An information recommendation method and device, electronic equipment and storage medium

Abstract

Embodiments of the present application provide an information recommendation method and device, electronic equipment and storage medium, belonging to the technical field of artificial intelligence. The method comprises: obtaining an object behavior sequence of a target object; the object behavior sequence comprises behavior information of the target object; inputting the object behavior sequence into an embedding layer model to obtain object encoding information; the embedding layer model is used for modeling a time interval at which the behavior information occurs; inputting the object encoding information into an encoder model to obtain a semantic representation vector; the encoder model is used for modeling semantic information in the object behavior sequence; and performing information recommendation to the target object based on the semantic representation vector. The embodiments of the present application are beneficial to improving the accuracy of information recommendation, and at the same time, improving the applicability of the method.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, and in particular to an information recommendation method and device, an electronic device, and a storage medium. BACKGROUND

[0002] In the fields of recommendation, advertising, risk control, and the like, an object behavior sequence is an important piece of information. Modeling the object behavior sequence to better perform information recommendation is an important method for improving service quality and satisfaction. In related technologies, the common practice is to model behavior features, focusing on the design process of the features, which requires more expert experience, and the model accuracy is not high, and the information recommendation effect is not ideal. SUMMARY

[0003] The main purpose of the embodiments of the present application is to propose an information recommendation method, device, electronic device, and storage medium, which aims to improve the accuracy of information recommendation.

[0004] To achieve the above-mentioned purpose, a first aspect of the embodiments of the present application proposes an information recommendation method, which comprises:

[0005] obtaining an object behavior sequence of a target object; the object behavior sequence comprises behavior information of the target object;

[0006] inputting the object behavior sequence into an embedding layer model to obtain object encoding information; the embedding layer model is used to model time intervals at which the behavior information occurs;

[0007] inputting the object encoding information into an encoder model to obtain a semantic representation vector; the encoder model is used to model semantic information in the object behavior sequence;

[0008] based on the semantic representation vector, performing information recommendation to the target object. By modeling the time intervals in the object behavior sequence, it is convenient to distinguish between continuous relevant behaviors and idle behaviors of the object when performing information recommendation, and at the same time, it is not dependent on expert experience, which is conducive to improving the recommendation accuracy; modeling based on the embedding layer is conducive to the fusion with other models, and improves the applicability of the model.

[0009] In some embodiments, the inputting the object behavior sequence into the embedding layer model to obtain the object encoding information comprises:

[0010] constructing a word sequence according to the object behavior sequence; the word sequence comprises interval-set behavior information and time interval information;

[0011] determining object encoding information according to the word sequence; the object encoding information comprises position encoding information, word type encoding information, and word value encoding information set based on the word sequence.

[0012] In some embodiments, the object behavior sequence is used to represent a plurality of behavior information arranged in chronological order; according to the object behavior sequence, a word sequence is constructed, including:

[0013] A time interval information is inserted between each two behavior information to construct a word sequence; the time interval information is used to represent the time interval between the two behavior information before and after the time interval information;

[0014] In an incremental manner, the index of each word information in the word sequence is updated; wherein the index is encoded from zero.

[0015] In some embodiments, according to the word sequence, the object encoding information is determined, including the following steps:

[0016] The index of the word information is embedded and coded to obtain the position encoding information at the corresponding position;

[0017] The type of the word information is embedded and coded to obtain the word type encoding information at the corresponding position; the type of the word information includes behavior and time interval.

[0018] In some embodiments, according to the word sequence, the object encoding information is determined, including the following steps:

[0019] If the word information in the word sequence belongs to behavior information, the word value encoding information is determined as the behavior feature information of the corresponding behavior information;

[0020] If the word information in the word sequence belongs to time interval information, the word value encoding information is determined by the following vector:

[0021]

[0022] Wherein, Used to represent the minimum time unit required to convert the time interval into a business; Used to represent the time interval information, Used to represent the preset domain vector.

[0023] In some embodiments, the object encoding information is input into an encoder model to obtain a semantic representation vector, including:

[0024] The object encoding information is processed by a transformer structure to obtain a semantic representation vector.

[0025] In some embodiments, the method further includes:

[0026] The semantic representation vector is updated by average pooling.

[0027] To achieve the above object, a second aspect of the embodiment of the present application provides an information recommendation device, which comprises:

[0028] A first module is configured to acquire an object behavior sequence of a target object; the object behavior sequence comprises behavior information of the target object;

[0029] A second module is configured to input the object behavior sequence into an embedding layer model to obtain object encoding information; the embedding layer model is configured to model time intervals at which the behavior information occurs;

[0030] A third module is configured to input the object encoding information into an encoder model to obtain a semantic representation vector; the encoder model is configured to model semantic information in the object behavior sequence;

[0031] A fourth module is configured to perform information recommendation to the target object based on the semantic representation vector.

[0032] To achieve the above object, a third aspect of the embodiment of the present application provides an electronic device, which comprises a memory and a processor; the memory stores a computer program; and the processor implements the information recommendation method of the first aspect when executing the computer program.

[0033] To achieve the above object, a fourth aspect of the embodiment of the present application provides a computer readable storage medium, which stores a computer program; and the computer program is executed by a processor to implement the information recommendation method of the first aspect.

[0034] The information recommendation method, device, electronic device and storage medium provided by the present application, the method comprises: acquiring an object behavior sequence of a target object; the object behavior sequence comprises behavior information of the target object; inputting the object behavior sequence into an embedding layer model to obtain object encoding information; the embedding layer model is configured to model time intervals at which the behavior information occurs; by modeling the time intervals in the object behavior sequence, it is convenient to distinguish continuous related behaviors and idle behaviors of the object when performing information recommendation, so as to improve the recommendation accuracy; modeling based on the embedding layer facilitates the fusion with other models and improves the applicability of the model. Meanwhile, the object encoding information is input into an encoder model to obtain a semantic representation vector; the encoder model is configured to model semantic information in the object behavior sequence; and information recommendation is performed to the target object based on the semantic representation vector. Through the method, the accuracy of information recommendation is improved, and the applicability of the method is improved. BRIEF DESCRIPTION OF DRAWINGS

[0035] Figure 1 is a flowchart of the information recommendation method provided by the embodiment of the present application;

[0036] Figure 2 is Figure 1 a flow chart of one embodiment of step S200 in

[0037] Figure 3 is Figure 2 a flow chart of one embodiment of step S210 in

[0038] Figure 4 is Figure 2 a flow chart of one embodiment of step S220 in

[0039] Figure 5 is a structural schematic diagram of an information recommendation device provided by an embodiment of the present application;

[0040] Figure 6 is a hardware structural schematic diagram of an electronic device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0041] In order to make the purposes, technical solutions and advantages of the present application clearer, the present application is further described in detail below in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, and are not used to limit the present application.

[0042] It should be noted that although the functional modules are divided in the device schematic diagram, and the logical order is shown in the flow chart, in some cases, the steps shown or described can be executed in a manner different from the module division in the device or the order in the flow chart. The terms "first", "second", etc. in the specification and claims and the above-described drawings are used to distinguish similar objects, and do not necessarily describe a specific order or sequence.

[0043] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs. The terms used herein are only for the purpose of describing the embodiments of the present application, and are not intended to limit the present application.

[0044] First, the terms involved in the present application are analyzed:

[0045] Artificial intelligence (AI): is a new technical science that studies, develops theories, methods, technologies and application systems for simulating, extending and expanding human intelligence; artificial intelligence is a branch of computer science, artificial intelligence attempts to understand the essence of intelligence, and produce a new intelligent machine that can react in a similar way to human intelligence, the research in this field includes robots, language recognition, image recognition, natural language processing and expert systems, etc. Artificial intelligence can simulate the information process of human consciousness and thinking. Artificial intelligence is also the theory, method, technology and application system of using digital computer or digital computer controlled machine to simulate, extend and expand human intelligence, perceive environment, acquire knowledge and use knowledge to obtain the best results.

[0046] Nature Language processing (NLP) is an important direction in the field of computer science and artificial intelligence. It studies various theories and methods that can realize effective communication between people and computers using natural language. Natural language processing is a science that integrates linguistics, computer science and mathematics. Therefore, the research in this field will involve natural language, i.e. the language used in daily life, so it is closely related to the study of linguistics. Natural language processing technology usually includes text processing, information extraction, semantic understanding, machine translation, robot question and answer, knowledge graph and many other directions.

[0047] Embedding: a way to convert discrete variables into continuous vector representation. In neural networks, embedding can reduce the spatial dimension of discrete variables, and can also represent the meaning of the vector, visualize the relationship between different discrete variables, input supervised learning tasks, find the nearest neighbor, etc.

[0048] With the continuous development of Internet applications, objects will leave a large number of object behaviors on the Internet and platforms. For recommendation, advertising, risk control and other fields, object behavior sequence is an important information, and modeling object behavior sequence can better recommend information, which is an important method to improve service quality and satisfaction. Object behavior sequence is usually a list composed of specific behaviors of objects, and each specific behavior in the list usually includes two items, the time when the behavior occurs and the specific behavior of the object.

[0049] In the related art, the history behavior sequence of an object is usually modeled by calculating statistical features of the behavior sequence. The focus of this kind of method is how to design the features, which requires more expert experience and feature engineering. Moreover, the current method for modeling the behavior sequence in deep learning only considers the sequence order, and ignores the modeling of the time interval between behaviors. The time interval contains a lot of information in the behavior sequence, for example, when a customer browses goods on an e-commerce platform, similar time intervals are usually in the same category of goods, and after a long time interval, the customer usually starts to browse other goods of another category; therefore, the object behavior can be predicted by modeling the time interval.

[0050] Based on this, the embodiments of the present application provide an information recommendation method and device, electronic equipment and storage medium, aiming to improve the accuracy of recommendation. First, the object behavior sequence of the target object is obtained; the object behavior sequence includes behavior information of the target object; the object behavior sequence is input into an embedding layer model to obtain object encoding information; the embedding layer model is used to model the time interval of the behavior information; by modeling the time interval in the object behavior sequence, it is convenient to distinguish the continuous related behavior and idle behavior of the object when performing information recommendation, so as to improve the recommendation accuracy; based on the embedding layer modeling, it is convenient to realize the fusion with other models, and improve the applicability of the model. Second, the object encoding information is input into an encoder model to obtain a semantic representation vector; the encoder model is used to model the semantic information in the object behavior sequence; based on the semantic representation vector, information is recommended to the target object. Through this method, it is beneficial to improve the accuracy of information recommendation, and at the same time, improve the applicability of the method.

[0051] The information recommendation method, device, electronic equipment and storage medium provided by the embodiments of the present application are specifically described by the following embodiments. First, the information recommendation method in the embodiments of the present application is described.

[0052] The embodiments of the present application can acquire and process related data based on artificial intelligence technology. Among them, artificial intelligence (Artificial Intelligence, AI) is to use digital computers or digital computer controlled machines to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results. Theory, method, technology and application system.

[0053] The basic technology of artificial intelligence generally includes technologies such as sensors, special artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction system, mechatronics, etc. The software technology of artificial intelligence mainly includes computer vision technology, robot technology, biometric technology, speech processing technology, natural language processing technology, and machine learning / deep learning, etc. Several major directions.

[0054] The information recommendation method provided by the embodiments of the present application relates to the technical field of artificial intelligence. The information recommendation method provided by the embodiments of the present application can be applied to a terminal, can also be applied to a server end, and can also be software running in the terminal or the server end. In some embodiments, the terminal can be a smart phone, a tablet computer, a notebook computer, a desktop computer, etc.; the server end can be configured as a stand-alone physical server, can also be configured as a server cluster or a distributed system composed of multiple physical servers, and can also be configured as a cloud server providing basic cloud computing services such as cloud service, cloud database, cloud computing, cloud function, cloud storage, network service, cloud communication, middleware service, domain name service, security service, CDN, and big data and artificial intelligence platform; and the software can be an application that implements the information recommendation method, but is not limited to the above forms.

[0055] The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld devices or portable devices, tablet devices, multi-processor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application can also be practiced in a distributed computing environment, in which tasks are performed by remote processing devices connected by a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media, including storage devices.

[0056] It should be noted that in each specific embodiment of the present application, when it is necessary to perform relevant processing according to user information, user behavior data, user historical data, and user location information, etc. related to the identity or characteristics of the user, the user's permission or consent will be obtained first, and the collection, use and processing of these data will comply with relevant laws, regulations and standards. In addition, when the embodiments of the present application need to obtain sensitive personal information of the user, the separate permission or separate consent of the user will be obtained through a pop-up window or a jump to a confirmation page, and after obtaining the separate permission or separate consent of the user, the necessary user-related data for the normal operation of the embodiments of the present application will be obtained.

[0057] Figure 1 is an optional flowchart of the information recommendation method provided by the embodiments of the present application, Figure 1 The method in can include but is not limited to steps S100 to S400.

[0058] Step S100, obtaining an object behavior sequence of a target object; the object behavior sequence comprises behavior information of the target object;

[0059] In some embodiments, the target object is an object that needs information recommendation. For example, in a shopping application, a terminal recommends a shopping category or a certain product to the target object. Alternatively, in a news application, a certain type of news is displayed on a prominent position on the screen of the terminal to achieve information recommendation to the target object. Meanwhile, the method of the present application can also be used in the fields of advertising, risk control, etc., and the present application is not limited in this regard. Specifically, by modeling and analyzing the object behavior sequence of the target object, the information needed by the target object is obtained; in some possible implementation manners, the object behavior sequence of the target object needs to be obtained. Specifically, the behavior sequence can be used to represent a sequence composed of a plurality of behavior events in a certain order in the time dimension. For example, in a short video application, the behavior information of the object can be a series of click actions of the object, which is used to represent that the object selects a short video category of interest. The present application models and analyzes the behavior of the object by obtaining the above click behavior. That is, by obtaining the object behavior sequence of the target object, the object behavior sequence is modeled and analyzed subsequently to determine the recommended information for the target object.

[0060] Step S200, inputting the object behavior sequence into an embedding layer model to obtain object encoding information; the embedding layer model is used to model the time interval at which the behavior information occurs;

[0061] In some embodiments, by using the embedding layer model, the time interval at which the behavior information occurs is modeled, and in the process of recommending information to the object behavior, the time interval influencing factor is added to improve the accuracy of information recommendation. In some possible implementation manners, by establishing an embedding layer model in the embedding layer, the time modeling is achieved, and it is also easy to combine with other behavior models for prediction or recommendation work, thereby improving the application breadth of information recommendation. Specifically, by using the embedding layer model, an embedding list is finally output, and the information related to the time interval is introduced in the embedding, thereby achieving information recommendation based on the time interval.

[0062] In some possible implementation manners, the input of the embedding layer model is the object behavior sequence. Specifically, the object behavior sequence can be a list, and the length of the list is L. Each behavior sequence in the list includes a timestamp and specific behavior information. The list has been sorted according to the time sequence of the behaviors, and the behaviors in the earlier part of the list occur earlier in time, and the behaviors in the later part of the list occur later in time. For the object behavior sequence input of the embedding layer model in the above embodiment, the output can be object encoding information E represented by embedding, and the size of a single sample is (M, H). M is the length of the embedding list, and M = 2*L-1. H is the length of the embedding hidden vector. For the short video application scenario in the above embodiment, for a series of click actions of an object, the time interval between every two different click actions can be used to represent the degree of love of the object for the selected short video category, and the behavior of the object can be predicted through the time interval; and then used in the information recommendation field.

[0063] In step S300, the object encoding information is input into an encoder model to obtain a semantic representation vector; the encoder model is used to model semantic information in the object behavior sequence.

[0064] In some embodiments, the semantic representation vector of the object behavior is obtained through the encoder model, to facilitate subsequent information recommendation. Specifically, taking click rate estimation in recommendation as an example, customer historical behavior modeling is an important method to realize click rate prediction. Generally, the customer historical behavior is modeled into a fixed-size semantic vector, which is then concatenated with other features to input into a machine learning training model to probabilistically predict the click rate, and then realize information recommendation for a target object. Therefore, the object encoding information containing the time interval information is converted into a fixed-size semantic vector through the encoder, to realize concatenation with other semantic features, and then realize information recommendation. Of course, the application does not limit the modeling process, the concatenation process, and the selection and training process of the model of other features.

[0065] In step S400, information is recommended to the target object based on the semantic representation vector.

[0066] In some embodiments, the time interval of the behavior information of the target object is included in the semantic representation vector, and information is recommended to the target object based on the semantic representation vector; the influence of the time interval of the object behavior on the prediction of the object behavior can be fully considered, to improve the accuracy of information recommendation. For example, the prediction of the object behavior can be realized through a DNN model, and then the recommended information is provided. It can be understood that the prediction of the object behavior can also be realized through other artificial intelligence models. The application does not limit the selection of the specific model.

[0067] It can be known that the information recommendation method provided in the application obtains an object behavior sequence of a target object, the object behavior sequence comprises behavior information of the target object, inputs the object behavior sequence into an embedding layer model to obtain object encoding information, the embedding layer model is used for modeling time intervals at which the behavior information occurs, models the time intervals in the object behavior sequence, and facilitates distinguishing continuous relevant behaviors and idle behaviors of the object when information is recommended, so as to improve recommendation accuracy; modeling based on the embedding layer facilitates fusion with other models and improves applicability of the model. Meanwhile, the object encoding information is input into an encoder model to obtain a semantic representation vector, the encoder model is used for modeling semantic information in the object behavior sequence, and information is recommended to the target object based on the semantic representation vector. Through the method, the accuracy of information recommendation is improved, and applicability of the method is improved.

[0068] Please refer to Figure 2 In some embodiments, step S200 can include, but is not limited to, steps S210 to S220:

[0069] Step S210, constructing a word sequence according to the object behavior sequence; the word sequence comprises interval set behavior information and time interval information;

[0070] Step S220, determining object encoding information according to the word sequence; the object encoding information comprises position encoding information, word type encoding information and word value encoding information set based on the word sequence.

[0071] In some embodiments, the embedding layer model is modeled based on time at which behavior information occurs, specifically, the time interval of the behavior information can be written into the object behavior sequence to generate a word sequence. Exemplarily, the word sequence can be a token sequence. The input is the object behavior sequence B , the length is L ; the corresponding output is a token sequence K , the length is M , M = 2*L - 1 . The object behavior sequence can be represented by the following vector B .

[0072] Formula (1)

[0073] Wherein, is the i th specific object behavior; i , , is specific behavior information, is the time at which the behavior occurs.

[0074] As can be understood from one embodiment of the above object behavior sequence, the object behavior sequence includes behavior information and the time of occurrence of the behavior information. The required time interval needs to be achieved by adding time interval information to the object behavior sequence, which facilitates subsequent analysis and recommendation of the time interval and behavior information. In some possible implementations, this application arranges the behavior information and time interval information sequentially to form a word sequence. Therefore, the word information in the word sequence may represent different types of data. For different types of data, they are uniformly represented by object encoding information, which facilitates subsequent semantic representation and information recommendation. Specifically, three encoding information is set based on the specific position, type, and value of the word information, namely position encoding information, word type encoding information, and word value encoding information. Through the above word sequence setting, time interval information can be modeled, improving the accuracy of information recommendation.

[0075] Please see Figure 3 In some embodiments, the object behavior sequence is used to characterize several behavioral information arranged in chronological order; step S210 may include, but is not limited to, steps S211 to S212:

[0076] S211, Insert time interval information between every two behavioral information to construct a word sequence; the time interval information is used to characterize the time interval between two behavioral information before and after the time interval information.

[0077] S212, update the subscript of each word in the word sequence in an incremental manner; wherein the subscript is encoded starting from zero.

[0078] In some embodiments, time interval information can be placed between action information to characterize the time interval between two consecutive action information. For the original object action sequence, after inserting new time interval information, the indexes corresponding to the action information need to be updated. Based on the object action sequence in the above embodiments... B The corresponding output word sequence is: token sequence K , length is M , M = 2*L - 1 .exist B Insert time interval information, token sequence K Represented by the following vectors:

[0079] Formula (2)

[0080] in, It is the first i Individual word information. Specifically, the token sequence. KThe embedding contains two tokens, one is the specific behavior of the object (behavior information), and the other is the time interval information. Exemplarily, In relation to the above embodiment , when i is even, is a specific behavior, ; when i is odd, represents the time interval information between two adjacent behaviors, and at this time . The construction of the word sequence is realized by the above embodiment, which facilitates subsequent modeling and analysis of the time interval information.

[0081] Please refer to Figure 4 , in some embodiments, step S220 can include but is not limited to steps S221 to S222:

[0082] Step S221, embedding encoding processing is performed on the index of the word information to obtain position encoding information at the corresponding position;

[0083] Step S222, embedding encoding processing is performed on the type of the word information to obtain word type encoding information at the corresponding position; the type of the word information includes behavior and time interval.

[0084] In some embodiments, based on the word information at different positions, position encoding and type encoding are performed. Specifically, the position encoding information can be obtained by the index of each word information, and each subscript position takes the corresponding embedding to obtain the position encoding information. That is, the position encoding information is represented by the following vector:

[0085] Equation (3)

[0086] Wherein, , is used to represent the embedding encoding of the position information. It should be noted that the embedding encoding of the position information is used to represent the position of the word information in the word sequence, wherein the embedding encoding can be obtained by training or calculated by using a preset formula, which is not limited by the present application.

[0087] Similarly, the word sequence is composed by the cross arrangement of the behavior information and the time interval information, which is used to distinguish the behavior part and the time interval part. For each different type, the embedding of each type in the word sequence is taken to obtain the word type encoding information. The word type encoding information is represented by the following vector:

[0088] Formula (4)

[0089] Among them, the Used to characterize the i The embedding encoding of the type to which each word belongs. Similarly, the embedding encoding of the type information is used to represent the specific type of the word information. The embedding encoding can be obtained through training or calculated using a preset formula, and this application does not impose any restrictions. It is understood that the positional encoding information and the word type encoding information after embedding encoding have the same dimension.

[0090] In some embodiments, step S220 may include, but is not limited to, steps S2201 to S2202:

[0091] S2201, If ​​the word information in the word sequence belongs to behavioral information, determine the word value encoding information as the behavioral feature information of the corresponding behavioral information;

[0092] S2202, If the word information in the word sequence belongs to time interval information, the word value encoding information is determined by the following vectors:

[0093] Formula (5)

[0094] in, Used to characterize the smallest unit of time required to convert time intervals into business operations; Used to characterize time interval information Used to characterize a predefined domain vector.

[0095] In some embodiments, the word sequence includes two types of word information: behavioral information and time interval information. The representation method of word value encoding information is different for different types. Specifically, for behavioral type word information, word value encoding information can take the corresponding embedding of the corresponding behavioral feature. For time interval type word information, word value encoding information can be represented by the vector shown in formula (5). Functions can be used to convert dates to specific data types, and date / time data can be displayed in different formats. In this application, Time interval Converting to the smallest time unit required for business operations is typically done in seconds. It's understandable that time differences vary considerably, ranging from a few seconds to days or even years; therefore, a log function is used for normalization. However, time differences can be less than a second or even zero seconds, and using the log function might result in large negative numbers or even errors. Therefore, [the following is omitted as it's not directly related to the previous sentence]. The vector representation is performed in a way that ensures the value after log is within a reasonable range. It should be noted that the above embodiments are exemplary and do not limit the specific representation method of word value encoding information.

[0096] In some possible implementations, word value encoding information is represented by the following vectors:

[0097] Formula (6)

[0098] Among them, the Used to characterize the i The word value encoding information for each word.

[0099] Based on positional encoding information, word type encoding information, and word value encoding information, the final object encoding information can be represented by the following vector:

[0100] Formula (7)

[0101] in, The vector representation of the object's encoded information is obtained through formula (7), where represents the object's behavior sequence over a time interval, which facilitates subsequent improvement of the accuracy of information recommendation based on the time interval.

[0102] In some embodiments, step S300 may include, but is not limited to, step S310:

[0103] Step S310: Process the object encoding information through the transformer structure to obtain a semantic representation vector.

[0104] In some embodiments, the transformer structure differs from CNNs and RNNs used in deep learning tasks. The BERT model is a model built upon the transformer structure. The transformer structure employs an encoder-decoder architecture, which is more complex and yields more significant results than the attention model. Specifically, in the embodiments provided in this application, the aforementioned object encoding information... E A list of mixed semantic vectors can be obtained by processing using a transformer structure. The size of a single sample is (L, H) Semantic representation facilitates subsequent information recommendation processing.

[0105] In some embodiments, the information recommendation method may include, but is not limited to, step S320:

[0106] Step S320: Update the semantic representation vector by average pooling.

[0107] In some embodiments, higher-order features are extracted through average pooling processing to improve the accuracy of information recommendation. Specifically, the list of mixed semantic vectors described above A is subjected to average pooling processing to obtain a semantic representation vector of the object behavior sequence, which is represented by the following vector:

[0108] Equation (8)

[0109] wherein, is the S th value of i .

[0110] The present application provides an information recommendation method, which comprises: obtaining an object behavior sequence of a target object; the object behavior sequence comprises behavior information of the target object; inputting the object behavior sequence into an embedding layer model to obtain object encoding information; the embedding layer model is used to model the time interval at which the behavior information occurs; by modeling the time interval in the object behavior sequence, it is convenient to distinguish between continuous relevant behaviors and idle behaviors of the object when performing information recommendation, so as to improve the recommendation accuracy; modeling based on the embedding layer facilitates the fusion with other models and improves the applicability of the model. Meanwhile, the object encoding information is input into an encoder model to obtain a semantic representation vector; the encoder model is used to model the semantic information in the object behavior sequence; based on the semantic representation vector, information is recommended to the target object. Through this method, the accuracy of information recommendation is improved, and the applicability of the method is improved.

[0111] It should be noted that the information recommendation method in the present application can also be used in video playing applications. By obtaining object behavior information such as sending a barrage of behaviors, sharing a video connection, etc., the time interval in the behavior sequence is modeled and analyzed by the method provided in the present application to improve the accuracy of information recommendation. It can be understood that the method provided in the present application is also applicable to other recommendation fields such as advertising and risk control.

[0112] Referring to Figure 5 , the present application also provides an information recommendation device, which can implement the information recommendation method described above. The device comprises:

[0113] A first module 510 is configured to obtain an object behavior sequence of a target object; the object behavior sequence comprises behavior information of the target object;

[0114] A second module 520 is configured to input the object behavior sequence into an embedding layer model to obtain object encoding information; the embedding layer model is used to model the time interval at which the behavior information occurs;

[0115] The third module 530 is configured to input the object coding information into an encoder model to obtain a semantic representation vector; the encoder model is configured to model semantic information in the object behavior sequence.

[0116] The fourth module 540 is configured to recommend information to the target object based on the semantic representation vector.

[0117] In some embodiments, the second module includes a first unit configured to construct a word sequence according to the object behavior sequence; the word sequence includes interval arranged behavior information and time interval information; determine object coding information according to the word sequence; the object coding information includes position coding information, word type coding information and word value coding information arranged based on the word sequence.

[0118] In some embodiments, the object behavior sequence is configured to represent a plurality of behavior information arranged in chronological order, and the first unit is further configured to insert time interval information between each two behavior information to construct a word sequence; the time interval information is configured to represent the time interval between the two behavior information before and after the time interval information; update the index of each word information in the word sequence in an incremental manner; wherein the index is encoded from zero.

[0119] In some embodiments, the first unit is further configured to perform embedding coding processing on the index of the word information to obtain position coding information at the corresponding position; perform embedding coding processing on the type of the word information to obtain word type coding information at the corresponding position; the type of the word information includes behavior and time interval.

[0120] In some embodiments, the second module is further configured to determine the word value coding information as the behavior feature information of the corresponding behavior information if the word information in the word sequence belongs to the behavior information; if the word information in the word sequence belongs to the time interval information, determine the word value coding information through the following vector:

[0121]

[0122] wherein, is configured to represent the time interval converted into the minimum time unit required by the business; is configured to represent the time interval information, is configured to represent a preset domain vector.

[0123] In some embodiments, the third module is further configured to process the object coding information through a transformer structure to obtain a semantic representation vector.

[0124] In some embodiments, the third module is further configured to update the semantic representation vector through average pooling.

[0125] The specific implementation of the information recommendation device is basically the same as the specific embodiment of the above information recommendation method, and will not be repeated here.

[0126] The device includes a first module, a second module, a third module, and a fourth module. Specifically, an object behavior sequence of a target object is obtained; the object behavior sequence includes behavior information of the target object; the object behavior sequence is input into an embedding layer model to obtain object encoding information; the embedding layer model is used to model time intervals at which the behavior information occurs; by modeling the time intervals in the object behavior sequence, the continuous relevant behavior and idle behavior of the object can be distinguished when information is recommended, thereby improving the recommendation accuracy; modeling based on the embedding layer facilitates the fusion with other models and improves the applicability of the model. Meanwhile, the object encoding information is input into an encoder model to obtain a semantic representation vector; the encoder model is used to model semantic information in the object behavior sequence; based on the semantic representation vector, information is recommended to the target object. Through the above information recommendation device, the accuracy of information recommendation is improved, and the applicability of the method is improved.

[0127] The embodiments of the present application also provide an electronic device including a memory and a processor. The memory stores a computer program, and the processor implements the above information recommendation method when executing the computer program. The electronic device can be any intelligent terminal including a tablet computer, a vehicle-mounted computer, etc.

[0128] Please refer to Figure 6 , Figure 6 The hardware structure of the electronic device of another embodiment is illustrated, and the electronic device includes:

[0129] The processor 601 can be implemented in a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, etc., and is used to execute related programs to implement the technical solutions provided by the embodiments of the present application.

[0130] The memory 602 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 602 can store an operating system and other application programs. When the technical solutions provided by the embodiments of the present application are implemented by software or firmware, the related program codes are saved in the memory 602 and are called and executed by the processor 601 to implement the information recommendation method of the embodiments of the present application.

[0131] The input / output interface 603 is configured to realize information input and output.

[0132] The communication interface 604 is configured to realize communication interaction between the device and other devices, and the communication can be realized in a wired manner (for example, a USB, a network cable, or the like) or in a wireless manner (for example, a mobile network, WIFI, Bluetooth, or the like).

[0133] The bus 605 is configured to transmit information between various components (for example, the processor 601, the memory 602, the input / output interface 603, and the communication interface 604) of the device.

[0134] The processor 601, the memory 602, the input / output interface 603, and the communication interface 604 are communicatively connected to each other in the device through the bus 605.

[0135] The embodiment of the present application further provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to realize the information recommendation method.

[0136] The memory is a non-transitory computer readable storage medium, and can be used to store a non-transitory software program and a non-transitory computer executable program. In addition, the memory can include a high-speed random access memory, and can further include a non-transitory memory, for example, at least one magnetic disk storage device, a flash memory device, or other non-transitory solid-state memory device. In some embodiments, the memory can optionally include a memory remotely arranged relative to the processor, and the remote memory can be connected to the processor through a network. Examples of the network include but are not limited to the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof.

[0137] The information recommendation method, device, electronic equipment, and storage medium provided by the present application obtain an object behavior sequence of a target object. The object behavior sequence includes behavior information of the target object. The object behavior sequence is input into an embedding layer model to obtain object encoding information. The embedding layer model is used to model time intervals at which the behavior information occurs. By modeling the time intervals in the object behavior sequence, it is convenient to distinguish continuous related behaviors and idle behaviors of the object when information is recommended, so as to improve the recommendation accuracy. Modeling based on the embedding layer facilitates the fusion with other models and improves the applicability of the model. Meanwhile, the object encoding information is input into an encoder model to obtain a semantic representation vector. The encoder model is used to model semantic information in the object behavior sequence. Based on the semantic representation vector, information is recommended to the target object. Through the method, the accuracy of information recommendation is improved, and the applicability of the method is improved.

[0138] The embodiments described in the specification of the present application are intended to more clearly illustrate the technical solutions of the present application, and do not constitute a limitation on the technical solutions provided by the present application. Those skilled in the art can know that, with the evolution of technology and the appearance of new application scenarios, the technical solutions provided by the present application are also applicable to similar technical problems.

[0139] Those skilled in the art can understand that, Figures 1-6 The technical solutions shown in the above description do not constitute a limitation on the present application, and can include more or fewer steps than shown, or combine certain steps, or different steps.

[0140] The device embodiments described above are merely illustrative, and units described as separate components can or can not be physically separated, i.e., can be located in one place, or can be distributed on multiple network units. Part or all of the modules can be selected according to actual needs to achieve the purpose of the present embodiment.

[0141] Those skilled in the art can understand that all or some of the steps in the above disclosed method, the functions of the modules / units in the system and the device can be implemented as software, firmware, hardware and their appropriate combinations.

[0142] The terms "first", "second", "third", "fourth" and the like (if any) in the specification of the present application and the above description are used to distinguish similar objects, and do not necessarily indicate a specific order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device that includes a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but can include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0143] It should be understood that, in the application, "at least one" refers to one or more, and "multiple" refers to two or more. "And / or" is used to describe the association relationship of the associated objects, which means that there can be three relationships, for example, "A and / or B" can represent three cases of only A, only B and A and B existing at the same time, wherein A and B can be singular or plural. The character " / " generally represents an "or" relationship between the front and rear associated objects. "At least one of the following" or the like means any combination of these items, including any combination of single or multiple items. For example, at least one of a, b or c can represent a, b, c, "a and b", "a and c", "b and c", or "a and b and c", wherein a, b and c can be single or multiple.

[0144] In several embodiments provided in the application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of the above units is only a logical function division, and actual implementation can have another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the displayed or discussed mutual ones can be indirect coupling or communication connection through some interfaces, devices or units, and can be electrical, mechanical or other forms.

[0145] The units described above as separate components can or can not be physically separated, and the components displayed as units can or can not be physical units, that is, they can be located in one place, or can be distributed on multiple network units. According to actual needs, part or all of the units can be selected to achieve the purpose of the embodiment scheme.

[0146] In addition, each functional unit in each embodiment of the application can be integrated in one processing unit, or each unit can be physically present separately, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0147] The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application, essentially or in other words, the part that contributes to the prior art or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes multiple instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various program storage media.

[0148] The preferred embodiments of the embodiments of the present application are described above with reference to the accompanying drawings, and are not limited to the scope of the embodiments of the present application. Any modifications, equivalent replacements and improvements made by those skilled in the art without departing from the scope and essence of the embodiments of the present application shall be within the scope of the embodiments of the present application.

Claims

1. An information recommendation method characterized by comprising: The method comprises: acquiring an object behavior sequence of a target object; the object behavior sequence comprises behavior information of the target object; inputting the object behavior sequence into an embedding layer model to obtain object encoding information; the embedding layer model is used for modeling time intervals at which the behavior information occurs; inputting the object encoding information into an encoder model to obtain a semantic representation vector; the encoder model is used for modeling semantic information in the object behavior sequence; based on the semantic representation vector, information is recommended to the target object; wherein the inputting of the object behavior sequence into the embedding layer model to obtain the object encoding information comprises: constructing a word sequence according to the object behavior sequence; the word sequence comprises interval-set behavior information and time interval information; determining object encoding information according to the word sequence; the object encoding information comprises position encoding information, word type encoding information and word value encoding information set based on the word sequence; wherein the object behavior sequence is used for representing a plurality of behavior information arranged in chronological order; the constructing of the word sequence according to the object behavior sequence comprises: inserting time interval information between each two behavior information to construct a word sequence; the time interval information is used for representing the time interval between the two behavior information before and after the time interval information; updating the index of each word information in the word sequence in an incremental manner; wherein the index is encoded from zero.

2. The recommendation method of claim 1, wherein, The determining of the object encoding information according to the word sequence comprises the following steps: embedding encoding processing is performed on the index of the word information to obtain position encoding information at the corresponding position; embedding encoding processing is performed on the type of the word information to obtain word type encoding information at the corresponding position; the type of the word information comprises behavior and time interval.

3. The recommendation method of claim 1, wherein, The determining of the object encoding information according to the word sequence comprises the following steps: if the word information in the word sequence belongs to behavior information, determining the word value encoding information as behavior feature information of the corresponding behavior information; if the word information in the word sequence belongs to time interval information, determining the word value encoding information through the following vector: ， wherein, for characterizing the minimum time unit required to transform a time interval into a service; for characterizing time interval information, for characterizing a preset domain vector.

4. The recommendation method of claim 1, wherein, the inputting of the object encoding information into the encoder model to obtain the semantic representation vector comprises: processing the object encoding information through a transformer structure to obtain a semantic representation vector.

5. The recommendation method according to claim 4, characterized in that, The method further comprises: updating the semantic representation vector through average pooling.

6. An information recommendation device characterized by comprising: The apparatus comprises: a first module for acquiring an object behavior sequence of a target object; the object behavior sequence comprises behavior information of the target object; a second module for inputting the object behavior sequence into an embedding layer model to obtain object encoding information; the embedding layer model is used for modeling time intervals at which the behavior information occurs; a third module for inputting the object encoding information into an encoder model to obtain a semantic representation vector; the encoder model is used for modeling semantic information in the object behavior sequence; a fourth module for recommending information to the target object based on the semantic representation vector; The second module is configured to input the object behavior sequence into an embedding layer model to obtain object encoding information, including: According to the object behavior sequence, a word sequence is constructed; the word sequence includes interval arranged behavior information and time interval information; According to the word sequence, object encoding information is determined; the object encoding information includes position encoding information, word type encoding information and word value encoding information arranged based on the word sequence; The object behavior sequence is used to represent a plurality of behavior information arranged in chronological order; the word sequence is constructed according to the object behavior sequence, including: Time interval information is inserted between each two behavior information to construct a word sequence; the time interval information is used to represent the time interval between the two behavior information before and after the time interval information; In an incremental manner, the index of each word information in the word sequence is updated; the index is encoded from zero.

7. An electronic device, comprising: The electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the information recommendation method of any one of claims 1 to 5 when executing the computer program.

8. A computer-readable storage medium storing a computer program, the computer-readable storage medium comprising: The computer program is executed by the processor to implement the information recommendation method of any one of claims 1 to 5.

Images