Scene service recommendation method, device, equipment and storage medium

Abstract

An embodiment of the present application provides a method, device, equipment and storage medium for scenario service recommendation. It relates to the field of artificial intelligence technology. The method includes: determining target service information according to target historical behavior information, and calling a target intent classification model corresponding to the target service information to perform intent prediction on the target historical behavior information and user profile information, so as to determine a target intent from multiple candidate intents corresponding to the target service information; obtaining at least one scenario service having an association relationship with the target intent based on a knowledge graph, determining a target scenario service from the at least one scenario service, and outputting the target scenario service. It can effectively improve the accuracy of the target scenario service. The present application may involve blockchain technology, such as writing the target scenario service into a blockchain. The present application also relates to the field of digital medical technology, such as the target historical behavior information includes information in the field of digital medical technology.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular, to a method, device, equipment and storage medium for scenario service recommendation. Background Art

[0002] With the rapid development of Internet technology, the demand for customer service personnel is increasing. Due to factors such as the long training cycle or high labor cost of human customer service, intelligent customer service robots have gradually entered people's lives.

[0003] However, currently, during the conversation between a user and an intelligent customer service robot, the intelligent customer service robot can often only search for corresponding reply content according to the questions raised by the user, and cannot give effective scenario services for related user behaviors. Summary of the Invention

[0004] Embodiments of this application provide a method, device, equipment and storage medium for scenario service recommendation, which can effectively improve the accuracy of target scenario services and enhance user viscosity.

[0005] In a first aspect, embodiments of this application provide a method for scenario service recommendation, and the method includes:

[0006] Obtain the target historical behavior information of the target user, and determine the target business information according to the target historical behavior information;

[0007] Obtain the user portrait information of the target user, and call the target intent classification model corresponding to the target business information to perform intent prediction on the target historical behavior information and the user portrait information, so as to determine the target intent from multiple candidate intents corresponding to the target business information;

[0008] Obtain at least one scenario service associated with the target intent based on the knowledge graph; the knowledge graph includes the association relationship between each candidate intent and the scenario service;

[0009] Determine the target scenario service from at least one scenario service, and output the target scenario service.

[0010] In a second aspect, embodiments of this application provide a scenario service recommendation device, and the device includes:

[0011] A determination unit, configured to obtain the target historical behavior information of the target user, and determine the target business information according to the target historical behavior information;

[0012] An intent prediction unit, configured to obtain the user portrait information of the target user, and call the target intent classification model corresponding to the target business information to perform intent prediction on the target historical behavior information and the user portrait information, so as to determine the target intent from multiple candidate intents corresponding to the target business information;

[0013] An acquisition unit, configured to acquire at least one scenario service associated with a target intent based on a knowledge graph; the knowledge graph includes the association relationships between each candidate intent and scenario services;

[0014] The determination unit is further configured to determine a target scenario service from the at least one scenario service and output the target scenario service.

[0015] In a third aspect, an embodiment of the present application provides a scenario service recommendation device, which includes an input interface and an output interface, and the scenario service recommendation device further includes:

[0016] A processor, adapted to implement one or more instructions; and,

[0017] A computer storage medium, which stores one or more instructions, and the one or more instructions are adapted to be loaded and executed by the processor to perform the method described in the first aspect above.

[0018] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium, which stores computer program instructions, and when the computer program instructions are executed by a processor, they are used to perform the method described in the first aspect.

[0019] In the embodiment of the present application, the scenario service recommendation device can determine target service information according to target historical behavior information, and call a target intent classification model corresponding to the target service information to perform intent prediction on the target historical behavior information and user portrait information, so as to determine a target intent from multiple candidate intents corresponding to the target service information, and acquire at least one scenario service associated with the target intent based on the knowledge graph, and then determine the target scenario service. In order to obtain the target scenario service, the target intent classification model can be called first to perform intent prediction on the target historical behavior information and user portrait information to determine the target intent from multiple candidate intents, and then the scenario services associated with the target intent can be acquired based on the knowledge graph, and the target scenario service to be pushed can be determined. It can adapt to user needs, intelligently output the target scenario service to the user, and improve the user experience. Moreover, in the present application, the target intent classification model is called first to determine the target intent, and then at least one scenario service associated with the target intent is determined according to the target intent, which can avoid too many candidate items in the classification task, avoid the long-tail phenomenon, and is easier to maintain. In addition, since different intent classification models are trained for different service information, more accurate target intent can be obtained through the target intent classification model corresponding to the target service information, so that the target scenario service determined according to at least one scenario service associated with the target intent is more accurate. It can effectively improve the accuracy of the target scenario service and enhance user stickiness. Description of the Drawings

[0020] To more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the accompanying drawings required for the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can be obtained based on these drawings.

[0021] Figure 1 is a schematic structural diagram of a scenario service recommendation system provided by an embodiment of the present application;

[0022] Figure 2 is a schematic flowchart of a scenario service recommendation method provided by an embodiment of the present application;

[0023] Figure 3 is a schematic flowchart of another scenario service recommendation method provided by an embodiment of the present application;

[0024] Figure 4 is a schematic structural diagram of a scenario service recommendation device provided by an embodiment of the present application;

[0025] Figure 5 is a schematic structural diagram of a scenario service recommendation device provided by an embodiment of the present application. Detailed implementation manners

[0026] The following will clearly and completely describe the technical solutions in the embodiments of the present application in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, rather than all embodiments. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present application.

[0027] With the rapid development of Internet technology, e-commerce has gradually entered people's daily lives, and people's demand for customer service personnel has become increasingly large. Due to factors such as the long training cycle or high labor cost of human customer service, intelligent customer service robots have gradually become a research focus. However, currently, in the process of the user's conversation with the intelligent customer service robot, the business process is complex, and how to determine the scenario services included in the business process is an urgent problem to be solved.

[0028] Currently, in the field of user behavior, there are various user behavior analysis theoretical models, such as behavior event analysis models, page click analysis models, user behavior path analysis models, funnel analysis models, user portrait analysis models, etc. By analyzing users, the user's behavior habits can be understood to perform more accurate and effective promotion.

[0029] Based on this, in order to effectively improve the accuracy of scenario services, an embodiment of this application proposes a scenario service recommendation solution. In this scenario service recommendation solution, by analyzing the historical behavior information of users and combining user portraits, in a specific business scenario, artificial intelligence (AI) technology can be used to match user intentions and push the scenario services corresponding to the user intentions. Specifically, the scenario service recommendation device can obtain the target historical behavior information of the target user, determine the target business information according to the target historical behavior information, obtain the user portrait information of the target user, and call the target intention classification model corresponding to the target business information to perform intention prediction on the target historical behavior information and the user portrait information, and determine the target intention from multiple candidate intentions corresponding to the target business information; obtain at least one scenario service associated with the target intention based on the knowledge graph, where the knowledge graph includes the association relationship between each candidate intention and the scenario service, and determine the target scenario service from at least one scenario service, and output the target scenario service. This can effectively improve the accuracy of the target scenario service and enhance user stickiness.

[0030] Among them, artificial intelligence is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results. Artificial intelligence basic technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, and mechatronics. Artificial intelligence software technologies mainly include several major directions such as computer vision technology, robotics technology, biometric technology, speech processing technology, natural language processing technology, and machine learning / deep learning. Optionally, the scenario service recommendation method can build an intention classification model based on machine learning algorithms, and can call the intention classification model to perform intention prediction on the historical behavior information and the user portrait information to determine the target intention.

[0031] In one embodiment, the scenario service recommendation method can be used for scenario service recommendation. The scenario service recommendation method can be applied in a scenario service recommendation system as Figure 1 shown, as Figure 1 shown, the scenario service recommendation system can at least include: a scenario service recommendation device 11 and a terminal device 12, where the scenario service recommendation device 11 is a device running an intention classification model, and the scenario service recommendation device 11 can be an intelligent customer service robot. Optionally, the scenario service recommendation device 11 can be as Figure 1The server shown can be an independent physical server, a server cluster or a 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, Content Delivery Network (CDN), middleware services, domain name services, security services, and big data and artificial intelligence platforms, etc. The scenario service recommendation device 11 can also be a terminal device. Among them, the terminal device can include, but is not limited to: smart phones, tablet computers, laptop computers, wearable devices, desktop computers, etc. Among them, the terminal device 12 can be a terminal device associated with the user.

[0032] Please refer to Figure 2 , which is a schematic flow chart of a scenario service recommendation method proposed in an embodiment of the present application. As Figure 2 shown, the scenario service recommendation method includes S201 - S204:

[0033] S201: Obtain the target historical behavior information of the target user, and determine the target business information according to the target historical behavior information.

[0034] Among them, the target user can be any user in the scenario service recommendation system. The scenario service recommendation device can obtain the target historical behavior information of the target user, and analyze the target historical behavior information based on the user behavior analysis theoretical model to determine the target business information.

[0035] In one embodiment, the historical behavior information can be used to indicate the behavior characteristics of the user on the page in the business field. Among them, the business field is the medical field, and the page in this business field can include medical data, such as personal health records, prescriptions, inspection reports and other data.

[0036] Among them, the historical behavior information is used to indicate the behavior characteristics of the user on the page. In one embodiment, the historical behavior information can include, but is not limited to, one or more of the number of page views, the duration of page views, and the number of page clicks within a time period. The scenario service recommendation device can obtain the number of page views, the duration of page views, and the number of page clicks of each page of the target user within the time period through data embedding, and analyze the number of page views, the duration of page views, and the number of page clicks of each page based on the user behavior analysis theoretical model to determine the target business information.

[0037] Among them, the target historical behavior information may involve behavior characteristics within multiple pages. The target historical behavior information may include behavior characteristics within at least one page. For example, in a scenario service recommendation system, there are pages A, B, C, and D. The target historical behavior information may include behavior characteristics within pages A, B, and C. The target historical behavior information may include that the number of views within page A is 1, the viewing duration within page A is 100 seconds, and the number of clicks within page A is 2; the number of views within page B is 1, the viewing duration is 200 seconds, and the number of clicks within page B is 5; the number of views within page C is 3, and the viewing duration is 100 seconds. And so on.

[0038] S202: Obtain the user profile information of the target user, and call the target intent classification model corresponding to the target business information to perform intent prediction on the target historical behavior information and the user profile information, so as to determine the target intent from multiple candidate intents corresponding to the target business information.

[0039] Among them, the user profile information may refer to information related to the user and used to describe the user's characteristics. Optionally, the user profile information may include user attribute information and / or user label information. Among them, the user attribute information may refer to the inherent attribute information of the user, and the user attribute information may include, but is not limited to, one or more of name, nickname, age, gender, place of residence, nationality, occupation, constellation, blood type, and identity identifier. The user label information refers to the abstraction and classification summary of a certain characteristic of the user. Specifically, the user label information may be obtained based on the analysis of user behavior data. User behavior data refers to the behavior data generated by the user when using the business product. For example, when the user uses the financial business, there is corresponding user behavior data within the financial business, which may include, but is not limited to, the user's deposits, the user's loan products, the number of the user's loans, the loan amount of the user, the financial products purchased by the user, the financial activities participated by the user, and so on. Another example is that when the user uses the social business, there is corresponding behavior data within the social business, which may include, but is not limited to, the user's social account, the user's level, the dynamics published by the user on the social platform, and so on.

[0040] In one embodiment, the scenario service recommendation device calls the target intent classification model corresponding to the target business information to perform intent prediction on the target historical behavior information and the user profile information, so as to determine the target intent from multiple candidate intents corresponding to the target business information, including: preprocessing the target historical behavior information and the user profile information to obtain feature vectors, calling the target intent classification model to process the feature vectors, obtaining the confidence levels of multiple candidate intents, determining the highest confidence level from the confidence levels of multiple candidate intents, and determining the candidate intent corresponding to the highest confidence level as the target intent.

[0041] Optionally, the scenario service recommendation device constructs a basic vector based on the target historical behavior information, constructs a user attribute vector based on the user profile information, fuses the basic vector and the user attribute vector to obtain an intermediate vector, and invokes a feature extraction model to process the intermediate vector to obtain a feature vector.

[0042] In one embodiment, the scenario service recommendation device can use an encoding method to perform encoding processing on the target historical behavior information to obtain a basic vector, and perform encoding processing on the user profile information to obtain a user attribute vector. Among them, the encoding method can be one-hot encoding (One-Hot encoding), etc.

[0043] In one embodiment, the scenario service recommendation device can directly splice the encoded basic vector and user attribute vector to obtain an intermediate vector. In another embodiment, the scenario service recommendation device can also fuse the basic vector and the user attribute vector based on the attention mechanism to obtain an intermediate vector. Specifically, the scenario service recommendation device can obtain the attention weight corresponding to the basic vector and the attention weight corresponding to the user attribute vector, and fuse the basic vector and the user attribute vector based on the attention weight corresponding to the basic vector and the attention weight corresponding to the user attribute vector to obtain an intermediate vector. Among them, the attention mechanism means that the attention can be concentrated on the actually important features through the attention weight. For example, when paying more attention to the target historical behavior information, the attention weight of the basic vector constructed by the target historical behavior information can be set to be greater than the attention weight of the user attribute vector constructed by the user profile information. Another example is that when paying more attention to the user profile information, the attention weight of the user attribute vector constructed by the user profile information can be set to be greater than the attention weight of the basic vector constructed by the target historical behavior information.

[0044] In one embodiment, the feature extraction model can be a convolutional neural network. That is, a convolutional neural network can be used to extract features from the intermediate vector to obtain a feature vector. Specifically, the intermediate vector can be divided into two-dimensional matrices of the same shape and input into the convolutional neural network. The convolutional neural network can perform sliding convolution on multiple two-dimensional matrices to obtain a feature vector. Among them, the convolutional neural network can have multiple convolutional layers. For example, the convolutional neural network includes four convolutional layers of 1ⅹ3, 1ⅹ4, 1ⅹ5, and 1ⅹ6.

[0045] Among them, the intent classification model can be a model trained based on machine learning algorithms. Among them, machine learning algorithms can include, but are not limited to, one or more of the following: Decision Tree (DT) algorithm, Rocchio algorithm, Xtreme Gradient Boosting (XGBooste) algorithm, Naive Bayes (NB) algorithm, Linear Discriminant Analysis (LDA), Support Vector Machine (SVM) algorithm, Random Forest (RF) algorithm, and Logistic Regression (LR) algorithm. For example, a decision tree model trained based on the decision tree algorithm, a Rocchio model trained based on the Rocchio algorithm, an XGBooste model trained based on the extreme gradient boosting algorithm, an NB model trained based on the naive bayes algorithm, an LDA model trained based on linear discriminant analysis, a support vector machine model trained based on the support vector machine algorithm, a random forest model trained based on the random forest, and a logistic regression model trained based on the logistic regression algorithm. And so on.

[0046] Among them, the confidence level can include the prediction probability. The higher the prediction probability, the higher the confidence level that the feature vector belongs to the candidate intent. The lower the prediction probability, the lower the confidence level that the feature vector belongs to the candidate intent.

[0047] In one embodiment, the scenario service recommendation system can support multiple services, such as payment services, investment and financial management, and credit cards in the financial field. For another example, medical treatment services in the medical field, and so on. For each service, the scenario services involved in the service are different. To effectively improve the accuracy of scenario service recommendation, different intent classification models can be trained for different services. When determining the target service information based on the target historical behavior information, the target intent classification model corresponding to the target service information can be directly called to perform intent prediction on the target historical behavior information and the user profile information.

[0048] In one embodiment, before calling the target intent classification model to perform intent prediction on the target historical behavior information and the user profile information, the initial model needs to be trained to obtain the target intent classification model. The training process of the intent classification model includes s11 - s12.

[0049] s11: Obtain the training sample pairs in the target service; the training sample pairs include training samples and the annotation information of the training samples.

[0050] Among them, the training samples include historical behavior information and user profile information, and the annotation information of the training samples includes the benchmark intention.

[0051] S12: Input the training sample pairs in the target service into the initial model for training to obtain a target intention classification model.

[0052] Specifically, for any training sample in the target service, the training sample can be input into the initial model to obtain the predicted intention of the training sample; the loss function corresponding to the initial model is determined according to the annotation information (benchmark intention) of the training sample and the predicted intention of the training sample, and the initial model is trained using this loss function to obtain a target intention classification model.

[0053] S203: Obtain at least one scenario service associated with the target intention based on the knowledge graph; wherein, the knowledge graph includes the association relationship between each candidate intention and the scenario service.

[0054] Among them, the scenario service can refer to the business process nodes set based on user needs. For example, the "study abroad foreign exchange purchase" business process can include business process nodes such as material pre-review, submission of foreign exchange purchase application form, and offline network appointment. Since the number of scenario services on the business platform is large, it is not convenient to directly construct a classification model. Moreover, some scenario services have strong correlations, so at least one scenario service with similar characteristics can be associated with a candidate intention, so that the target intention can be determined first, and then the scenario service associated with the target intention can be obtained.

[0055] In one embodiment, the candidate intention can be associated with the scenario service through the scenario service management interface. Specifically, the scenario service management interface can be displayed, and the scenario service management interface includes an intention input column and a scenario service input column; input the intention in the intention input column and input the scenario service in the scenario service input column, so that the intention input in the intention input column is associated with the scenario service input in the scenario service input column.

[0056] In another embodiment, business data can be obtained, and the business data includes scenario services, attributes of the scenario services, and the association relationship between the scenario services, and a knowledge graph is determined based on the scenario services, attributes of the scenario services, and the association relationship between the scenario services. Among them, the knowledge graph includes the association relationship between each candidate intention and the scenario service. Specifically, various scenario services can be classified into multiple scenario service sets according to the attributes of the scenario services and the association relationship between the scenario services, and each scenario service set is associated with each candidate intention. The scenario service recommendation device can obtain at least one scenario service in the scenario service set associated with the target intention based on the knowledge graph.

[0057] Among them, a knowledge graph is a graph organization form that associates various entities or concepts existing in the real world through semantics, mainly formed by nodes, edges, and node attributes. Among them, each entity or concept is identified by a globally unique and determined code, each attribute-value pair is used to describe the internal attributes of the entity or concept, and the edge is used to connect two entities or concepts to describe the association between them. In the embodiments of the present application, the knowledge graph can be a scenario service knowledge graph.

[0058] S204: Determine a target scenario service from at least one scenario service and output the target scenario service.

[0059] In one embodiment, at least one scenario service can be directly determined as the target scenario service. In another embodiment, a selection instruction for at least one scenario service can also be obtained, and the target scenario service is determined from at least one scenario service based on the selection instruction. In still another embodiment, the relevance between at least one scenario service and the target business information can be obtained respectively, and the target scenario service is determined according to the relevance between at least one scenario service and the target business information. The relevance between the target scenario service and the target business information is greater than a preset threshold, or the target scenario service is the scenario service with the highest relevance among at least one scenario service. The target scenario service with a higher matching degree with the target business information can be screened out from at least one scenario service to improve the accuracy of the target scenario service.

[0060] In one embodiment, the target scenario service can be directly output through a scenario service recommendation device. After the target scenario service is determined, the target scenario service can also be sent to the terminal device associated with the user, so that the terminal device associated with the user outputs the target scenario service.

[0061] Furthermore, each scenario service can also correspond to a component, and the component includes the execution instruction of the scenario service and the description information of the scenario service. Specifically, when determining the target scenario service, the scenario service recommendation device can obtain the target component corresponding to the target scenario service and output the target component corresponding to the target scenario service while outputting the target scenario service.

[0062] In the embodiments of the present application, the scenario service recommendation device may determine target service information based on target historical behavior information, and call a target intent classification model corresponding to the target service information to perform intent prediction on the target historical behavior information and user profile information, so as to determine a target intent from multiple candidate intents corresponding to the target service information, and obtain at least one scenario service associated with the target intent based on a knowledge graph, and further determine a target scenario service. In order to obtain the target scenario service, the target intent classification model may be called first to perform intent prediction on the target historical behavior information and user profile information to determine the target intent from multiple candidate intents, and then the scenario services associated with the target intent may be obtained based on the knowledge graph, and the target scenario service to be pushed may be determined. It can adapt to user needs, intelligently output the target scenario service to the user, and improve the user experience. Moreover, in the present application, the target intent classification model is called first to determine the target intent, and then at least one scenario service associated with the target intent is determined according to the target intent, which can avoid too many candidate items in the classification task, avoid the long-tail phenomenon, and is easier to maintain. In addition, since different intent classification models are trained for different service information, more accurate target intents can be obtained through the target intent classification model corresponding to the target service information, so that the target scenario service determined according to at least one scenario service associated with the target intent is more accurate. It can effectively improve the accuracy of the target scenario service and enhance user stickiness.

[0063] Referring to the above Figure 2 description of the relevant method embodiments shown, Figure 2 the scenario service recommendation method shown may call a target intent classification model corresponding to the target service information to perform intent prediction on the target historical behavior information and user profile information. Further, in order to more comprehensively represent the behavior characteristics of the target user on each page, the target historical behavior information may be determined based on long-term first historical behavior information and short-term second historical behavior information. Based on this, the embodiments of the present application propose a schematic flowchart of another scenario service recommendation method. As Figure 3 shown, the schematic flowchart of the scenario service recommendation method includes S301-S306:

[0064] S301: Obtain the first historical behavior information of the target user within a first time period, and obtain the second historical behavior information of the target user within a second time period; the duration of the first time period is greater than the duration of the second time period.

[0065] Specifically, the scenario service recommendation device can obtain the first historical behavior information of the target user within the first time period, and obtain the second historical behavior information of the target user within the second time period, and determine the target historical behavior information according to the first historical behavior information and the second historical behavior information. Among them, the duration of the first time period is greater than the duration of the second time period. For example, the first time period can be the day closest to the current time, and the second time period can be the week closest to the current time. The longer duration of the first time period can capture the long-term behavior characteristics of the target user. That is, the first historical behavior information can include, but is not limited to, the number of views of each page, the viewing duration of each page, and the number of clicks on each page within the first time period. The shorter duration of the second time period can capture the short-term behavior characteristics of the target user. That is, the second historical behavior information can include, but is not limited to, the number of views of each page, the viewing duration of each page, and the number of clicks on each page within the second time period.

[0066] S302: Determine the target historical behavior information according to the first historical behavior information and the second historical behavior information.

[0067] Among them, determining the target historical behavior information of the target user according to the first historical behavior information and the second historical behavior information can include various forms. Optionally, the target historical behavior information can be determined based on the attention mechanism. Specifically, the attention weights corresponding to the first historical behavior information and the second historical behavior information can be obtained respectively, and the first historical behavior information and the second historical behavior information can be processed based on the attention weights corresponding to the first historical behavior information and the second historical behavior information to obtain the target historical behavior information. Among them, the attention mechanism means that the attention can be focused on the actually important features through the attention weights. For example, when more attention is paid to the long-term behavior characteristics, the attention weight of the first historical behavior information can be set to be greater than the attention weight of the second historical behavior information. Another example is that when more attention is paid to the short-term behavior characteristics, the attention weight of the second historical behavior information can be set to be greater than the attention weight of the first historical behavior information.

[0068] Optionally, the first historical behavior information and the second historical behavior information can also be directly added to obtain the target historical behavior information. For example, assume that the first historical behavior information includes 1 view in page A, a viewing duration of 100 seconds in page A, and 1 click in page A; 1 view in page B, a viewing duration of 200 seconds; 1 view in page C, a viewing duration of 80 seconds. The second historical behavior information can include 1 click in page A; 5 clicks in page B; 2 views in page C, with a viewing duration of 20 seconds. Then, the first historical behavior information and the second historical behavior information can be directly added to obtain the target historical behavior information as follows: 1 view in page A, a viewing duration of 100 seconds, and 2 clicks in page A; 1 view in page B, a viewing duration of 200 seconds, and 5 clicks in page B; 3 views in page C, a viewing duration of 100 seconds.

[0069] S303: Determine the target service information based on the target historical behavior information.

[0070] S304: Obtain the user profile information of the target user, and call the target intent classification model corresponding to the target service information to perform intent prediction on the target historical behavior information and the user profile information, so as to determine the target intent from multiple candidate intents corresponding to the target service information.

[0071] S305: Obtain at least one scenario service associated with the target intent based on the knowledge graph; the knowledge graph includes the association relationship between each candidate intent and the scenario service.

[0072] S306: Determine the target scenario service from at least one scenario service and output the target scenario service.

[0073] It should be noted that the specific implementation process of S303 - S306 can be referred to Figure 2 the specific description of the relevant embodiments, which will not be elaborated here.

[0074] In one embodiment, the target scenario service can be written into the blockchain, so that the target scenario service can be directly output later. Specifically, the scenario service recommendation device can verify the target scenario service. If the verification passes, the consensus node in the blockchain network will perform consensus verification on the target scenario service; if the consensus verification passes, the target scenario service will be encapsulated into a block and the block will be written into the blockchain.

[0075] Among them, blockchain is a chain - type data structure formed by combining data blocks in sequence according to the time sequence, and is a distributed ledger that ensures the data cannot be tampered with and forged in a cryptographic manner. Multiple independent distributed nodes store the same records. Blockchain technology has achieved decentralization and has become the cornerstone for the storage, transfer, and trading of trustworthy digital assets.

[0076] In the embodiment of the present application, the scenario service recommendation device can determine the target historical behavior information based on the long - term first historical behavior information and the short - term second historical behavior information, then determine the target business information according to the target historical behavior information, and call the target intention classification model corresponding to the target business information to perform intention prediction on the target historical behavior information and the user portrait information, so as to determine the target intention from multiple candidate intentions corresponding to the target business information, and obtain at least one scenario service associated with the target intention based on the knowledge graph, and further determine the target scenario service. Since the target historical behavior information is determined based on the first historical behavior information and the second historical behavior information, comprehensively considering the long - term characteristics and short - term characteristics of the user, it can more comprehensively represent the behavior characteristics of the target user on each page, so that the target intention obtained by performing intention prediction on the target historical behavior information and the user portrait information by calling the target intention classification model is more accurate, and thus the target scenario service determined from at least one scenario service associated with the target intention is also more accurate.

[0077] The embodiment of the present application also discloses a scenario service recommendation device, and the scenario service recommendation device can be a computer program (including program code) running in one of the above - mentioned scenario service recommendation devices. This scenario service recommendation device can execute Figure 2 or Figure 3 the method shown. Please refer to Figure 4 and this scenario service recommendation device can run the following units:

[0078] A determination unit 401, configured to obtain the target historical behavior information of the target user, and determine the target business information according to the target historical behavior information;

[0079] An intention prediction unit 402, configured to obtain the user portrait information of the target user, and call the target intention classification model corresponding to the target business information to perform intention prediction on the target historical behavior information and the user portrait information, so as to determine the target intention from multiple candidate intentions corresponding to the target business information;

[0080] An acquisition unit 403, configured to obtain at least one scenario service associated with the target intention based on the knowledge graph; the knowledge graph includes the association relationship between each candidate intention and the scenario service;

[0081] The determination unit 401 is further configured to determine a target scenario service from at least one scenario service and output the target scenario service.

[0082] In a feasible implementation manner, the intent prediction unit 402 is configured to call a target intent classification model corresponding to the target service information to perform intent prediction on the target historical behavior information and the user profile information, so as to determine a target intent from multiple candidate intents corresponding to the target service information, including:

[0083] Preprocess the target historical behavior information and the user profile information to obtain feature vectors;

[0084] Call the target intent classification model to process the feature vectors to obtain the confidence levels of multiple candidate intents;

[0085] Determine the highest confidence level from the confidence levels of multiple candidate intents, and determine the candidate intent corresponding to the highest confidence level as the target intent.

[0086] In a feasible implementation manner, the intent prediction unit 402 is configured to preprocess the target historical behavior information and the user profile information to obtain feature vectors, including:

[0087] Construct a basic vector based on the target historical behavior information and construct a user attribute vector based on the user profile information;

[0088] Fuse the basic vector and the user attribute vector to obtain an intermediate vector;

[0089] Call a feature extraction model to process the intermediate vector to obtain feature vectors.

[0090] In a feasible implementation manner, before the acquisition unit 403 is configured to acquire at least one scenario service having an association relationship with the target intent based on the knowledge graph, the acquisition unit 403 is further configured to:

[0091] Acquire service data, where the service data includes scenario services, attributes of scenario services, and association relationships between scenario services;

[0092] Determine a knowledge graph based on the scenario services, attributes of scenario services, and association relationships between scenario services.

[0093] In a feasible implementation manner, the determination unit 401 is configured to determine a target scenario service from at least one scenario service, including:

[0094] Respectively acquire the relevance between at least one scenario service and the target service information;

[0095] Determine the target scenario service according to the relevance between at least one scenario service and the target service information.

[0096] In a feasible implementation manner, the determining unit 401 is configured to obtain target historical behavior information of a target user, including:

[0097] Obtain first historical behavior information of the target user within a first time period, and obtain second historical behavior information of the target user within a second time period; the duration of the first time period is greater than the duration of the second time period;

[0098] Determine the target historical behavior information according to the first historical behavior information and the second historical behavior information.

[0099] In a feasible implementation manner, the determining unit 401 is configured to determine the target historical behavior information according to the first historical behavior information and the second historical behavior information, including:

[0100] Obtain the attention weight corresponding to the first historical behavior information and the attention weight corresponding to the second historical behavior information respectively;

[0101] Based on the attention weight corresponding to the first historical behavior information and the attention weight corresponding to the second historical behavior information, process the first historical behavior information and the second historical behavior information to obtain the target historical behavior information.

[0102] According to another embodiment of the present application, Figure 4 Each unit in the scene service recommendation device shown can be separately or all combined into one or several other units to form, or some of the units can be further split into multiple smaller units with more functions to form, which can achieve the same operation without affecting the realization of the technical effects of the embodiments of the present application. The above units are divided based on logical functions. In actual applications, the function of one unit can also be realized by multiple units, or the functions of multiple units can be realized by one unit. In other embodiments of the present application, the scene service recommendation device can also include other units. In actual applications, these functions can also be assisted by other units and can be realized by the cooperation of multiple units.

[0103] According to another embodiment of the present application, it can include processing elements and storage elements such as a central processing unit (CPU), a random access storage medium (RAM), and a read-only storage medium (ROM). For example, on a general computing device of a computer, a computer program (including program code) that can execute the respective steps involved in the corresponding method shown in Figure 2 or Figure 3 is run to construct such as Figure 4The described scenario service recommendation device and the scenario service recommendation method of the embodiments of the present application are implemented. The computer program can be recorded on, for example, a computer-readable recording medium, loaded into the above scenario service recommendation device through the computer-readable recording medium, and run therein.

[0104] In the embodiments of the present application, the scenario service recommendation device can determine target service information based on target historical behavior information, and call the target intent classification model corresponding to the target service information to perform intent prediction on the target historical behavior information and user profile information, so as to determine a target intent from multiple candidate intents corresponding to the target service information, and obtain at least one scenario service associated with the target intent based on the knowledge graph, and then determine the target scenario service. In order to obtain the target scenario service, the target intent classification model can be called first to perform intent prediction on the target historical behavior information and user profile information to determine the target intent from multiple candidate intents, and then the scenario services associated with the target intent are obtained based on the knowledge graph, and the target scenario service to be pushed is determined. It can adapt to user needs, intelligently output the target scenario service to the user, and improve the user experience. Moreover, in the present application, the target intent classification model is called first to determine the target intent, and then at least one scenario service associated with the target intent is determined according to the target intent, which can avoid too many candidate items in the classification task, avoid the long-tail phenomenon, and is easier to maintain. In addition, since different intent classification models are trained for different service information, more accurate target intents can be obtained through the target intent classification model corresponding to the target service information, so that the target scenario service determined according to at least one scenario service associated with the target intent is more accurate. It can effectively improve the accuracy of the target scenario service and enhance user stickiness.

[0105] Based on the description of the embodiments of the above scenario service recommendation method, the embodiments of the present application also disclose a scenario service recommendation device. Please refer to Figure 5 This scenario service recommendation device at least includes a processor 501, an input interface 502, an output interface 503, and a computer storage medium 504, which can be connected through a bus or other means.

[0106] The computer storage medium 504 is a memory device in the scenario service recommendation device, used to store programs and data. It can be understood that the computer storage medium 504 here can include both the built-in storage medium of the scenario service recommendation device and, of course, the extended storage medium supported by the scenario service recommendation device. The computer storage medium 504 provides a storage space, and the operating system of the scenario service recommendation device is stored in this storage space. Moreover, one or more instructions suitable for being loaded and executed by the processor 501 are stored in this storage space, and these instructions can be one or more computer programs (including program codes). It should be noted that the computer storage medium here can be a high-speed RAM memory; optionally, it can also be at least one computer storage medium far from the aforementioned processor. The processor can be called a Central Processing Unit (CPU), which is the core and control center of the scenario service recommendation device, suitable for implementing one or more instructions, specifically loading and executing one or more instructions to implement the corresponding method flow or function.

[0107] In one embodiment, one or more instructions stored in the computer storage medium 504 can be loaded and executed by the processor 501 to implement the execution of each step involved in the corresponding method as shown in Figure 2 or Figure 3 In the specific implementation, one or more instructions in the computer storage medium 504 are loaded and executed by the processor 501 to perform the following steps:

[0108] Obtain the target historical behavior information of the target user, and determine the target business information according to the target historical behavior information;

[0109] Obtain the user profile information of the target user, and call the target intent classification model corresponding to the target business information to perform intent prediction on the target historical behavior information and the user profile information, so as to determine the target intent from multiple candidate intents corresponding to the target business information;

[0110] Based on the knowledge graph, obtain at least one scenario service associated with the target intent; the knowledge graph includes the association relationship between each candidate intent and the scenario service;

[0111] Determine the target scenario service from at least one scenario service, and output the target scenario service.

[0112] In a feasible implementation manner, the processor 501 is used to call the target intent classification model corresponding to the target business information to perform intent prediction on the target historical behavior information and the user profile information, so as to determine the target intent from multiple candidate intents corresponding to the target business information, including:

[0113] Preprocess the target historical behavior information and user profile information to obtain feature vectors;

[0114] Call the target intent classification model to process the feature vectors to obtain the confidence levels of multiple candidate intents;

[0115] Determine the highest confidence level from the confidence levels of multiple candidate intents, and determine the candidate intent corresponding to the highest confidence level as the target intent.

[0116] In a feasible implementation, the processor 501 is used to preprocess the target historical behavior information and user profile information to obtain feature vectors, including:

[0117] Construct a basic vector based on the target historical behavior information, and construct a user attribute vector based on the user profile information;

[0118] Fuse the basic vector and the user attribute vector to obtain an intermediate vector;

[0119] Call the feature extraction model to process the intermediate vector to obtain feature vectors.

[0120] In a feasible implementation, before the processor 501 is used to obtain at least one scenario service associated with the target intent based on the knowledge graph, the processor 501 is further used to:

[0121] Obtain business data, which includes scenario services, attributes of scenario services, and association relationships between scenario services;

[0122] Determine the knowledge graph based on the scenario services, attributes of scenario services, and association relationships between scenario services.

[0123] In a feasible implementation, the processor 501 is used to determine the target scenario service from at least one scenario service, including:

[0124] Respectively obtain the relevance degrees of at least one scenario service to the target business information;

[0125] Determine the target scenario service according to the relevance degrees of at least one scenario service to the target business information.

[0126] In a feasible implementation, the processor 501 is used to obtain the target historical behavior information of the target user, including:

[0127] Obtain the first historical behavior information of the target user in the first time period, and obtain the second historical behavior information of the target user in the second time period; the duration of the first time period is greater than the duration of the second time period;

[0128] Determine the target historical behavior information according to the first historical behavior information and the second historical behavior information.

[0129] In a feasible embodiment, the processor 501 is configured to determine target historical behavior information according to the first historical behavior information and the second historical behavior information, including:

[0130] Obtain the attention weight corresponding to the first historical behavior information and the attention weight corresponding to the second historical behavior information respectively;

[0131] Based on the attention weight corresponding to the first historical behavior information and the attention weight corresponding to the second historical behavior information, process the first historical behavior information and the second historical behavior information to obtain the target historical behavior information.

[0132] In the embodiments of the present application, the scenario service recommendation device may determine target service information according to the target historical behavior information, and call the target intent classification model corresponding to the target service information to perform intent prediction on the target historical behavior information and the user profile information, so as to determine the target intent from multiple candidate intents corresponding to the target service information, and obtain at least one scenario service associated with the target intent based on the knowledge graph, and then determine the target scenario service. In order to obtain the target scenario service, the target intent classification model may be called first to perform intent prediction on the target historical behavior information and the user profile information to determine the target intent from multiple candidate intents, and then the scenario services associated with the target intent are obtained based on the knowledge graph, and the target scenario service to be pushed is determined. It can adapt to the user's needs and intelligently output the target scenario service to the user, improving the user experience. Moreover, in the present application, the target intent classification model is called first to determine the target intent, and then at least one scenario service associated with the target intent is determined according to the target intent, which can avoid too many candidate items in the classification task, avoid the long-tail phenomenon, and is easier to maintain. In addition, since different intent classification models are trained for different service information, more accurate target intents can be obtained by performing intent prediction through the target intent classification model corresponding to the target service information, so that the target scenario service determined according to at least one scenario service associated with the target intent is more accurate. It can effectively improve the accuracy of the target scenario service and enhance user stickiness.

[0133] It should be noted that the embodiments of the present application also provide a computer program product or a computer program. The computer program product or the computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. The processor of the scenario service recommendation device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the scenario service recommendation device executes the steps performed in the above-mentioned embodiment of the scenario service recommendation method Figure 2 or Figure 3 the steps executed in

[0134] The above-disclosed is only a preferred embodiment of the present application. Of course, it cannot be used to limit the scope of rights of the present application. Those of ordinary skill in the art can understand all or part of the processes of implementing the above embodiments, and the equivalent changes made according to the claims of the present application still fall within the scope covered by the application.

Claims

1. A method for recommending scenario services, characterized in that, the method includes: Obtain the target historical behavior information of the target user, and determine the target business information according to the target historical behavior information; the target historical behavior information is used to indicate the behavior characteristics of the target user on the pages in the business field; the acquisition method of the target historical behavior information is: obtain the first historical behavior information of the target user within the first time period, and obtain the second historical behavior information of the target user within the second time period; the duration of the first time period is greater than the duration of the second time period; respectively obtain the attention weight corresponding to the first historical behavior information and the attention weight corresponding to the second historical behavior information, if the long-term behavior characteristics are more concerned in the business field, then set the attention weight of the first historical behavior information to be greater than the attention weight of the second historical behavior information, if the short-term behavior characteristics are more concerned in the business field, then set the attention weight of the second historical behavior information to be greater than the attention weight of the first historical behavior information; based on the attention weight corresponding to the first historical behavior information and the attention weight corresponding to the second historical behavior information, process the first historical behavior information and the second historical behavior information to obtain the target historical behavior information; Obtain the user portrait information of the target user, and call the target intention classification model corresponding to the target business information to perform intention prediction on the target historical behavior information and the user portrait information, so as to determine the target intention from multiple candidate intentions corresponding to the target business information; the target intention classification model is the intention classification model corresponding to the target business information among multiple intention classification models, and each intention classification model corresponds to a kind of business, and the businesses include payment business, investment and financial management business, and credit card business in the financial field, and medical treatment business in the medical field; Obtain at least one scenario service associated with the target intention based on the knowledge graph; the knowledge graph includes the association relationship between each candidate intention and the scenario service; Determine the target scenario service from the at least one scenario service, and output the target scenario service.

2. The method according to claim 1, characterized in that, the step of calling the target intention classification model corresponding to the target business information to perform intention prediction on the target historical behavior information and the user portrait information, so as to determine the target intention from multiple candidate intentions corresponding to the target business information, includes: Preprocess the target historical behavior information and the user portrait information to obtain feature vectors; Call the target intention classification model to process the feature vectors to obtain the confidence levels of the multiple candidate intentions; Determine the highest confidence level from the confidence levels of the multiple candidate intentions, and determine the candidate intention corresponding to the highest confidence level as the target intention.

3. The method according to claim 2, characterized in that, the step of preprocessing the target historical behavior information and the user portrait information to obtain feature vectors includes: Construct a basic vector based on the target historical behavior information and construct a user attribute vector based on the user profile information; Fuse the basic vector and the user attribute vector to obtain an intermediate vector; Call a feature extraction model to process the intermediate vector to obtain the feature vector.

4. The method according to claim 1, wherein, before obtaining at least one scenario service associated with the target intention based on the knowledge graph, the method further includes: Obtain business data, where the business data includes scenario services, attributes of the scenario services, and association relationships between the scenario services; Determine the knowledge graph based on the scenario services, attributes of the scenario services, and association relationships between the scenario services.

5. The method according to claim 1, wherein, determining the target scenario service from the at least one scenario service includes: Obtain the relevance of each of the at least one scenario service to the target business information respectively; Determine the target scenario service according to the relevance of the at least one scenario service to the target business information.

6. A scenario service recommendation device, wherein, the device includes: A determination unit, configured to obtain the target historical behavior information of a target user and determine target business information according to the target historical behavior information; the target historical behavior information is used to indicate the behavior characteristics of the target user on the pages in the business field; the acquisition method of the target historical behavior information is: obtain the first historical behavior information of the target user in the first time period and obtain the second historical behavior information of the target user in the second time period; the duration of the first time period is greater than the duration of the second time period; obtain the attention weight corresponding to the first historical behavior information and the attention weight corresponding to the second historical behavior information respectively. If the business field pays more attention to long-term behavior characteristics, set the attention weight of the first historical behavior information to be greater than the attention weight of the second historical behavior information. If the business field pays more attention to short-term behavior characteristics, set the attention weight of the second historical behavior information to be greater than the attention weight of the first historical behavior information; process the first historical behavior information and the second historical behavior information based on the attention weight corresponding to the first historical behavior information and the attention weight corresponding to the second historical behavior information to obtain the target historical behavior information; An intention prediction unit, configured to obtain the user profile information of the target user and call the target intention classification model corresponding to the target business information to perform intention prediction on the target historical behavior information and the user profile information, so as to determine a target intention from multiple candidate intentions corresponding to the target business information; the target intention classification model is the intention classification model corresponding to the target business information among multiple intention classification models, and each intention classification model corresponds to a kind of business, and the businesses include payment business, investment and financial management business, and credit card business in the financial field and medical treatment business in the medical field; An acquisition unit, configured to acquire at least one scenario service associated with the target intent based on a knowledge graph; the knowledge graph includes the association relationships between each candidate intent and scenario services; The determination unit is further configured to determine a target scenario service from the at least one scenario service and output the target scenario service.

7. A scenario service recommendation device Characterized in that it includes a processor and a memory, wherein the memory is used to store a computer program, the computer program includes a program, and the processor is configured to call the program to execute the scenario service recommendation method according to any one of claims 1-5.

8. A computer-readable storage medium Characterized in that the computer-readable storage medium stores a computer program, the computer program includes program instructions, and when the program instructions are executed by a processor, the processor is caused to execute the scenario service recommendation method according to any one of claims 1-5.

Images