step of selecting L negative samples of recommendation information corresponding to the N maximum current interest values ​​from the M pieces of recommendation information includes: Selecting, from the M pieces of recommendation information, N pieces of recommendation information corresponding to the N maximum current interest values; Selecting P pieces of recommendation information of the first recommendation object sample that have not been converted from the M pieces of recommendation information, where P is a positive integer and the term "not converted" means that no conversion has occurred; Obtaining L common recommendation information between the N recommendation information and the P recommendation information; The L public recommendation information are determined as L negative samples of the recommendation information.

4. The method according to claim 1, wherein The step of combining the first recommended object sample and the L negative samples of the recommendation information to construct a negative sample dataset includes: Obtaining object association information of the first recommended object sample; Combining the object association information with the L negative samples of the recommendation information to obtain L negative sample information, wherein the negative sample information includes the object association information, the negative sample of the recommendation information, and the unconverted label of the first recommended object sample for the negative sample of the recommendation information; The L pieces of negative sample information are determined as the negative sample data set.

5. The method according to claim 1, wherein The step of training the first model to be trained based on the negative sample data set to obtain an information recommendation model includes: From the M pieces of recommendation information, select Q positive recommendation information samples that have been converted from the first recommendation object sample, where Q is a positive integer; Combining the first recommended object sample and the Q positive samples of the recommendation information to construct a positive sample data set; The first to-be-trained model is trained in combination with the positive sample data set and the negative sample data set to obtain the information recommendation model.

6. The method according to claim 5, characterized in that The step of combining the first recommended object sample and the Q positive samples of the recommendation information to construct a positive sample dataset includes: Combining the object association information of the first recommended object sample with the Q positive recommendation information samples to obtain Q positive sample information, wherein the positive sample information includes the object association information, the positive recommendation information sample, and the converted label of the first recommended object sample for the positive recommendation information sample; The Q pieces of positive sample information are determined as the positive sample data set.

7. The method according to any one of claims 1 to 6, characterized in that The step of training the first model to be trained based on the negative sample data set to obtain an information recommendation model includes: For each negative sample in the negative sample dataset, perform the following processing: Using the first to-be-trained model to predict the object association information in the negative sample information and the negative sample of the recommendation information, to obtain a first estimated conversion rate of the first recommended object sample with respect to the negative sample of the recommendation information; Calculating a first loss function value based on a difference between the first estimated conversion rate and an unconverted label in the negative sample information; The first to-be-trained model is trained based on the first loss function value to obtain the information recommendation model.

8. The method according to claim 7, characterized in that Before calculating the first loss function value based on the difference between the first estimated conversion rate and the unconverted label in the negative sample information, the method further includes: For each positive sample in the positive sample dataset, perform the following processing: Using the first to-be-trained model to predict the object association information and the positive sample of recommendation information in the positive sample information, to obtain a second estimated conversion rate of the first recommended object sample with respect to the positive sample of recommendation information; The calculating a first loss function value based on a difference between the first estimated conversion rate and the unconverted label in the negative sample information includes: The first loss function value is calculated by combining the difference between the first estimated conversion rate and the unconverted labels in the negative sample information, and the difference between the second estimated conversion rate and the converted labels in the positive sample information.

9. The method according to any one of claims 1 to 6, characterized in that After training the first to-be-trained model based on the negative sample data set to obtain the information recommendation model, the method further includes: In response to the information recommendation request, obtaining the object to be recommended and the information to be recommended; Extracting features of the object to be recommended to obtain features of the object to be recommended; Extracting features of the information to be recommended to obtain features of the information to be recommended; Using the information recommendation model to predict the characteristics of the object to be recommended and the characteristics of the information to be recommended to obtain a target conversion rate; Based on the target conversion rate, a target recommendation result is determined, where the target recommendation result indicates whether the information to be recommended is recommended to the object to be recommended.

10. The method according to any one of claims 1 to 6, characterized in that The predicting, based on the historical action data sequence, M current interest values ​​of the first recommended object sample for the M pieces of recommended information is achieved by an interest prediction model; The interest prediction model is trained by the following steps: Obtaining a recommendation information sample sequence of an action performed by a second recommended object sample and a recommendation information label of an action performed by the second recommended object sample at a next action execution time, where the next action execution time is a time immediately following the last action execution time corresponding to the recommendation information sample sequence; Using the second to-be-trained model to predict the recommendation information sample sequence, to obtain estimated recommendation information of the action to be performed by the second recommended object sample at the next action execution time; Calculating a second loss function value based on a difference between the estimated recommendation information and the recommendation information label; The second model to be trained is trained based on the second loss function value to obtain the interest prediction model.

11. The method according to claim 10, characterized in that The obtaining of a sequence of recommendation information samples of an action performed by the second recommended object sample and a recommendation information label of an action performed by the second recommended object sample at the next action execution moment includes: Obtaining an initial sequence of recommendation information for the action performed by the second recommended object sample and the number K of recommendation information in the initial sequence, where K is an integer greater than 1; Starting from the lowest number of sequences, iterate k and perform the following process for each iteration, k≤K-1: Starting from the first recommendation information in the initial sequence of recommendation information, select k pieces of recommendation information in sequence; constructing the selected k pieces of recommendation information into the recommendation information sample sequence; The k+1th recommendation information in the initial sequence of recommendation information is determined as the recommendation information label of the action executed by the second recommendation object sample at the next action execution time.

12. The method according to claim 10, characterized in that The number of model parameters of the interest prediction model is greater than the number of model parameters of the information recommendation model.

13. An information processing device, characterized in that: The information processing device includes: A data collection module is used to obtain a historical action data sequence performed by the first recommended object sample on the recommended information; an interest prediction module, configured to predict, based on the historical action data sequence, M current interest values ​​of the first recommended object sample for the M pieces of recommendation information, where M>1 and M is an integer; A sample construction module, configured to select N maximum current interest values ​​from the M current interest values, where N is a positive integer; The sample construction module is further configured to select L negative samples of recommendation information corresponding to the N maximum current interest values ​​from the M recommendation information, where L is a positive integer and L≤N; The sample construction module is further configured to combine the first recommended object sample and the L negative samples of the recommendation information to construct a negative sample dataset; The model training module is used to train a first model to be trained based on the negative sample data set to obtain an information recommendation model, wherein the information recommendation model is used to predict the conversion rate of the recommended object to the recommended information.

14. An electronic device for information processing, characterized in that: The electronic device comprises: a memory for storing computer-executable instructions or computer programs; The processor is configured to implement the information processing method according to any one of claims 1 to 12 when executing the computer-executable instructions or computer programs stored in the memory.

15. A computer-readable storage medium storing computer-executable instructions or a computer program, characterized in that: When the computer-executable instructions or computer programs are executed by a processor, the information processing method according to any one of claims 1 to 12 is implemented.

16. A computer program product comprising computer executable instructions or a computer program, characterized in that: When the computer-executable instructions or computer programs are executed by a processor, the information processing method according to any one of claims 1 to 12 is implemented.

Images