Information pushing method, device and electronic equipment

Abstract

This application discloses an information push method, an information push device, and an electronic device. The method includes: determining at least one intention of the interactive object based on interaction data with the interactive object; analyzing the interaction data to determine a matching degree parameter of the at least one intention; determining the role information of the interactive object; determining a push strategy for data to be pushed based on the matching degree parameter and the role information of the interactive object; and pushing the data to be pushed using the push strategy.

Description

Technical Field

[0001] This application relates to the field of push technology, and in particular to an information push method, information push device and electronic device. Background Technology

[0002] In related technologies, push systems typically push surveys based on certain trigger signals. For example, when an order is delivered, the system generates a trigger signal to push the survey based on the delivery person's click on the "Complete Delivery" button on the order page. Responding to this signal, the survey is pushed. It can be understood that the trigger signal is generated at a fixed delivery point (the click of the delivery button). This survey push scheme can be considered a scheme that pushes surveys based on a fixed trigger point. This scheme is relatively fixed and lacks flexibility. Summary of the Invention

[0003] This application provides an information push method, an information push device, and an electronic device to at least solve the above-mentioned technical problems existing in the prior art.

[0004] According to a first aspect of this application, an information push method is provided, comprising:

[0005] Based on the interaction data with the interaction object, determine at least one intention of the interaction object;

[0006] Analyze the interaction data to determine the matching degree parameter of the at least one intent;

[0007] Determine the role information of the interactive object;

[0008] Based on the matching degree parameter and the role information of the interaction object, determine the push strategy for the data to be pushed;

[0009] The push strategy is used to push the data to be pushed.

[0010] In one possible implementation, analyzing the interaction data to determine the matching degree parameter of the at least one intent includes:

[0011] Obtain target parameters for the at least one intention, wherein the target parameters for the at least one intention characterize a reference probability that the at least one intention will change;

[0012] The at least one intent, the target parameters of the at least one intent, and the first profile data of the interactive object are input into the first prediction model to obtain a target prediction result. The target prediction result indicates whether the at least one intent output generated by the interactive object in the interaction data is consistent with the target prediction result.

[0013] At least one prediction result for matching data corresponding to an intent;

[0014] Based on the target prediction result, a matching degree parameter for the at least one intent is determined.

[0015] In one possible implementation, determining the role information of the interactive object includes:

[0016] Based on the interaction data, the role information of the interaction object is determined.

[0017] In one possible implementation, the step of determining the role information of the interactive object based on the interaction data...

[0018] Information, including:

[0019] The interaction data, the at least one intent, and the second profile data of the interaction object are input into the second prediction model to obtain the role information of the interaction object.

[0020] In one possible implementation, obtaining the target parameter of the at least one intent includes: 5. Obtaining a historical reference value of the at least one intent based on historical interaction data, wherein the historical reference value is...

[0021] Historical reference values ​​represent the probability that at least one stated intention has changed in history;

[0022] The historical reference value of the at least one intention is used as the target parameter of the at least one intention.

[0023] In one possible implementation, obtaining the at least one intention based on historical interaction data...

[0024] The historical reference values ​​of the graph include: 0 for any intent among the at least one intent;

[0025] From historical interaction data, a first type of target event and a second type of target event are identified. The first type of target event represents an event in the historical interaction data where the arbitrary intention changes to a first type of optional intention; the second type of target event represents an event in the historical interaction data where the arbitrary intention changes to a first type of optional intention.

[0026] Events leading to the second type of optional intent;

[0027] 5. Based on the reference attributes of the first type of target event and the second type of target event in the historical interaction data, obtain the historical reference value of the arbitrary intention;

[0028] The reference attribute represents the number of times or frequency of occurrence in the historical interaction data.

[0029] In one possible implementation, the data to be pushed is determined based on the role information of the interactive object.

[0030] In one possible implementation, determining the data to be pushed based on the role information of the interactive object includes:

[0031] When the interaction object is a positive role, the first data is determined to be the data to be pushed;

[0032] When the interaction object is a non-positive role, second data is determined to be the data to be pushed, wherein the second data is data generated based on the interaction data. According to a second aspect of this application, an information push device is provided, comprising:

[0033] The first determining unit is configured to determine at least one intention of the interactive object based on interaction data with the interactive object;

[0034] The second determining unit is used to analyze the interaction data to determine the matching degree parameter of the at least one intention;

[0035] The third determining unit is used to determine the role information of the interactive object;

[0036] The fourth determining unit is used to determine the push strategy for the data to be pushed based on the matching degree parameter and the role information of the interaction object;

[0037] The push unit is used to push the data to be pushed using the push strategy.

[0038] According to a third aspect of this application, an electronic device is provided, comprising:

[0039] At least one processor; and

[0040] A memory communicatively connected to the at least one processor; wherein,

[0041] The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method described in this application.

[0042] According to a fourth aspect of this application, a non-transitory computer-readable storage medium is provided storing computer instructions for causing the computer to perform the methods described in this application.

[0043] The present application discloses an information push method, apparatus, and electronic device storage medium, wherein the method includes: determining at least one intent of the interactive object based on interaction data with the interactive object; analyzing the interaction data to determine a matching degree parameter of the at least one intent; determining the role information of the interactive object; determining a push strategy for data to be pushed based on the matching degree parameter and the role information of the interactive object; and pushing the data to be pushed using the push strategy.

[0044] The technical solution of this application determines the push strategy based on the matching degree parameter and the role, and can push the information to be pushed at the appropriate time, which has the flexibility of push.

[0045] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent from the following description. Attached Figure Description

[0046] The above and other objects, features, and advantages of exemplary embodiments of this application will become readily apparent from the following detailed description taken in conjunction with the accompanying drawings. Several embodiments of this application are illustrated in the drawings by way of example and not limitation, wherein:

[0047] In the accompanying drawings, the same or corresponding reference numerals indicate the same or corresponding parts.

[0048] Figure 1 This document illustrates the implementation flow of the information push method in an embodiment of this application. Figure 1 ;

[0049] Figure 2 This document illustrates the implementation flow of the information push method in an embodiment of this application. Figure 2 ;

[0050] Figure 3 A schematic diagram of the input of the first prediction model in an embodiment of this application is shown;

[0051] Figure 4 A schematic diagram of the input of the second prediction model in an embodiment of this application is shown;

[0052] Figure 5 This document illustrates the implementation flow of the information push method in an embodiment of this application. Figure 3 ;

[0053] Figure 6 A schematic diagram illustrating the implementation flow of the information push method in an embodiment of this application is shown;

[0054] Figure 7 A schematic diagram of the composition structure of the information push device in an embodiment of this application is shown;

[0055] Figure 8 A schematic diagram of the composition structure of the electronic device in an embodiment of this application is shown. Detailed Implementation

[0056] To make the objectives, features, and advantages of this application more apparent and understandable, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0057] To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings. The described embodiments should not be regarded as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0058] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0059] In the following description, the terms "first" and "second" are used merely to distinguish similar objects and do not represent a specific ordering of objects. It is understood that "first" and "second" may be interchanged in a specific order or sequence where permitted, so that the embodiments of this application described herein can be implemented in an order other than that illustrated or described herein.

[0060] 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 this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0061] It should be understood that in the various embodiments of this application, the sequence number of each implementation process does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0062] The processing logic of the information push method in this application embodiment can be deployed on any reasonable device. This device can be a server or a terminal. The server includes ordinary servers, cloud servers, server clusters, and servers used in specialized fields, such as customer service servers. The terminal includes, but is not limited to, tablet computers, all-in-one computers, desktop computers, etc.

[0063] In some preferred embodiments, the processing logic of the information push method of this application can be deployed in a customer service system, such as an intelligent customer service system, so as to push customer service questionnaires at appropriate times, thereby achieving flexible push of customer service questionnaires. This satisfies the user's customer service experience and highlights the intelligence and flexibility of the customer service system.

[0064] Figure 1 This document illustrates the implementation flow of the information push method in an embodiment of this application. Figure 1 Information push methods can be applied to push systems. In practical applications, intelligent customer service systems can be used as push systems. For example... Figure 1 As shown, the method includes:

[0065] S101: Based on the interaction data with the interaction object, determine at least one intention of the interaction object;

[0066] In this step, the interaction object is any person or robotic device that can interact with the push system. The interaction object can input what it wants to inquire about or learn about into the push system, and the push system will output the data expected by the interaction object based on the input. The aforementioned input from the interaction object and output from the push system can be regarded as the interaction data generated by the interaction between the push system and the interaction object.

[0067] It is understandable that in practical applications, the content input by the interactive object in an interaction may be one or more different aspects. That is, the interactive object has the intention to consult or understand multiple different aspects during this interaction. Thus, the interactive object and the push system can generate one or more rounds of question and answer in one interaction.

[0068] For example, in a Q&A session, the user inputs the topic or question "How to restore my phone to factory settings?" In response, the push system outputs a reply or answer such as "Open your phone's Settings icon, click the System and Updates page, select the Factory Reset function, and click Reset Phone." In this Q&A session, the user's intention is to restore their phone to factory settings.

[0069] For example, in a Q&A session, the user inputs the question "How do I change my phone password?" In response, the push system outputs a reply such as "Open your phone's settings, tap the fingerprint, face, or password option, enter the selected fingerprint settings page, face recognition page, or password page, and enter your new password." In this Q&A session, the user is inquiring about their intention to set a phone password.

[0070] In an interaction between an interactive object and a push system, a question-and-answer session like the one described above might consist of one round, such as simply asking to restore the phone to factory settings or simply asking to set the phone password. It could also consist of multiple rounds, first asking to set the phone password, then asking to reset the phone. In different rounds, the interactive object's intent might be the same or different, such as different rounds of question-and-answer sessions corresponding to different intents. The data generated from each question-and-answer session can be considered interaction data.

[0071] In this application, a pre-defined intent recognition method is used to identify the intent in the interaction data, thereby obtaining one, two, or more intents generated by the interactive object during a single interaction. The pre-defined intent recognition method includes any reasonable method capable of realizing the intent. Examples include intent rule-based recognition methods, neural network-based intent recognition methods, and slot-filling-based intent recognition methods.

[0072] S102: Analyze the interaction data to determine the matching degree parameter of the at least one intent.

[0073] In this application, the matching degree parameter characterizes the satisfaction level of the interactive object with the response to the question provided by the push system under each identified intent. It can be understood that in a single interaction, the more intents the interactive object has, the more questions it inputs. The fewer intents, the fewer questions it inputs. In the same interaction, the more times the push system responds to each question input by the interactive object, the higher the satisfaction level. The fewer times the push system responds, the lower the satisfaction level. If the push system responds to all the questions input by the interactive object, then from the interactive object's perspective, it is equivalent to all questions being resolved, resulting in the highest level of satisfaction.

[0074] During implementation, the interaction data is analyzed. For example, the number of times the user inputs a question, and the number of times the push system responds to each question, are analyzed. Based on the ratio of responses to inputs, a matching parameter is determined for at least one intent. A higher ratio indicates more responses, higher satisfaction, and a larger matching parameter. Conversely, a lower ratio indicates fewer responses, lower satisfaction, and a smaller matching parameter.

[0075] S103: Determine the role information of the interactive object.

[0076] In this application, the role information represents the interaction object as either a positive role or a negative role. Specifically, positive review users can be used as positive roles, while negative review users can be used as negative roles.

[0077] In implementation, if the role information during the interaction process is considered as the current role information, then the historical role of the interaction object can be used as the current role information. For example, if the interaction object was a positive reviewer during a historical interaction, then the current role information of the interaction object is determined to be a positive reviewer. If the interaction object was a negative reviewer during a historical interaction, then the current role information of the interaction object is determined to be a negative reviewer.

[0078] In practice, the role information of the interactive object can also be determined based on the interaction data. Specifically, the role information of the interactive object can be predicted by analyzing its emotions when it inputs a question, and / or its response to the push system, and the feedback content given by the interactive object.

[0079] S104: Based on the matching degree parameter and the role information of the interaction object, determine the push strategy for the data to be pushed.

[0080] In this application, the matching degree parameter and the role information of the interaction object are combined to determine the push strategy. That is, the push strategy is determined by combining the role information of the interaction object and its satisfaction with the push system's response.

[0081] Push strategies are used to indicate when to push data to be pushed. For example, a push strategy might indicate that the data to be pushed should be pushed when the interaction target is a user with positive reviews and high satisfaction. Alternatively, it might indicate that the data to be pushed should not be pushed or should be postponed when the interaction target is a user with negative reviews and low satisfaction. Or, a push strategy might indicate adaptive adjustments to the data to be pushed (such as a gentler tone, more friendly wording, or adjustments to the questionnaire content).

[0082] In this application, the data to be pushed can vary based on the different roles of the interaction objects. In implementation, the data to be pushed can be determined based on the role information of the interaction objects. Specifically, when the interaction object is a positive role, first data is determined as the data to be pushed; when the interaction object is a negative role, second data is determined as the data to be pushed, wherein the second data is data generated based on the interaction data. The first data and the second data are different. The first data can be considered as a pre-set questionnaire for users who give positive reviews. This questionnaire can be a template questionnaire. The second data can be considered as a questionnaire generated based on the actual interaction between users who give negative reviews and the push system. Because this questionnaire is generated based on the actual interaction, it is more targeted and can soothe and inquire about the reasons for giving negative reviews, thereby guiding the push system to improve user satisfaction.

[0083] The second data can be viewed as a customized questionnaire generated for interaction objects with non-positive roles. This approach, which determines different data to be pushed to interaction objects with different roles, enables targeted questionnaires to be pushed to interaction objects with different roles, making it highly practical and feasible.

[0084] S105: The push strategy is used to push the data to be pushed.

[0085] In practical applications, when the interaction target is a user with positive reviews and high satisfaction, a template questionnaire designed for that user can be pushed to them. Conversely, when the interaction target is a user with negative reviews and low satisfaction, a questionnaire generated for that user can be pushed to them.

[0086] In steps S101 to S105, during the interaction process, the matching degree parameter of at least one (interaction) intent and the role of the interaction object are combined to determine the push strategy for the data to be pushed, and the push strategy is then used to push the data to be pushed. Essentially, the push strategy is determined by combining the interaction object's satisfaction with the push system's response to the question during the interaction process and the interaction object's role. Compared with related technologies, the push strategy determined based on the matching degree parameter and role information can push the information to be pushed at the appropriate time, providing push flexibility.

[0087] In addition, this application enables targeted push notifications by allowing different roles to receive different data.

[0088] In addition to the aforementioned scheme of analyzing interaction data to determine the matching degree parameter of at least one intent, this application can also determine the matching degree parameter of at least one intent through the following scheme. This application provides a machine learning model for predicting whether all questions input by an interactive object to a push system have been resolved. For distinction, the machine learning model for predicting whether all questions input by an interactive object to a push system have been resolved is considered a first prediction model. The first prediction model can determine the matching degree parameter of at least one intent based on interaction data.

[0089] like Figure 2 As shown, during implementation, the schemes shown in S201 to S203 can be executed.

[0090] S201: Obtain the target parameters of the at least one intention, wherein the target parameters of the at least one intention characterize the reference probability of the at least one intention changing.

[0091] In this application, a reference probability of intention change needs to be obtained, and the reference probability of intention change is used as an input to the first prediction model, thereby realizing the prediction of whether the interactive object input problem has been completely solved.

[0092] In this application, the scheme for the reference probability of a change in intent is implemented by the following scheme: a historical reference value of the at least one intent obtained based on historical interaction data is obtained, the historical reference value representing the probability of the at least one intent changing in history; the historical reference value of the at least one intent is used as the target parameter of the at least one intent.

[0093] In other words, this application calculates the probability of changes in each possible intent generated by an interactive object in the past based on historical interaction data, and uses the probability of changes in each possible intent generated in the past as a reference probability of changes in each possible intent. This method of obtaining the reference probability of intent changes is simple and easy to implement, and can be easily promoted in engineering.

[0094] In this application, the objects interacting with the push system on the network are mostly different users. Within the network, each topic input by any of the interacting objects corresponds to an intent. Changes in the topic content input by an interacting object signify a change in the interacting object's intent. For example, if the input topic changes from "How to restore the phone to factory settings" to "How to change the phone password," the interacting object's intent changes from asking about factory settings to changing the password. Therefore, in this application, situations where intent changes include the interacting object changing from inputting one topic to inputting another.

[0095] In implementation, analysis of historical interaction data, such as the analysis of input topics from various interaction objects throughout history, can be used to calculate the probability that each possible intention has changed from one intention to another in the past. For example, calculating the probability of intention... Figure 1 Historically from Italy Figure 1 The probability of changing to other intentions, if the intention is calculated. Figure 1 Historically, the probability of changing to other intentions is 70%, and 70% is intentional. Figure 1 Historical reference values. Figure 1 70% of the historical reference volume can be used as a reference. Figure 1 Reference probability of change (meaning) Figure 1 The target parameters). Among them, historical interaction data can be the interaction data generated between various possible users in the network and the push system within a past period (such as the past three months).

[0096] In this step, the scheme for obtaining the historical reference value of the at least one intent based on historical interaction data includes: for any intent among the at least one intent; determining a first type of target event and a second type of target event from the historical interaction data, wherein the first type of target event represents an event in the historical interaction data in which the arbitrary intent changes to a first type of optional intent; and the second type of target event represents an event in the historical interaction data in which the arbitrary intent changes to a second type of optional intent; and obtaining the historical reference value of the arbitrary intent based on the reference attributes of the first type of target event and the second type of target event in the historical interaction data; wherein the reference attributes represent the number of times or frequency of occurrence in the historical interaction data.

[0097] In practical applications, two scenarios exist. Scenario 1: After an interactive object inputs a topic corresponding to a certain intent into the push system, it may input the same topic under another intent. Scenario 2: After an interactive object inputs a topic corresponding to a certain intent into the push system, it leaves the push system and no longer inputs topics. To distinguish between these two scenarios, this application considers Scenario 1, appearing in historical interaction data, as an event indicating a shift from a certain intent to a first type of optional intent. Scenario 2, appearing in historical interaction data, is considered an event indicating a shift from a certain intent to a second type of optional intent.

[0098] It is understandable that the number of times or frequencies of occurrence of the aforementioned scenario one and / or scenario two under a certain intention can be statistically determined in the interaction data (historical interaction data) over a past period. Adding the occurrences of scenario one and scenario two over the past period yields the sum. Dividing the occurrences of scenario one over the past period by the sum gives the probability that the intention has changed historically, i.e., the historical reference value of the intention. This content can be used as a textual explanation of formula (1).

[0099] In implementation, it can be done using G<Source,Target,Weight> This expression stores the changes in an intent and the number of times these changes occur. Here, Source represents any possible intent generated in the historical interaction data. If Source = Intent I, then Target represents other intents that can change from Intent I, such as Intent J (Target = Intent J). Weight represents the number of times the interaction data changes from Intent I to Intent J. If the interaction object leaves after Intent I and no other intent is generated, then Target = None.

[0100] According to formula (1), calculate the probability that intention I has changed in history.

[0101] P(I)=ΣWeight(i)(G<I,Target(i),Weight(i)> |Target≠None) / Σ

[0102] G<I,Target(i),Weight(i)> |Target≠None and Target=None); (1)

[0103] In Equation (1), Target(i) represents at least one of the other intentions that Intention I may change to in the historical interaction data. Weight(i) represents the number of times Intention I changes to at least one of the other intentions in the historical interaction data.

[0104] For example, for Intention I, in the historical interaction data, Intention I can change to Intention B, Intention C, and Intention D. Specifically, it changes to Intention B 30 times, to Intention C 40 times, and to Intention D 10 times. Meanwhile, in the historical interaction data, Intention I remains unchanged 20 times. Therefore, the G of Intention I...<Source,Target,Weight> Including G<I,Target=B,Weight=30> G<I,Target=C,Weight=40> G<I,Target=D,Weight=10> and G<I,Target=None,Weight=20> According to formula (1), the probability that intention I has changed in history is (30+40+10) / (30+40+10+20) = 80%.

[0105] In this way, based on historical interaction data, the probability of each possible intent changing over time can be obtained and used as a reference probability for the intent. This reference probability calculation scheme can be updated once over a period of time. For example, if the past period is 3 months, the probability of each possible intent changing over the past 3 months can be updated every 3 months. This ensures the accuracy of the reference probability of the intent, thereby providing a guarantee for pushing push data at the appropriate time.

[0106] S202: Input the at least one intention, the target parameters of the at least one intention, and the first profile data of the interactive object into the first prediction model to obtain a target prediction result. The target prediction result represents whether it is a prediction result of the matching data corresponding to the at least one intention output generated by the interactive object in the interaction data.

[0107] In this step, outputting matching data corresponding to the intent refers to the push system responding to the questions or topics entered by the interactive object under the intent. The response to the questions or topics entered under the intent is considered the matching data corresponding to the intent. During an interaction, the interactive object's intent may be one intent, or it may be two or more intents. This application provides a machine learning model for predicting whether all of the interactive object's questions will be responded to or receive feedback from the push system. That is, whether all of the interactive object's questions will be resolved. This machine learning model is considered the first prediction model.

[0108] like Figure 3 As shown, the inputs to the first prediction model (Model 1) mainly include several types of inputs, such as intent features, device features, and user features. Intent features include the intent obtained through analysis of interaction data, and the reference probability of that intent changing (the target parameter of the intent). Device features include the brand, years of use, and warranty status of the mobile phone, computer, or other terminal used by the interactive object to interact with the push system. User features mainly include two aspects: the basic information of the interactive object and the historical service information of the interactive object. Basic information includes the country of the interactive object, the language used during interaction, and the access channel used to interact with the push system, such as accessing via telephone or using an app. The historical service information of the interactive object includes historical access information, historical interaction information, and historical feedback information. Historical access information includes the number of times the interactive object accessed the push system historically and the average access duration. Historical interaction information includes the average number of topic changes, the average number of interaction rounds, and the sentiment of the interactive object historically.

[0109] Features and whether there has been interaction in the last 5 minutes, etc. Historical feedback information includes whether the interacting object has given positive (good) feedback to the push system in the past 5 minutes. Among them, device features and user features are included.

[0110] It can be used as the initial profile data for interactive objects.

[0111] In this application, the first prediction model is a machine learning model. Machine learning models possess certain robustness, robustness, and stability. Using the first prediction model to predict whether all problems have been solved can guarantee prediction accuracy. This provides a certain guarantee for achieving flexible push notifications.

[0112] 0. In practical applications, when Intent I, its target parameters, and the first profile data of the interacting object are input into the first prediction model, the first prediction model will predict whether Intent I changes based on its own input. That is, it will predict whether the interacting object will input other topics corresponding to Intent I after inputting the topic corresponding to Intent I into the push system. If Intent I is predicted not to change, it means the interacting object will no longer request...

[0113] If other topics are asked, and all questions from the interaction object are resolved during this interaction, the interaction with interaction object 5 can end. The first prediction model can produce a target prediction result representing that all questions raised by the interaction object in this interaction data have been resolved. If prediction I changes, it means that the interaction object will continue to ask other topics, and other questions raised by the interaction object still need to be resolved during this interaction. The first prediction model can produce a target prediction result representing that not all questions raised by the interaction object in this interaction data have been resolved.

[0114] 0. That is, the first prediction model in this application is based on the prediction of whether the current intent of the input will change.

[0115] The test results are used to determine whether all problems arising during the interaction have been resolved. The reference probability of a change in the current intent serves as an input to the first prediction model. This reference probability is derived from historical interaction data and has a certain degree of reliability.

[0116] To a certain extent, this can assist the first prediction model in accurately predicting whether the current intent has changed. This, in turn, provides a guarantee for flexible push notifications.

[0117] S203: Based on the target prediction result, determine the matching degree parameter of the at least one intention.

[0118] In this step, the target prediction result includes matching data for at least one intent output generated by the interactive object in the interaction data and matching data for at least one intent corresponding to at least one intent, as well as matching data for at least one intent output not generated by the interactive object in the interaction data and matching data for at least one intent corresponding to at least one intent. That is, the target prediction result includes whether all problems generated by the interactive object in the interaction data are resolved, and whether all problems generated by the interactive object in the interaction data are not resolved. If the target prediction result indicates that all problems are resolved, a larger value is assigned to the matching degree parameter. If the target prediction result indicates that all problems are not resolved, a smaller value is assigned to the matching degree parameter. A larger matching degree parameter value indicates higher user satisfaction with the interaction process. A smaller matching degree parameter value indicates lower user satisfaction with the interaction process.

[0119] In steps S201 to S203, the intent, its target parameters, and the first profile data of the interactive object are used as inputs to the first prediction model, enabling accurate prediction of the target prediction result and thus determining accurate matching parameters. This provides a guarantee for realizing the flexible push functionality of this application.

[0120] In addition to the aforementioned scheme for determining the role information of interactive objects, this application can also predict the role information of interactive objects through the following scheme. This application provides a machine learning model for predicting the role information of interactive objects. For distinction, the machine learning model for predicting the role information of interactive objects is considered the second prediction model in this application. The second prediction model can also be used to determine the role information of interactive objects based on interaction data. In implementation, the interaction data, the at least one intent, and the second profile data of the interactive object can be input into the second prediction model. The second prediction model predicts the role information of the interactive object and outputs the prediction result. The role information of the interactive object is obtained by obtaining the prediction result of the second prediction model.

[0121] Combination Figure 4 As shown, the inputs to the second prediction model (Model 2) mainly include several types of inputs such as intent features, device features, and user features. Intent features include the intent obtained through analysis of interaction data. Device features include the brand, age, warranty status, and historical repair history of the mobile phone, computer, and other terminals used by the user to interact with the push system. User features mainly include two aspects: basic information of the user, historical interaction information between the user and the push system, current interaction information, and historical service information of the user. Basic information includes the user's country of origin, the language used during interaction, and the access channel used to interact with the push system. Historical service information includes historical interaction information and historical feedback information. Historical interaction information includes the average number of times the user was transferred to human assistance, the reason for the most recent transfer, the average duration of interaction with human customer service, the user's historical "like" rate, and questionnaire ratings. Historical interaction information between the user and the push system includes the historical number of visits, average visit duration, and visit time. The current interaction information between the interactive object and the push system includes the number of interaction rounds, emotional characteristics, whether there has been interaction in the last 5 minutes, whether there is positive feedback, the number of times the topic has changed, and the number of times the robot does not understand.

[0122] Among them, the basic information of device characteristics and user characteristics, the historical interaction information between the interactive object and the push system, and the historical service information can be used as the second profile data of the interactive object.

[0123] In this application, the second prediction model is a machine learning model. Machine learning models possess certain robustness, robustness, and stability. Using the second prediction model to predict the role information of interactive objects can ensure the accuracy of role information prediction. This provides a certain guarantee for the implementation of flexible push notifications.

[0124] The aforementioned solution involves using two prediction models to predict matching parameters and role information; that is, applying two prediction models. It can be understood that before applying the two prediction models, they need to be trained and optimized. After optimization, the two prediction models are deployed online. Based on the interaction data between the interactive object and the push system, the two prediction models are used to predict matching parameters and role information.

[0125] During implementation, it can be executed. Figure 5 The process shown is to complete the training and deployment of two prediction models.

[0126] First, user profile data is designed and created. During implementation, first and second profile data of all possible users in the network can be collected as the user profile data in this application.

[0127] Secondly, data statistics and feature engineering were performed on the user profile data. This involved collecting, statistically analyzing, standardizing, removing outliers, and performing dimensionality reduction on the user profile data of users who had completed questionnaires in the last three months.

[0128] In addition, statistics were collected on the various possible intentions of users who filled out the questionnaire in the past three months and the number of times each intention appeared online, and the reference probability of the change of various possible intentions was calculated according to formula (1).

[0129] Next, we obtain training data and collect the training data to train the model to be trained. Here, since it involves training two prediction models, it involves obtaining training data for each prediction model.

[0130] The acquisition of training data for the first prediction model includes obtaining first profile data, partial intents, and reference probabilities of changes in each intent within the partial intents, which serve as first sample data.

[0131] The process of obtaining the labels for the first sample data is as follows: Based on the definition of a human-computer dialogue system (a system that enables interaction between an interactive object and a push system can be considered a human-computer dialogue system), users whose push system responded to their topics and who provided simple feedback ("like / dislike") to the responses were selected. If the user subsequently interacted with the push system (such as asking other questions, clicking buttons, etc.), the user was marked as "problem not fully resolved," and the label for the first sample data was Y = 1. If the user subsequently left directly or did not interact, the user was marked as "problem fully resolved," and the label for the first sample data was Y = 0.

[0132] In this application, the training data for the first prediction model includes first sample data and the labels of the first sample data. From several candidate binary classification supervised learning models, one model is selected as the first model to be trained in this application. This first model is trained using the first sample data and its labels to obtain the first prediction model. Specifically, the candidate binary classification supervised learning models include KNN, decision tree, random forest, SVM, Adaboost, and Xgboost. The first sample data is used as the input to these models, and the labels of the first sample data are used as the output of each model. Each model provides an F-score based on its input and output. From these models, the Xgboost model with the highest F-score is selected as the first model to be trained. The first sample data is used as the input to the first model to be trained, and the labels of the first sample data are used as the output of the first model to be trained. The first model is then trained to adjust its parameters and improve its accuracy, thereby obtaining the first prediction model.

[0133] The acquisition of training data for the second prediction model includes collecting second profile data, partial intent data, and interaction data generated between interactive objects who have filled out questionnaires in the past three months and the push system, which are used as second sample data.

[0134] The process of obtaining the labels for the second sample data is as follows: The questionnaire results of users who completed the questionnaire in the last three months are used as the Y-label. Assuming the maximum satisfaction score in the questionnaire is 10 points, if a user's satisfaction score in the questionnaire is greater than or equal to a preset threshold, such as 5 points, then the user is considered a positive user, and the label for the second sample data is recorded as Y = 1. Conversely, if the score is less than or equal to a negative user, then the user is considered a negative user, and the label for the second sample data is recorded as Y = 0.

[0135] In this application, the training data for the second prediction model includes the second sample data and the labels of the second sample data. Similar to the first prediction model, which requires selecting a suitable model from several binary classification supervised learning models as the first model to be trained, in this application, the second sample data is used as the input to the above models, and the labels of the second sample data are used as the output of the above models. Each model provides an F-score based on its input and output. The XGBoost model with the highest F-score is selected from the above models as the second model to be trained. The second sample data is used as the input to the second model to be trained, and the labels of the second sample data are used as the output of the second model to be trained. The second model is then trained to adjust its parameters and improve its accuracy, thereby obtaining the second prediction model.

[0136] In this application, both the first model to be trained and the second model to be trained are XGBoost models. Naturally, the first and second prediction models that have been trained or have completed training are XGBoost models.

[0137] Finally, the two trained prediction models are deployed online and applied.

[0138] The applications provided in this application can be found in [reference needed]. Figure 6 As shown. In Figure 6 In this system, the first and second profile data of the interacting object are used as offline profile data. Taking the push system as an intelligent customer service system, such as an intelligent customer service robot, the interaction data between the interacting object and the intelligent customer service robot is recorded from the beginning of the dialogue. This data is combined with the first profile data of the interacting object, and a first prediction model is used to predict whether the user's problem has been fully resolved. The second profile data of the interacting object is combined with the second prediction model to predict the role information of the interacting object. Based on the prediction results of whether the user's problem has been fully resolved and the prediction results of the role information of the interacting object, a push strategy is determined, and the push data to be pushed is pushed using the push strategy. This achieves the push of push information at the appropriate time, realizing flexible push. Based on this, Figure 6 The proposed solution can be considered as a new questionnaire delivery solution provided in this application.

[0139] exist Figure 6 In the illustrated process, the current user and the intelligent customer service robot can engage in one, two, or more rounds of interaction during a single interaction. Each round corresponds to a topic, and if a topic corresponds to an intent, then one, two, or more intents can be addressed during a single interaction. See the description of the scheme below for details on each round of interaction.

[0140] In one round of interaction, the current user (the interaction target) accesses the intelligent customer service chatbot. The user inputs a topic into the chatbot, which then identifies whether the input is casual conversation or a problem requiring the chatbot's assistance. If the chatbot identifies a need for assistance, it determines whether to provide an answer.

[0141] If the intelligent customer service robot has provided an answer to the question, it determines whether the user has provided any feedback on the response. In practice, if the answer provided by the intelligent customer service robot is helpful to the user, the user will click the "Helpful" button. If the answer provided by the intelligent customer service robot is not helpful to the user, the user will click the "UnHelpful" button or provide no feedback. The intelligent customer service robot determines whether the time since the user stopped interacting with the robot has exceeded a preset value, such as 5 minutes. If it has not exceeded this time, it calls the first prediction model to predict whether all of the user's questions have been resolved. If it has exceeded this time, the process returns to the stage of waiting for user input, waiting for the user to input the next topic, and entering the next round of interaction. Each round of interaction is similar; see the explanation section.

[0142] Specifically, the intelligent customer service robot identifies that the user's input topic indicates a need for the robot's assistance in resolving a problem. Based on the current topic input, it identifies the user's current intent. Furthermore, it reads a reference probability of changes in the current intent, derived from historical interaction data.

[0143] When the answer provided by the intelligent customer service robot is helpful to the user—that is, when the user clicks the Helpful button—the intelligent customer service robot invokes the first prediction model. The first prediction model is fed with the identified current intent generated under the current topic, the reference probability of the current intent changing, and the user's first profile data. The first prediction model then predicts whether all of the user's questions or topics have been resolved. In essence, if the first prediction model, based on the current intent, the reference probability of the current intent changing, and the user's first profile data, predicts a high probability that the current intent will actually change during the current user's interaction, it means that the user, while inputting the current topic, is likely to input a different topic, expecting the intelligent customer service robot to provide a solution. Thus, the first prediction model predicts that the user may input other topics, and all the questions generated during the current interaction have not yet been resolved. Therefore, the intelligent customer service robot waits for the user to input another topic, or pushes other related topics to the user based on the current topic, until the first prediction model, based on subsequent intentions, the reference probability of that intention changing, and the first profile data, predicts that the intention will not change, i.e., predicting that all questions have been resolved.

[0144] If the first prediction model, based on the current intent, the reference probability of the current intent changing, and the current user's first profile data, predicts that the probability of the current intent actually changing during the current user's current interaction is small, it means that the current user will not input any other topics, and all of the current user's questions will be resolved.

[0145] In the aforementioned scheme, the user clicking the Helpful button can serve as a trigger event for invoking the first prediction model.

[0146] If the first prediction model predicts that all of the user's problems will be resolved, then a larger value can be assigned to the matching degree parameter. A higher value for the matching degree parameter indicates a higher level of user satisfaction with the current interaction.

[0147] Once all issues for the current user have been resolved, the second prediction model is invoked. All interaction data generated by the current user during this interaction, each intent, and the current user's second profile data are input into the second prediction model. The second prediction model then predicts whether the current user will leave a positive or negative review. If a positive review is received, the intelligent customer service chatbot pushes a standardized satisfaction questionnaire to the current user. If a negative review is received, a customized satisfaction questionnaire is generated based on all interaction data generated by the current user during this interaction, inquiring about the reasons for the negative review.

[0148] In this application, a unified satisfaction questionnaire is pushed to users when all issues raised during a single interaction are resolved and the user leaves a positive review. Conversely, a customized satisfaction questionnaire is pushed to users when all issues raised during a single interaction are resolved and the user leaves a negative review. Essentially, the satisfaction questionnaire is pushed out when all issues are resolved. Compared to related technologies that push questionnaires based on fixed trigger nodes, this application's technical solution can push satisfaction questionnaires at appropriate times, achieving flexible push functionality.

[0149] Furthermore, pushing the satisfaction survey when all issues have been resolved can avoid the negative user experience caused by pushing the survey when there are still unresolved issues. Pushing the survey at an appropriate time, as proposed in this application, can greatly improve the user's experience with the intelligent customer service robot.

[0150] Furthermore, the technical solution of this application, once all problems are resolved, can push different satisfaction questionnaires based on the roles of the interacting objects, achieving targeted delivery and strong practicality. Specifically, for users who leave negative reviews, a customized satisfaction questionnaire generated for that user can be pushed, guiding the intelligent customer service robot to improve its customer service satisfaction.

[0151] As can be seen from the preceding content, the technical solution of this application has at least the following advantages.

[0152] First, in the technical solution of this application, through human-computer dialogue (interaction between the user and the intelligent customer service robot), the analysis of the user's intent understanding, the possibility of intent jump, and other human-computer interaction characteristics (such as the first profile data) is used to predict whether the user's questions have been fully resolved. This makes the timing of questionnaire push dynamic, depending on the interaction state between the user and the human-computer interaction, and does not rely on fixed trigger nodes as in related technologies to trigger questionnaire push.

[0153] This application presents a novel questionnaire delivery solution, which can be considered an intelligent questionnaire delivery solution that can deliver questionnaires at appropriate times. Furthermore, because this intelligent questionnaire delivery solution is based on a machine learning model, it can be viewed as a machine learning-based intelligent questionnaire delivery solution. Due to the robustness, stability, and effectiveness of machine learning models, accurate delivery of questionnaires at the appropriate time can be guaranteed.

[0154] Secondly, the technical solution of this application is practical.

[0155] The proposed intelligent questionnaire push method for user satisfaction is similar to a smart customer service robot with certain human characteristics. It allows the robot to predict user psychology (whether the problem has been resolved and whether the user tends to give a positive review) before deciding whether to push the questionnaire at the appropriate time. This increases the completion rate and volume of user satisfaction questionnaires without affecting user experience.

[0156] With iterative optimization of the two prediction models, and combined with data analysis of user questionnaire results with higher confidence from business personnel, we can further understand the reasons for user dissatisfaction, thereby improving the service level of the intelligent customer service robot and increasing questionnaire satisfaction.

[0157] This application also provides an information push device, such as... Figure 7 As shown, it includes:

[0158] The first determining unit 701 is used to determine at least one intention of the interactive object based on the interaction data with the interactive object;

[0159] The second determining unit 702 is used to analyze the interaction data to determine the matching degree parameter of the at least one intention;

[0160] The third determining unit 703 is used to determine the role information of the interactive object;

[0161] The fourth determining unit 704 is used to determine the push strategy for the data to be pushed based on the matching degree parameter and the role information of the interaction object;

[0162] The push unit 705 is used to push the data to be pushed using the push strategy.

[0163] In some embodiments, the second determining unit 702 is used for

[0164] Obtain target parameters for the at least one intention, wherein the target parameters for the at least one intention characterize a reference probability that the at least one intention will change;

[0165] The at least one intention, the target parameters of the at least one intention, and the first profile data of the interactive object are input into the first prediction model to obtain a target prediction result. The target prediction result represents whether it is a prediction result of the matching data of the at least one intention output generated by the interactive object in the interaction data and the at least one intention corresponding to the at least one intention.

[0166] Based on the target prediction result, a matching degree parameter for the at least one intent is determined.

[0167] In some embodiments, the third determining unit 703 is used to determine the role information of the interactive object based on the interaction data.

[0168] In some embodiments, the third determining unit 703 is used to input the interaction data, the at least one intention, and the second profile data of the interaction object into the second prediction model to obtain the role information of the interaction object.

[0169] In some embodiments, the second determining unit 702 is configured to obtain a historical reference value of the at least one intention based on historical interaction data, the historical reference value representing the probability that the at least one intention has changed in history; and to use the historical reference value of the at least one intention as a target parameter of the at least one intention.

[0170] In some embodiments, the second determining unit 702 is configured to:

[0171] For any of the at least one intent;

[0172] From historical interaction data, a first type of target event and a second type of target event are identified. The first type of target event represents an event in the historical interaction data in which the arbitrary intent changes to a first type of optional intent. The second type of target event represents an event in the historical interaction data in which the arbitrary intent changes to a second type of optional intent.

[0173] Based on the reference attributes of the first type of target event and the second type of target event in the historical interaction data, the historical reference quantity of the arbitrary intent is obtained; the reference attribute represents the number of times or frequency of occurrence in the historical interaction data.

[0174] In some embodiments, the fourth determining unit 704 is used to determine the data to be pushed based on the role information of the interactive object.

[0175] In some embodiments, the fourth determining unit 704 is used to determine the first data as data to be pushed when the interaction object is a positive role;

[0176] When the interaction object is a non-positive role, the second data is determined to be the data to be pushed, wherein the second data is data generated based on the interaction data.

[0177] It should be noted that the information push device in this application embodiment solves the problem in a similar way to the aforementioned information push method. Therefore, the implementation process and implementation principle of the information push device can be found in the description of the implementation process and implementation principle of the aforementioned method, and the repeated parts will not be repeated.

[0178] It should be noted that the description of the information push device in this application embodiment is similar to the description of the information push device method embodiment described above, and has similar beneficial effects as the method embodiment, therefore it will not be repeated. For any technical details not covered in the information push device provided in this application embodiment, please refer to... Figures 1 to 6 The meaning is understood in accordance with the description of any of the accompanying drawings.

[0179] According to embodiments of this application, this application also provides an electronic device and a readable storage medium.

[0180] Figure 8 A schematic block diagram of an example electronic device 800 that can be used to implement embodiments of this application is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the application described and / or claimed herein.

[0181] like Figure 8As shown, device 800 includes a computing unit 801, which can perform various appropriate actions and processes based on a computer program stored in read-only memory (ROM) 802 or a computer program loaded from storage unit 808 into random access memory (RAM) 803. RAM 803 may also store various programs and data required for the operation of device 800. The computing unit 801, ROM 802, and RAM 803 are interconnected via bus 804. Input / output (I / O) interface 805 is also connected to bus 804.

[0182] Multiple components in device 800 are connected to I / O interface 805, including: input unit 806, such as keyboard, mouse, etc.; output unit 807, such as various types of monitors, speakers, etc.; storage unit 808, such as disk, optical disk, etc.; and communication unit 809, such as network card, modem, wireless transceiver, etc. Communication unit 809 allows device 800 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0183] The computing unit 801 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 801 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 801 performs the various methods and processes described above, such as the information push method. For example, in some embodiments, the information push method may be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 808. In some embodiments, part or all of the computer program may be loaded and / or installed on device 800 via ROM 802 and / or communication unit 809. When the computer program is loaded into RAM 803 and executed by the computing unit 801, one or more steps of the information push method described above may be performed. Alternatively, in other embodiments, the computing unit 801 may be configured to perform the information push method by any other suitable means (e.g., by means of firmware).

[0184] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific labeled products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0185] The program code used to implement the methods of this application may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing device, such that when executed by the processor or controller, the functions / operations specified in the flowcharts and / or block diagrams are implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0186] In the context of this application, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. Machine-readable media can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0187] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0188] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with embodiments of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0189] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact via communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other. Servers can be cloud servers, servers in distributed systems, or servers incorporating blockchain technology.

[0190] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. An information push method, comprising: Based on the interaction data with the interaction object, determine at least one intention of the interaction object; Obtain target parameters for the at least one intention, wherein the target parameters for the at least one intention characterize a reference probability that the at least one intention will change; The at least one intention, the target parameters of the at least one intention, and the first profile data of the interactive object are input into the first prediction model to obtain a target prediction result. The target prediction result represents whether it is a prediction result of the matching data of the at least one intention output generated by the interactive object in the interaction data and the at least one intention corresponding to the at least one intention. Based on the target prediction result, determine the matching degree parameter of the at least one intent; Determine the role information of the interactive object; Based on the matching degree parameter and the role information of the interaction object, determine the push strategy for the data to be pushed; The push strategy is used to push the data to be pushed.

2. The method according to claim 1, wherein determining the role information of the interactive object includes: Based on the interaction data, the role information of the interaction object is determined.

3. The method according to claim 2, wherein determining the role information of the interactive object based on the interaction data includes: The interaction data, the at least one intent, and the second profile data of the interaction object are input into the second prediction model to obtain the role information of the interaction object.

4. The method according to claim 1, wherein obtaining the target parameter of the at least one intention comprises: Obtain a historical reference value for the at least one intention based on historical interaction data, wherein the historical reference value represents the probability that the at least one intention has changed in history; The historical reference value of the at least one intention is used as the target parameter of the at least one intention.

5. The method according to claim 4, wherein obtaining the historical reference value of the at least one intent based on historical interaction data comprises: For any of the at least one intent; From historical interaction data, a first type of target event and a second type of target event are identified. The first type of target event represents an event in the historical interaction data in which the arbitrary intent changes to the first type of optional intent. The second type of target event represents an event in historical interaction data where the arbitrary intent changes to the second type of optional intent; Based on the reference attributes of the first type of target event and the second type of target event in the historical interaction data, the historical reference value of the arbitrary intention is obtained; The reference attribute represents the number of times or frequency of occurrence in the historical interaction data.

6. The method according to any one of claims 1 to 5, further comprising: Based on the role information of the interactive object, the data to be pushed is determined.

7. The method according to claim 6, wherein determining the data to be pushed based on the role information of the interactive object includes: When the interaction object is a positive role, the first data is determined to be the data to be pushed; When the interaction object is a non-positive role, the second data is determined to be the data to be pushed, wherein the second data is data generated based on the interaction data.

8. An information push device, comprising: The first determining unit is configured to determine at least one intention of the interactive object based on interaction data with the interactive object; The second determining unit is used to obtain target parameters of the at least one intention, wherein the target parameters of the at least one intention characterize a reference probability that the at least one intention changes; The at least one intention, the target parameters of the at least one intention, and the first profile data of the interactive object are input into the first prediction model to obtain a target prediction result. The target prediction result represents whether it is a prediction result of the matching data of the at least one intention output generated by the interactive object in the interaction data and the at least one intention corresponding to the at least one intention. Based on the target prediction result, determine the matching degree parameter of the at least one intent; The third determining unit is used to determine the role information of the interactive object; The fourth determining unit is used to determine the push strategy for the data to be pushed based on the matching degree parameter and the role information of the interaction object; The push unit is used to push the data to be pushed using the push strategy.

9. An electronic device, comprising: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-7.

Images