Intention prediction method and device, server and storage medium

Abstract

The invention discloses an intention prediction method and device, a server and a storage medium, and relates to the technical field of artificial intelligence. The method comprises the steps of obtaining user data corresponding to at least two users and scene data corresponding to scenes where the at least two users are located, inputting the user data corresponding to the at least two users, the scene data corresponding to the scenes where the at least two users are located and a plurality of preset collaborative intentions into a collaborative intention prediction model, the probability of each preset collaborative intention in a plurality of preset collaborative intentions generated by the at least two users and output by the collaborative intention prediction model is obtained, and the collaborative intentions are used for indicating intentions generated by the at least two users due to the associated purposes. According to the method and the device, a plurality of collaborative intentions are defined, and the probability that different users generate each collaborative intention is identified through the multi-user data, so that the scene which can be applied to follow-up content pushing is improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and more specifically, to a method, device, server, and storage medium for predicting intention. Background Art

[0002] With the rapid development of artificial intelligence technology, current electronic devices (such as smartphones) can accurately identify and predict user intentions based on natural language processing, speech recognition, image recognition, intelligent prediction and other technologies. However, current intention recognition and prediction technologies are mostly based on the user's personal data to identify the user's personal intentions, and the application scenarios are limited. Summary of the invention

[0003] In view of the above problems, the present application proposes an intention prediction method, device, server and storage medium to solve the above problems.

[0004] In a first aspect, an embodiment of the present application provides a method for predicting intentions, the method comprising: obtaining user data corresponding to at least two users and scene data corresponding to scenes in which the at least two users are located; inputting the user data corresponding to the at least two users, the scene data corresponding to the scenes in which the at least two users are located, and a plurality of preset collaborative intentions into a collaborative intention prediction model, and obtaining the probability of the at least two users generating each of the plurality of preset collaborative intentions output by the collaborative intention prediction model, wherein the collaborative intention is used to indicate the intentions generated by at least two users due to associated purposes.

[0005] In a second aspect, an embodiment of the present application provides an intention prediction device, the device comprising: a data acquisition module, used to acquire user data corresponding to at least two users, and scene data corresponding to the scenes in which the at least two users are located; an intention prediction module, used to input the user data corresponding to the at least two users, the scene data corresponding to the scenes in which the at least two users are located, and a plurality of preset collaborative intentions into a collaborative intention prediction model, and obtain the probability of the at least two users generating each of the plurality of preset collaborative intentions output by the collaborative intention prediction model, wherein the collaborative intention is used to indicate the intention generated by at least two users due to associated purposes.

[0006] In a third aspect, an embodiment of the present application provides a server, comprising a memory and a processor, wherein the memory is coupled to the processor, the memory stores instructions, and when the instructions are executed by the processor, the processor executes the above method.

[0007] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, in which a program code is stored, and the program code can be called by a processor to execute the above method.

[0008] The intention prediction method, device, server and storage medium provided in the embodiments of the present application obtain user data corresponding to at least two users and scene data corresponding to the scenes in which the at least two users are located, input the user data corresponding to the at least two users, the scene data corresponding to the scenes in which the at least two users are located, and multiple preset collaborative intentions into a collaborative intention prediction model, and obtain the probability of each preset collaborative intention among the multiple preset collaborative intentions generated by the at least two users output by the collaborative intention prediction model, wherein the collaborative intention is used to indicate the intention generated by at least two users due to related purposes, thereby defining multiple collaborative intentions and identifying the probability of different users generating each collaborative intention through multi-user data, so as to improve the scenarios in which subsequent content push can be applied. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without paying any creative work.

[0010] Figure 1 A schematic diagram showing a flow chart of a method for predicting intentions provided in an embodiment of the present application is shown;

[0011] Figure 2 A schematic diagram showing a flow chart of a method for predicting intentions provided in an embodiment of the present application is shown;

[0012] Figure 3 A schematic diagram showing a flow chart of a method for predicting intentions provided in an embodiment of the present application is shown;

[0013] Figure 4 A schematic diagram showing a flow chart of a method for predicting intentions provided in an embodiment of the present application is shown;

[0014] Figure 5 A schematic diagram showing a flow chart of a method for predicting intentions provided in an embodiment of the present application is shown;

[0015] Figure 6 A schematic diagram showing a flow chart of a method for predicting intentions provided in an embodiment of the present application is shown;

[0016] Figure 7 A schematic diagram showing model training provided in an embodiment of the present application and intent recognition based on the trained model;

[0017] Figure 8 A schematic diagram of an interface of an electronic device provided in an embodiment of the present application is shown;

[0018] Fig. 9 A module block diagram of an intention prediction device provided by an embodiment of the present application is shown;

[0019] Fig.10 A block diagram of a server according to an embodiment of the present application for executing a prediction method according to an embodiment of the present application is shown;

[0020] Fig.11 A storage unit for storing or carrying program codes for implementing the prediction method according to the intention of the embodiment of the present application is shown. DETAILED DESCRIPTION

[0021] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0022] Among them, electronic devices (such as smartphones and smart watches) are infiltrating more and more life scenarios, which reflects that electronic devices can meet more user needs. On the other hand, the data that can be collected by electronic devices and their accessories are becoming more and more abundant. How to apply the collected data to meet user needs requires electronic devices to understand user intentions.

[0023] User intent, in simple terms, is what the user wants to do. For example, when a user uses a search engine to search for "fun parks nearby", the user's intent may be "travel"; when a user "turns on the Bluetooth headset on the Bluetooth face" on the way to work, the user's intent may be "make a call" or "listen to music". With the rapid development of artificial intelligence technology, current electronic devices can accurately recognize and predict user intent based on natural language processing, speech recognition, image recognition, intelligent prediction and other technologies, thereby realizing the transformation from "people looking for services" to "services looking for people".

[0024] However, most of the current intent recognition and prediction technologies are based on the user's personal device usage data and behavior habits to identify the user's personal intent. However, personal intent is limited in application scenarios, for example, it cannot be applied to carpooling scenarios, shared rental scenarios, and group order scenarios.

[0025] In response to the above problems, the inventors have discovered after long-term research and proposed the intention prediction method, device, server and storage medium provided in the embodiments of this application, which defines multiple collaborative intentions and identifies the probability of different users generating each collaborative intention through multi-user data, so as to improve the scenarios in which subsequent content push can be applied. Among them, the specific intention prediction method is described in detail in the subsequent embodiments.

[0026] See also Figure 1 , Figure 1 The flowchart of the method for predicting intent provided by an embodiment of the present application is shown. The method is used to define multiple collaborative intents and identify the probability of different users generating each collaborative intent through multi-user data to improve the scenarios in which subsequent content push can be applied. In a specific embodiment, the method for predicting intent is applied to Fig. 9 The intention prediction device 200 and the server 300 equipped with the intention prediction device ( Fig.10 ). The following will take the server as an example to illustrate the specific process of this embodiment. Of course, it can be understood that the server used in this embodiment may include a cloud server or a traditional server, etc., which is not limited here. Figure 1 The process shown is described in detail, and the intention prediction method may specifically include the following steps:

[0027] Step S110: Obtain user data corresponding to at least two users, and scene data corresponding to the scenes in which the at least two users are located.

[0028] In this embodiment, user data corresponding to at least two users can be obtained, and scene data corresponding to the scenes that the at least two users are in can be obtained. Optionally, the user data corresponding to the user can be used to reflect the attributes of the user, for example, the gender, age, occupation, hobbies, behavior, etc. of the user; the scene data corresponding to the scene that the user is in can be used to reflect the dimension of the scene that the user is in, for example, the time dimension, location dimension, etc. of the user, which is not limited here.

[0029] In some embodiments, if at least two users each have an electronic device (such as a smart phone), user data of the corresponding users can be collected through the electronic devices corresponding to the at least two users. Accordingly, the server can obtain the user data of the corresponding users uploaded by the electronic devices corresponding to the at least two users based on the communication connection between the electronic devices corresponding to the at least two users, thereby obtaining the user data corresponding to the at least two users.

[0030] Optionally, the electronic devices corresponding to at least two users respectively collecting user data of the corresponding users may include: the electronic devices corresponding to at least two users respectively receiving user data input by the corresponding users, or the electronic devices corresponding to at least two users respectively collecting user data of the corresponding users through sensors.

[0031] As an implementable method, obtaining user data corresponding to at least two users may include: obtaining behavioral data corresponding to at least two users, and obtaining user portraits corresponding to at least two users, pre-processing the behavioral data corresponding to at least two users and the user portraits corresponding to at least two users, and obtaining user data corresponding to at least two users.

[0032] Optionally, a user profile can be analyzed to obtain user feature labels, such as "white-collar worker", "music lover", "gender", "age", etc., and then the corresponding user profile is determined by the user feature labels.

[0033] Optionally, behavior data (User Behavior) can be collected through sensors (sensors of electronic devices or sensors of IOT devices), applications and system tracking points. Among them, behavior data can include system data of electronic devices, such as screen on and off, alarm status, Bluetooth status, WiFi status, etc.; behavior data can include application usage of electronic devices, such as conference software usage, weather software usage, game software usage, etc.; behavior data can include user behavior identification data, such as walking, running, etc.; behavior data can include geographic fence information, such as location, subway identification, etc.; behavior data can include IOT device data, such as watch data, bracelet data, headphone data, etc., which are not limited here.

[0034] As an approach, preprocessing the behavior data corresponding to at least two users and the user portraits corresponding to at least two users may include: filtering, denoising, deduplicating, and format conversion of one or a combination of these.

[0035] In some embodiments, if at least two users each have an electronic device (such as a smart phone), scene data corresponding to the scene in which the corresponding user is located can be collected through the electronic device corresponding to the at least two users. Accordingly, the server can obtain the scene data corresponding to the scene in which the corresponding user is located uploaded by the electronic device corresponding to the at least two users based on the communication connection between the electronic devices corresponding to the at least two users, thereby obtaining the user data corresponding to the scene in which the at least two users are located.

[0036] Optionally, the electronic devices corresponding to at least two users respectively collecting user data of the corresponding users may include: the electronic devices corresponding to at least two users respectively receiving user data input by the corresponding users, or the electronic devices corresponding to at least two users respectively collecting user data of the corresponding users through sensors.

[0037] As an implementable method, scene data refers to the scene that the user is in. In the early stage, the user data can be analyzed to summarize the mapping relationship between the user data and the scene data. Through the mapping relationship, the scene data corresponding to the scene that the user is currently in can be determined based on the user data of a certain user. Among them, common scenes include: a meeting room at a certain location, being on the subway at a certain time, sleeping at a certain location at a certain time, shopping in a shopping mall, etc., which are not limited here.

[0038] Step S120: Input the user data corresponding to each of the at least two users, the scene data corresponding to the scenes in which the at least two users are located, and a plurality of preset collaborative intentions into a collaborative intention prediction model, and obtain the probability of the at least two users generating each of the plurality of preset collaborative intentions output by the collaborative intention prediction model, wherein the collaborative intention is used to indicate the intention generated by at least two users for related purposes.

[0039] In this embodiment, when user data corresponding to at least two users and scene data corresponding to the scenes in which at least two users are located are obtained, the user data corresponding to at least two users, the scene data corresponding to the scenes in which at least two users are located, and multiple preset collaborative intentions can be input into a collaborative intention prediction model to obtain the probability of the at least two users generating each of the multiple preset collaborative intentions output by the collaborative intention output model.

[0040] Optionally, collaborative intent is used to indicate the intent generated by at least two users for related purposes. For example, "carpooling", "order sharing", "rental sharing", "second-hand car trading", "second-hand house trading", etc. In this process, these intentions generated by at least two users can be called "collaborative intent".

[0041] Optionally, the multiple preset collaborative intentions may include "carpooling", "order sharing", "rental sharing", "second-hand car trading", "second-hand house trading", etc., which are not limited here. In some embodiments, the multiple preset collaborative intentions can be saved after being pre-set, and when the collaborative intentions of at least two users need to be identified, the saved multiple preset collaborative intentions are obtained.

[0042] Optionally, the collaborative intention prediction model is obtained through machine learning. Specifically, a training data set is first collected, wherein the attributes or characteristics of one type of data in the training data set are different from those of another type of data. Then, the collected training data set is used to train and model a neural network according to a preset algorithm, thereby summarizing the rules based on the training data set to obtain the collaborative intention prediction model. In this embodiment, the training data set may, for example, be training user data corresponding to each of at least two users when they have a collaborative intention, and training scene data corresponding to the scenes in which each of the at least two users is located when they have a collaborative intention.

[0043] In some embodiments, the probability of at least two users generating each of the multiple preset collaborative intentions may be all the same, partially the same, or completely different. For example, assuming that the multiple preset collaborative intentions include "carpooling", "order sharing", and "rent sharing", the probability of at least two users generating the "carpooling" collaborative intention may be 0.8, the probability of generating the "order sharing" collaborative intention may be 0.5, and the probability of generating the "rent sharing" collaborative intention may be 0.1.

[0044] An intention prediction method provided in an embodiment of the present application obtains user data corresponding to at least two users and scene data corresponding to scenes in which at least two users are located, inputs the user data corresponding to at least two users, the scene data corresponding to scenes in which at least two users are located, and multiple preset collaborative intentions into a collaborative intention prediction model, and obtains the probability of each preset collaborative intention among the multiple preset collaborative intentions generated by the at least two users output by the collaborative intention prediction model, wherein the collaborative intention is used to indicate the intention generated by at least two users due to related purposes, thereby defining multiple collaborative intentions and identifying the probability of different users generating each collaborative intention through multi-user data, so as to improve the scenarios in which subsequent content push can be applied.

[0045] See also Figure 2 , Figure 2 The flowchart of the method for predicting intention provided by an embodiment of the present application is shown in FIG. Figure 2 The process shown is described in detail, and the intention prediction method may specifically include the following steps:

[0046] Step S210: Obtain user data corresponding to at least two users, and scene data corresponding to the scenes in which the at least two users are located.

[0047] Step S220: Input the user data corresponding to each of the at least two users, the scene data corresponding to the scenes in which the at least two users are located, and a plurality of preset collaborative intentions into a collaborative intention prediction model, and obtain the probability of the at least two users generating each of the plurality of preset collaborative intentions output by the collaborative intention prediction model, wherein the collaborative intention is used to indicate the intention generated by at least two users for related purposes.

[0048] For the detailed description of step S210 to step S220, please refer to step S110 to step S120, which will not be repeated here.

[0049] Step S230: Pushing content based on the probability that the at least two users generate the preset collaborative intentions.

[0050] In this embodiment, when the probability of at least two users generating each preset collaborative intention is determined, content can be pushed based on the probability of the at least two users generating each preset collaborative intention. For example, carpooling content is pushed to at least two users to increase the success rate of carpooling for the at least two users; order-sharing content is pushed to at least two users to increase the success rate of order-sharing for the at least two users; or house-sharing content is pushed to at least two users to increase the success rate of house-sharing for the at least two users, etc., which are not limited here.

[0051] As a method, when the server determines the probability that at least two users generate each preset collaboration intention, the server can directly push content based on the probability that the at least two users generate each preset collaboration intention.

[0052] As another way, when the server determines the probability that at least two users will generate each preset collaborative intention, the server can send the probability that at least two users will generate each preset collaborative intention to the corresponding manufacturer server (such as a carpooling manufacturer server, a group buying manufacturer server, etc.), so as to push content through each manufacturer server based on the probability that at least two users will generate the corresponding preset collaborative intention.

[0053] In some embodiments, when the probability of at least two users generating each preset collaborative intention is determined, the corresponding preset collaborative intention with the highest probability can be determined from multiple preset collaborative intentions, and content related to the corresponding preset collaborative intention with the highest probability can be pushed. As an implementable manner, when the probability of at least two users generating each preset collaborative intention is determined, the probabilities of generating each preset collaborative intention can be compared to determine the corresponding preset collaborative intention with the highest probability from multiple preset collaborative intentions, and then the content related to the corresponding preset collaborative intention with the highest probability can be pushed directly or through the manufacturer server.

[0054] For example, assuming that multiple preset collaborative intentions include "carpooling", "group buying", and "shared renting", and the probability of at least two users generating a "carpooling" collaborative intention is 0.8, the probability of generating a "group buying" collaborative intention is 0.5, and the probability of generating a "shared renting" collaborative intention is 0.1. Then, it can be determined that the probability of the "carpooling" collaborative intention is the highest, and content related to "carpooling" can be pushed.

[0055] In some embodiments, when the probability of at least two users generating each preset collaborative intention is determined, the content related to the multiple preset collaborative intentions can be pushed in sequence according to the order of the probabilities corresponding to the multiple preset collaborative intentions from high to low. As an implementable manner, when the probability of at least two users generating each preset collaborative intention is determined, the probabilities of generating each preset collaborative intention can be compared to determine the order of the probabilities corresponding to the multiple preset collaborative intentions from high to low, and the content related to the multiple preset collaborative intentions can be pushed in sequence according to the order of the probabilities corresponding to the multiple preset collaborative intentions from high to low.

[0056] For example, assuming that multiple preset collaborative intentions include "carpooling", "group buying", and "shared renting", and the probability of at least two users generating the collaborative intention of "carpooling" is 0.8, the probability of generating the collaborative intention of "group buying" is 0.5, and the probability of generating the collaborative intention of "shared renting" is 0.1. Then, it can be determined that the probabilities of "carpooling", "group buying", and "shared renting" are from high to low, and the content related to "carpooling", the content related to "group buying", and the content related to "shared renting" can be pushed in sequence.

[0057] As an implementable method, when pushing content associated with a collaborative intent corresponding to a first probability, if a confirmation instruction input by the user is received, content associated with a collaborative intent corresponding to a second probability will no longer be pushed, wherein the first probability is greater than the second probability.

[0058] Compared with the intention prediction method provided by an embodiment of the present application, Figure 1 The intention prediction method shown in the figure, this embodiment also pushes content based on the probability that at least two users generate each preset collaborative intention, so that the pushed content is more in line with the common purpose of at least two users, thereby improving the content push effect and user experience.

[0059] See also Figure 3 , Figure 3 The flowchart of the method for predicting intention provided by an embodiment of the present application is shown in FIG. Figure 3 The process shown is described in detail, and the intention prediction method may specifically include the following steps:

[0060] Step S310: Obtain user data corresponding to at least two users, and scene data corresponding to the scenes in which the at least two users are located.

[0061] The specific description of step S310 can be found in step S110 , which will not be repeated here.

[0062] Step S320: When the scene data corresponding to the scenes in which the at least two users are located match, determine the probability that the at least two users have a collaborative intention.

[0063] In this embodiment, when scene data corresponding to the scenes in which at least two users are respectively located are obtained, the scene data corresponding to the scenes in which at least two users are respectively located can be compared to determine whether the scene data corresponding to the scenes in which the at least two users are respectively located match.

[0064] Optionally, if it is determined that the scene data corresponding to the scenes where at least two users are located do not match, it can be determined that there is no probability that the at least two users have a collaborative intention, and therefore, the subsequent steps can be skipped. If it is determined that the scene data corresponding to the scenes where at least two users are located match, it can be determined that there is a probability that the at least two users have a collaborative intention, and therefore, the subsequent steps can be executed.

[0065] Among them, it can be understood that if the scene data corresponding to the scenes where at least two users are located match, it can be considered that different users have certain needs in the same location (such as shopping in the same mall and having a need to take a taxi), or different users have certain needs at the same time point (such as searching for rental houses at the same time), therefore, it can be considered that at least two users have a probability of generating collaborative intentions. If the scene data corresponding to the scenes where at least two users are located do not match, it can be considered that different users have certain needs in different locations (such as one user shopping in a mall in Shenzhen and one user shopping in a mall in Shanghai), or different users have certain needs at different time points (such as one user searching for milk tea group orders in the morning and one user searching for milk tea group orders in the afternoon), therefore, it can be considered that at least two users do not have a probability of generating the same intention.

[0066] In some embodiments, the scene data corresponding to the scenes in which at least two users are located may include: the time corresponding to at least two users. The scene data matching corresponding to the scenes in which at least two users are located may include: the time difference corresponding to at least two users is less than the time difference threshold. That is, if the time difference between the times when different users generate a certain intention is less than the time difference threshold, it can be considered that the time when different users generate a certain intention is very close, there is a probability of generating collaborative intentions between different users, and it can be considered that the scene data corresponding to the scenes in which at least two users are located match.

[0067] In some embodiments, the scene data corresponding to the scenes where at least two users are located may include: the positions corresponding to at least two users. The scene data matching corresponding to the scenes where at least two users are located may include: the distance between the positions corresponding to at least two users is less than a distance threshold. That is, if the distance between the positions where different users generate a certain intention is less than a distance threshold, it can be considered that the positions where different users generate a certain intention are very close, there is a probability of generating collaborative intentions between different users, and it can be considered that the scene data corresponding to the scenes where at least two users are located match.

[0068] In some embodiments, the scene data corresponding to the scenes in which at least two users are located may include: the time and location corresponding to the at least two users. The scene data matching corresponding to the scenes in which at least two users are located may include: the time difference between the times corresponding to the at least two users is less than the time difference threshold, and the distance between the locations is less than the distance threshold. That is, if the time difference between the times when different users generate a certain intention is less than the time difference threshold and the distance between the locations is less than the distance threshold, it can be considered that the time and location when different users generate a certain intention are very close, there is a probability of generating collaborative intentions between different users, and it can be considered that the scene data corresponding to the scenes in which at least two users are located match.

[0069] Step S330: Input the user data corresponding to the at least two users, the scene data corresponding to the scenes in which the at least two users are located, and multiple preset collaborative intentions into the collaborative intention prediction model, and obtain the probability of the at least two users generating each of the multiple preset collaborative intentions output by the collaborative intention prediction model.

[0070] Among them, it can be understood that since the scene data corresponding to the scenes in which at least two users are respectively located match, it can be understood that the scene data corresponding to the scenes in which at least two users are respectively located are the same. When inputting the scene data into the collaborative intention prediction model, the scene data corresponding to the scenes in which at least two users are respectively located can be input into the collaborative intention prediction model, or any one of the scene data corresponding to the scenes in which at least two users are respectively located can be input into the collaborative intention prediction model, and there is no limitation here.

[0071] Compared with the intention prediction method provided by an embodiment of the present application, Figure 1 The intention prediction method shown in the figure, this embodiment also determines the probability that at least two users have a collaborative intention when the scene data corresponding to the scenes in which at least two users are located match, and performs subsequent collaborative intention recognition, thereby making the collaborative intention recognition more accurate and more adapted to the user's scene, thereby improving the user experience.

[0072] See also Figure 4 , Figure 4 The flowchart of the method for predicting intention provided by an embodiment of the present application is shown in FIG. Figure 4 The process shown is described in detail, and the intention prediction method may specifically include the following steps:

[0073] Step S410: Obtain user data corresponding to at least two users, and scene data corresponding to the scenes in which the at least two users are located.

[0074] For the detailed description of step S410, please refer to step S110, which will not be repeated here.

[0075] Step S420: Acquire multiple preset collaboration intentions.

[0076] In this embodiment, multiple preset collaborative intentions may be obtained. Optionally, multiple preset collaborative intentions may be saved after being preset, and accordingly, when performing intention prediction later, the saved multiple preset collaborative intentions may be directly obtained.

[0077] Step S430: Filter out multiple target collaboration intentions from the multiple preset collaboration intentions based on preset rules.

[0078] In some embodiments, preset rules may be pre-set and saved, and the preset rules are used to screen preset collaborative intentions. In this embodiment, when multiple preset collaborative intentions are obtained, multiple target collaborative intentions may be screened out from the multiple preset collaborative intentions based on the preset rules. Among them, the number corresponding to the multiple target collaborative intentions is less than the number corresponding to the multiple preset collaborative intentions. It can be understood that by screening out a smaller number of multiple target collaborative intention parameters from multiple preset collaborative intentions for subsequent prediction of collaborative intentions, the amount of data processing can be effectively reduced and the prediction speed of collaborative intentions can be accelerated.

[0079] That is, in this embodiment, multiple target collaborative intentions can be screened out from multiple preset collaborative intentions through pre-set rules. The rules can be flexibly formulated, for example, based on the user's corresponding behavior data and / or the scene data corresponding to the user's scene, and can be based on other reasonable rules (such as time, date, etc.).

[0080] In some embodiments, multiple target collaborative intentions can be screened out from multiple preset collaborative intentions based on user data corresponding to at least two users. As an implementable method, the adaptability of multiple preset collaborative intentions to the user data corresponding to at least two users can be determined respectively based on the user data corresponding to at least two users, and multiple target collaborative intentions can be screened out from multiple preset collaborative intentions based on the order of adaptability from high to low. For example, assuming that multiple preset collaborative intentions include "carpooling", "group buying", and "shared rent", and the user data corresponding to at least two users represent that the at least two users are shopping in a mall, "carpooling" and "group buying" can be screened as target collaborative intentions.

[0081] In some embodiments, multiple target collaborative intentions can be screened out from multiple preset collaborative intentions based on scene data corresponding to scenes in which at least two users are respectively located. As an implementable method, the degree of adaptability of multiple preset collaborative intentions and user data corresponding to at least two users can be determined respectively based on scene data corresponding to scenes in which at least two users are respectively located, and multiple target collaborative intentions can be screened out from multiple preset collaborative intentions based on the order of adaptability from high to low. For example, assuming that multiple preset collaborative intentions include "carpooling", "group buying", and "shared rent", and the user data corresponding to at least two users respectively represent that the at least two users are in their respective homes, "shared rent" can be screened as the target collaborative intention.

[0082] In some embodiments, multiple target collaborative intentions can be screened out from multiple preset collaborative intentions based on user data corresponding to at least two users and scene data corresponding to the scenes in which they are located. As an implementable method, the adaptability of multiple preset collaborative intentions to the user data corresponding to at least two users can be determined based on the user data corresponding to at least two users and scene data corresponding to the scenes in which they are located, and multiple target collaborative intentions can be screened out from multiple preset collaborative intentions based on the order of adaptability from high to low. For example, assuming that multiple preset collaborative intentions include "carpooling", "group buying", and "shared rent", and the user data corresponding to at least two users represent that the at least two users are shopping in a mall and have just been to a restaurant, "group buying" can be screened as the target collaborative intention.

[0083] Step S440: Input the user data corresponding to the at least two users, the scene data corresponding to the scenes in which the at least two users are located, and multiple target collaborative intentions into the collaborative intention prediction model, and obtain the probability of the at least two users generating each of the multiple target collaborative intentions output by the collaborative intention prediction model.

[0084] In this embodiment, when multiple target collaborative intentions are obtained, the user data corresponding to at least two users, the scene data corresponding to the scenes in which at least two users are located, and multiple target collaborative intentions can be input into the collaborative intention prediction model to obtain the probability of the at least two users generating each of the multiple target collaborative intentions output by the collaborative intention output model. Based on this, only the probability of the at least two users generating each of the multiple selected target collaborative intentions can be determined, thereby reducing the data processing volume of the intention collaborative model and accelerating the prediction efficiency of the collaborative intention.

[0085] Compared with the intention prediction method provided by an embodiment of the present application, Figure 1 The intention prediction method shown in the figure, this embodiment also obtains multiple preset collaborative intentions, and screens multiple target collaborative intentions from the multiple preset collaborative intentions based on preset rules for subsequent traversal, which effectively reduces the data processing volume and can improve the efficiency of intention recognition while ensuring the accuracy of collaborative intention recognition.

[0086] See also Figure 5 , Figure 5 The flowchart of the method for predicting intention provided by an embodiment of the present application is shown in FIG. Figure 5 The process shown is described in detail, and the intention prediction method may specifically include the following steps:

[0087] Step S510: Obtain user data corresponding to at least two users, and scene data corresponding to the scenes in which the at least two users are located.

[0088] For the detailed description of step S510, please refer to step S110, which will not be repeated here.

[0089] Step S520: Obtain the intention data corresponding to each of the at least two users.

[0090] In this embodiment, the intention data corresponding to each of the at least two users may be obtained.

[0091] In some implementations, when at least two users are determined, separate intention predictions may be performed on the at least two users to obtain intention data corresponding to each of the at least two users. As an implementable approach, when at least two users are determined, separate user intention recognition and predictions may be performed on the at least two users based on natural language processing, speech recognition, image recognition, intelligent prediction, and other technologies to obtain intention data corresponding to each of the at least two users.

[0092] As one approach, when user data corresponding to at least two users are obtained, the intention data corresponding to the at least two users can be determined based on the user data corresponding to the at least two users.

[0093] Step S530: Input the user data corresponding to the at least two users, the scene data corresponding to the scenes in which the at least two users are located, the intention data corresponding to the at least two users, and a plurality of preset collaborative intentions into a collaborative intention prediction model, and obtain the probability of the at least two users generating each of the plurality of preset collaborative intentions output by the collaborative intention prediction model.

[0094] In this embodiment, when user data corresponding to at least two users, scene data corresponding to the scenes in which at least two users are located, and intention data corresponding to at least two users are obtained, the user data corresponding to at least two users, scene data corresponding to the scenes in which at least two users are located, intention data corresponding to at least two users, and multiple preset collaborative intentions can be input into a collaborative intention prediction model to obtain the probability of the at least two users generating each of the multiple preset collaborative intentions output by the collaborative intention output model. Based on this, the accuracy of the determined probability of at least two users generating each preset collaborative intention can be improved.

[0095] Compared with the intention prediction method provided by an embodiment of the present application, Figure 1The intention prediction method shown in the figure, this embodiment also obtains the intention data corresponding to at least two users, and inputs the intention data corresponding to at least two users into the intention prediction model to participate in the recognition of the collaborative intention of at least two users, thereby improving the recognition accuracy of the collaborative intention.

[0096] See also Figure 6 , Figure 6 The flowchart of the method for predicting intention provided by an embodiment of the present application is shown in FIG. Figure 6 The process shown is described in detail, and the intention prediction method may specifically include the following steps:

[0097] Step S610: Obtain user training data corresponding to each of the multiple training users when they generate collaborative intentions and scene training data corresponding to the scenes in which they are located.

[0098] like Figure 7 As shown, in this embodiment, user training data corresponding to each of the multiple training users when they generate collaborative intentions and scene training data corresponding to the scenes in which they are located can be obtained.

[0099] As an implementable method, multiple seed users can be selected as multiple training users, and the behavioral data and user portraits corresponding to the multiple training users when they generate collaborative intentions can be obtained. By preprocessing the behavioral data and user portraits corresponding to the multiple training users when they generate collaborative intentions, they can be converted into user training data that meets the specifications.

[0100] Step S620: adding labels to the user training data and the scenario training data based on multiple preset collaborative intentions to obtain a training data set.

[0101] In this embodiment, when multiple user training data corresponding to each of the collaborative intents and scene training data corresponding to each of the scenes are obtained, the user training data and the scene training data can be labeled based on multiple preset collaborative intents to obtain a training data set. That is, when the user training data and the scene training data are obtained, these data can be labeled in combination with the full set of predefined collaborative intents to mark the collaborative intent of the user in a specific scene to form a training data set.

[0102] Each training data in the training data set includes at least user training data corresponding to at least two users, scene training data corresponding to at least two users, and corresponding preset collaborative intent labels. For example, it includes user data corresponding to two users (user1 and user2) (user portraits and behavior data corresponding to each user), and generates collaborative intent (collaborative_intent) in a specific scene (scene).

[0103] In some embodiments, adding labels to user training data and scenario training data based on multiple preset collaborative intentions to obtain a training data set may include: adding a first label to the user training data and scenario training data based on multiple preset collaborative intentions to obtain multiple positive sample data; for each positive sample data in the multiple positive sample data, determining a second label different from the first label from multiple preset collaborative intentions, and adding the second label to the user training data and scenario training data to obtain multiple negative sample data; based on the multiple positive sample data and the multiple negative sample data, obtaining a training data set.

[0104] In some implementations, after labeling the user training data and multiple preset collaborative intentions, a training data set can be formed. After each sample i is encoded (such as one-hot embedding), the following vector representation can be obtained:

[0105] s (i) : The scene information vector of the training data;

[0106] User profile and behavior sequence information of user A in the training data;

[0107] User profile and behavior sequence information of user B in the training data;

[0108] y (i) : Collaborative intent label.

[0109] Let the number of labeled samples be n, then after vector encoding, a supervised data set is formed Then use neural network algorithms (such as MLP, RNN and Transformers, etc.) to learn a user (x A and x B )Probability prediction model for generating a specific collaborative intention y in a specific scenario s:

[0110] M:P(s,x A ,x B ,y)=NeuralNet(s,x A ,x B ,y;Θ)

[0111] Among them, Θ is the neural network parameter to be trained. When training the model, Contains all positive sample data that generate collaborative intentions. For each data sample We can randomly select the collaborative intention set except y(i) Other intentions other than

[0112] Let the number of all positive and negative samples be N. In order to distinguish whether a sample j∈{1,2,3…,N} is a positive sample or a negative sample, a new identifier is introduced

[0113]

[0114] According to this construction, an expanded training data set containing both positive and negative samples can be obtained.

[0115]

[0116] The predicted probability value of the model for the jth sample is Then the negative log-likelihood function can be used as the loss function to update the model parameters:

[0117]

[0118] Step S630: Process the training data set based on a neural network algorithm to obtain a collaborative intention prediction model.

[0119] In this embodiment, when a training data set is obtained, the training data set can be processed based on a neural network algorithm to obtain a collaborative intention prediction model. Optionally, when a training data set is obtained, a neural Wang Kelly algorithm (including a feedforward neural network MLP, a recurrent neural network RNN, and a Transformer, etc.) can be used to train a collaborative intention prediction model.

[0120] Step S640: Obtain user data corresponding to at least two users respectively, and scene data corresponding to the scenes in which the at least two users are respectively located.

[0121] Step S650: Input the user data corresponding to each of the at least two users, the scene data corresponding to the scenes in which the at least two users are located, and a plurality of preset collaborative intentions into a collaborative intention prediction model, and obtain the probability of the at least two users generating each of the plurality of preset collaborative intentions output by the collaborative intention prediction model, wherein the collaborative intention is used to indicate the intention generated by at least two users for associated purposes.

[0122] For the detailed description of step S640 to step S650, please refer to step S110 to step S120, which will not be repeated here.

[0123] Compared with the intention prediction method provided by an embodiment of the present application, Figure 1 The intention prediction method shown in the figure, this embodiment also obtains user training data corresponding to multiple training users when they generate collaborative intentions and scene training data corresponding to their scenes, adds labels to the user training data and scene training data based on multiple preset collaborative intentions, obtains a training data set, processes the training data set based on a neural network algorithm, and obtains a collaborative intention prediction model, thereby improving the accuracy of collaborative intent identification through the collaborative intention prediction model.

[0124] As an example, assume that user 1 has his own unique habits, such as listening to music while walking, often visiting certain shopping malls, often ordering coffee in a certain store, and accustomed to taking taxis during certain time periods, etc. These usage habits will generate behavioral data and calculate the corresponding user portrait. Based on the behavioral data and user portrait, the user data of user 1 can be determined. When it is necessary to calculate the collaborative intention of two users, the user data of the two users can be simultaneously combined with the scene where the current user is located (such as shopping malls) and input into the agreement intention prediction model for calculation. In addition, in order to improve the operation efficiency, the collaborative intention prediction model does not need to calculate the probability of all possible preset collaborative intentions, but only needs to calculate its subset. Therefore, based on the pre-set rules, a part of the preset collaborative intentions can be extracted as candidate subsets, and input into the collaborative intention prediction model for traversal. For example, assuming that the user is shopping in a mall and has just been to a restaurant, in this case, the user's preset collaborative intentions such as "carpooling" and "group ordering" will be adopted as candidate subsets, while "renting together" and "second-hand transactions" will not be in the candidate subsets. Afterwards, the algorithm inputs "carpooling" and "group ordering" as parameters into the collaborative intention prediction model, and performs inference calculations one by one, and finally obtains the probability that two users have collaborative intentions. For example, the probability of "carpooling" is 0.8, and the probability of "group ordering" is 0.5.

[0125] In summary, the intention prediction method proposed in this application can infer user intentions and apply them to multiple user scenarios, improve the coverage of service scenarios and the novelty of service recommendations, expand the applicable scenarios of intent recognition, and provide more service recommendations for downstream, and even meet the user's needs that are currently difficult to achieve.

[0126] Among them, electronic equipment manufacturers can complete the recommendation of services or content by themselves, or they can open interfaces to cooperate with third parties. The third parties can push services or content at the right time by calling the service interface, which can improve service exposure, traffic and conversion.

[0127] For example, when there are scenarios such as "half price for the second cup of milk tea" and "one yuan more for a bottle of Coke", it is often difficult for a single consumer to complete the group order because there may not be another user who needs to buy the product at the same time, or it may be embarrassing to find someone on the spot. However, the collaborative intent solution can link users in similar time and space, allowing users to group order and realize their needs that were originally difficult to achieve.

[0128] For another example, when the system predicts that the user intends to carpool, the system can push a carpool reminder. For users, there are at least the following benefits: First, it is easy to realize the demand: users do not necessarily choose carpool when using taxi apps, and may directly take a special car because the waiting time for carpooling may be too long. If at this time, it is calculated based on the model algorithm that there are users nearby who already have carpooling needs, the waiting time will be reduced, and carpooling will also reduce the user's taxi expenses. Second, the operation interface is simpler: the carpooling interface can be exposed to users in a more concise way, and users can click directly to enter without having to jump into the application, such as Figure 8 For merchants, there are at least the following benefits: First, increase the exposure of the enterprise: many non-head taxi companies that provide carpooling services can also be exposed; second, increase the exposure of functions: the functions that the enterprise hopes to promote (in this case, carpooling) can be promoted to users through the card entrance.

[0129] See also Fig. 9 , Fig. 9 The module block diagram of the intention prediction device provided by an embodiment of the present application is shown. The intention prediction device is applied to the above-mentioned server. Fig. 9 As shown in the block diagram, the intention prediction device 200 includes: a data acquisition module 210 and an intention prediction module 220, wherein:

[0130] The data acquisition module 210 is used to acquire user data corresponding to at least two users and scene data corresponding to the scenes in which the at least two users are located.

[0131] Furthermore, the data acquisition module 210 includes: a first data acquisition submodule and a second data acquisition submodule, wherein:

[0132] The first data acquisition submodule is used to acquire the behavior data corresponding to each of the at least two users, and to acquire the user portraits corresponding to each of the at least two users.

[0133] The second data acquisition submodule is used to pre-process the behavior data corresponding to each of the at least two users and the user portraits corresponding to each of the at least two users to obtain the user data corresponding to each of the at least two users.

[0134] The intention prediction module 220 is used to input the user data corresponding to each of the at least two users, the scene data corresponding to the scenes in which the at least two users are located, and a plurality of preset collaborative intentions into a collaborative intention prediction model, and obtain the probability of the at least two users generating each of the plurality of preset collaborative intentions output by the collaborative intention prediction model, wherein the collaborative intention is used to indicate the intention generated by at least two users due to related purposes.

[0135] Furthermore, the intention prediction module 220 includes: a collaborative intention determination submodule and a first intention prediction submodule, wherein:

[0136] The collaborative intention determination submodule is used to determine the probability that the at least two users have a collaborative intention when the scene data corresponding to the scenes in which the at least two users are respectively located match.

[0137] The first intention prediction submodule is used to input user data corresponding to each of the at least two users, scene data corresponding to the scenes in which the at least two users are located, and multiple preset collaborative intentions into a collaborative intention prediction model, and obtain the probability of the at least two users generating each of the multiple preset collaborative intentions output by the collaborative intention prediction model.

[0138] Furthermore, the intention prediction module 220 includes: a collaborative intention acquisition submodule, a collaborative intention screening submodule and a second intention prediction submodule, wherein:

[0139] The collaborative intention acquisition submodule is used to acquire multiple preset collaborative intentions.

[0140] The collaborative intention screening submodule is used to screen out a plurality of target collaborative intentions from the plurality of preset collaborative intentions based on preset rules.

[0141] Furthermore, the collaborative intention screening submodule includes: a first collaborative intention screening submodule and a second collaborative intention screening submodule, wherein:

[0142] The first collaborative intention screening submodule is used to screen out the multiple target collaborative intentions from the multiple preset collaborative intentions based on the user data corresponding to each of the at least two users.

[0143] The second collaborative intention screening submodule is used to screen out the multiple target collaborative intentions from the multiple preset collaborative intentions based on the scene data corresponding to the scenes in which the at least two users are respectively located.

[0144] The second intention prediction submodule is used to input user data corresponding to each of the at least two users, scene data corresponding to the scenes in which the at least two users are located, and multiple target collaborative intentions into a collaborative intention prediction model, and obtain the probability of the at least two users generating each of the multiple target collaborative intentions output by the collaborative intention prediction model.

[0145] Furthermore, the intention prediction module 220 includes: an intention data acquisition submodule and a third intention prediction submodule, wherein:

[0146] The intention data acquisition submodule is used to obtain the intention data corresponding to each of the at least two users.

[0147] The third intention prediction submodule is used to input user data corresponding to each of the at least two users, scene data corresponding to the scenes in which the at least two users are located, intention data corresponding to each of the at least two users, and multiple preset collaborative intentions into a collaborative intention prediction model, and obtain the probability of the at least two users generating each of the multiple preset collaborative intentions output by the collaborative intention prediction model.

[0148] Furthermore, the intention prediction device 200 further includes: a content push module, wherein:

[0149] The content push module is used to push content based on the probability that the at least two users generate the preset collaborative intentions.

[0150] Furthermore, the content push module includes: a first content push submodule and a second content push submodule, wherein:

[0151] The first content push submodule is used to determine a corresponding preset collaborative intention with the highest probability from the multiple preset collaborative intentions, and push content related to the corresponding preset collaborative intention with the highest probability.

[0152] The second content push submodule is used to push the contents related to the multiple preset collaborative intentions in sequence according to the probability corresponding to the multiple preset collaborative intentions from high to low.

[0153] Furthermore, the intention prediction device 200 further includes: a training data acquisition module, a training data set acquisition module and a module training module, wherein:

[0154] The training data acquisition module is used to obtain the user training data corresponding to each of the multiple training users when they generate collaborative intentions and the scene training data corresponding to the scenes they are in.

[0155] The training data set acquisition module is used to add labels to the user training data and the scenario training data based on multiple preset collaborative intentions to obtain a training data set.

[0156] Furthermore, the training data set acquisition module includes: a positive sample data acquisition submodule, a negative sample data acquisition submodule and a training data set acquisition submodule, wherein:

[0157] The positive sample data acquisition submodule is used to add a first label to the user training data and the scenario training data based on the multiple preset collaborative intentions to obtain multiple positive sample data.

[0158] The negative sample data acquisition submodule is used to determine, for each positive sample data among the multiple positive sample data, a second label different from the first label from the multiple preset collaborative intentions, and add the second label to the user training data and the scenario training data to obtain multiple negative sample data.

[0159] The training data set acquisition submodule is used to acquire the training data set based on the multiple positive sample data and the multiple negative sample data.

[0160] The model training module is used to process the training data set based on a neural network algorithm to obtain a collaborative intention prediction model.

[0161] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described devices and modules can refer to the corresponding processes in the aforementioned method embodiments, and will not be repeated here.

[0162] In several embodiments provided in the present application, the coupling between modules may be electrical, mechanical or other forms of coupling.

[0163] In addition, each functional module in each embodiment of the present application can be integrated into a processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The above integrated modules can be implemented in the form of hardware or software functional modules.

[0164] See also Fig.10, which shows a structural block diagram of a server 100 provided in an embodiment of the present application. The server 100 may be a server capable of running applications, such as a smart phone, a tablet computer, an e-book, etc. The server 100 in the present application may include one or more of the following components: a processor 110, a memory 120, and one or more applications, wherein the one or more applications may be stored in the memory 120 and configured to be executed by one or more processors 110, and the one or more programs are configured to execute the method described in the aforementioned method embodiment.

[0165] Among them, the processor 110 may include one or more processing cores. The processor 110 uses various interfaces and lines to connect various parts within the entire server 100, and executes various functions and processes data of the server 100 by running or executing instructions, programs, code sets or instruction sets stored in the memory 120, and calling data stored in the memory 120. Optionally, the processor 110 can be implemented in at least one hardware form of digital signal processing (Digital Signal Processing, DSP), field programmable gate array (Field-Programmable Gate Array, FPGA), and programmable logic array (Programmable Logic Array, PLA). The processor 110 can integrate one or a combination of a central processing unit (Central Processing Unit, CPU), a graphics processing unit (Graphics Processing Unit, GPU) and a modem. Among them, the CPU mainly processes the operating system, user interface and application programs; the GPU is responsible for rendering and drawing the content to be displayed; and the modem is used to process wireless communications. It can be understood that the above-mentioned modem may not be integrated into the processor 110, but may be implemented separately through a communication chip.

[0166] The memory 120 may include a random access memory (RAM) or a read-only memory (ROM). The memory 120 may be used to store instructions, programs, codes, code sets or instruction sets. The memory 120 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for implementing functions (such as a touch function, a sound playback function, an image playback function, etc.), instructions for implementing the following various method embodiments, etc. The data storage area may also store data (such as a phone book, audio and video data, chat record data) created by the server 100 during use.

[0167] See also Fig.11, which shows a structural block diagram of a computer-readable storage medium provided in an embodiment of the present application. The computer-readable medium 300 stores program codes, which can be called by a processor to execute the method described in the above method embodiment.

[0168] The computer readable storage medium 300 may be an electronic memory such as a flash memory, an EEPROM (electrically erasable programmable read-only memory), an EPROM, a hard disk, or a ROM. Optionally, the computer readable storage medium 300 includes a non-transitory computer-readable storage medium. The computer readable storage medium 300 has storage space for program code 310 that performs any method steps in the above method. These program codes can be read from or written to one or more computer program products. The program code 310 can be compressed, for example, in an appropriate form.

[0169] In summary, the intention prediction method, device, server and storage medium provided in the embodiments of the present application obtain user data corresponding to at least two users and scene data corresponding to the scenes in which at least two users are respectively located, and input the user data corresponding to at least two users, the scene data corresponding to the scenes in which at least two users are respectively located, and multiple preset collaborative intentions into a collaborative intention prediction model to obtain the probability of each preset collaborative intention among the multiple preset collaborative intentions generated by the at least two users output by the collaborative intention prediction model, wherein the collaborative intention is used to indicate the intention generated by at least two users due to related purposes, thereby defining multiple collaborative intentions and identifying the probability of different users generating each collaborative intention through multi-user data to improve the scenarios in which subsequent content push can be applied.

[0170] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for predicting intention, characterized in that: The method comprises: Acquire user data corresponding to at least two users respectively, and scene data corresponding to the scenes in which the at least two users respectively are located; The user data corresponding to each of the at least two users, the scene data corresponding to the scenes in which the at least two users are located, and a plurality of preset collaborative intentions are input into a collaborative intention prediction model, and the probability of the at least two users generating each of the plurality of preset collaborative intentions is obtained from the collaborative intention prediction model output, wherein the collaborative intention is used to indicate the intention generated by at least two users due to associated purposes.

2. The method according to claim 1, characterized in that After obtaining the probability of the at least two users generating each of the plurality of preset collaborative intentions output by the collaborative intention prediction model, the method further includes: Content is pushed based on the probability that the at least two users generate the preset collaborative intentions.

3. The method according to claim 2, characterized in that The pushing of content based on the probability of the at least two users generating the preset collaboration intentions includes: Determine a corresponding preset collaborative intent with the highest probability from the multiple preset collaborative intents, and push content related to the corresponding preset collaborative intent with the highest probability; or Contents related to the multiple preset collaborative intentions are pushed in sequence according to the order of probabilities corresponding to the multiple preset collaborative intentions from high to low.

4. The method according to claim 1, characterized in that: The step of inputting user data corresponding to each of the at least two users, scene data corresponding to the scenes in which the at least two users are located, and a plurality of preset collaborative intentions into a collaborative intention prediction model, and obtaining the probability of each of the plurality of preset collaborative intentions generated by the at least two users output by the collaborative intention prediction model comprises: In a case where the scene data corresponding to the scenes in which the at least two users are located respectively match, determining a probability that the at least two users have a collaborative intention; The user data corresponding to the at least two users, the scene data corresponding to the scenes in which the at least two users are located, and a plurality of preset collaborative intentions are input into a collaborative intention prediction model to obtain the probability of the at least two users generating each of the plurality of preset collaborative intentions output by the collaborative intention prediction model.

5. The method according to claim 4, characterized in that The scene data corresponding to the scenes where the at least two users are located are matched, including: The time difference between the times corresponding to the at least two users is less than the time difference threshold; and / or The distance between the positions corresponding to the at least two users is less than a distance threshold.

6. The method according to claim 1, characterized in that The step of inputting user data corresponding to each of the at least two users, scene data corresponding to the scenes in which the at least two users are located, and a plurality of preset collaborative intentions into a collaborative intention prediction model, and obtaining the probability of each of the plurality of preset collaborative intentions generated by the at least two users output by the collaborative intention prediction model comprises: Get multiple preset collaborative intentions; Filtering out a plurality of target collaboration intentions from the plurality of preset collaboration intentions based on preset rules; The user data corresponding to the at least two users, the scene data corresponding to the scenes in which the at least two users are located, and multiple target collaborative intentions are input into a collaborative intention prediction model to obtain the probability of the at least two users generating each of the multiple target collaborative intentions output by the collaborative intention prediction model.

7. The method according to claim 6, characterized in that The step of selecting a plurality of target collaboration intentions from the plurality of preset collaboration intentions based on a preset rule includes: Filtering the multiple target collaboration intentions from the multiple preset collaboration intentions based on the user data corresponding to the at least two users; and / or The multiple target collaborative intentions are filtered out from the multiple preset collaborative intentions based on the scene data corresponding to the scenes in which the at least two users are respectively located.

8. The method according to claim 1, characterized in that The step of inputting user data corresponding to each of the at least two users, scene data corresponding to the scenes in which the at least two users are located, and a plurality of preset collaborative intentions into a collaborative intention prediction model, and obtaining the probability of each of the plurality of preset collaborative intentions generated by the at least two users output by the collaborative intention prediction model comprises: Acquire intention data corresponding to each of the at least two users; The user data corresponding to the at least two users, the scene data corresponding to the scenes in which the at least two users are located, the intention data corresponding to the at least two users, and a plurality of preset collaborative intentions are input into a collaborative intention prediction model to obtain the probability of the at least two users generating each of the plurality of preset collaborative intentions output by the collaborative intention prediction model.

9. The method according to claim 1, characterized in that: The obtaining of user data corresponding to at least two users respectively includes: Obtaining behavior data corresponding to each of the at least two users, and obtaining user portraits corresponding to each of the at least two users; The behavior data corresponding to each of the at least two users and the user portraits corresponding to each of the at least two users are preprocessed to obtain user data corresponding to each of the at least two users.

10. The method according to any one of claims 1 to 9, characterized in that: Before acquiring the user data corresponding to the at least two users and the scene data corresponding to the scenes in which the at least two users are located, the method further includes: Obtaining user training data corresponding to each of the multiple training users when they generate a collaborative intention and scene training data corresponding to the scenes in which they are located; Adding labels to the user training data and the scenario training data based on multiple preset collaborative intentions to obtain a training data set; The training data set is processed based on a neural network algorithm to obtain a collaborative intention prediction model.

11. The method according to claim 10, characterized in that The step of adding labels to the user training data and the scenario training data based on a plurality of preset collaborative intentions to obtain a training data set includes: Adding a first label to the user training data and the scenario training data based on the multiple preset collaborative intentions to obtain multiple positive sample data; For each positive sample data among the plurality of positive sample data, determine a second label different from the first label from the plurality of preset collaborative intentions, and add the second label to the user training data and the scenario training data to obtain a plurality of negative sample data; The training data set is obtained based on the multiple positive sample data and the multiple negative sample data.

12. An intention prediction device, characterized in that: The device comprises: A data acquisition module, used to acquire user data corresponding to at least two users respectively, and scene data corresponding to the scenes in which the at least two users respectively are located; An intention prediction module is used to input user data corresponding to each of the at least two users, scene data corresponding to the scenes in which the at least two users are located, and multiple preset collaborative intentions into a collaborative intention prediction model, and obtain the probability of the at least two users generating each of the multiple preset collaborative intentions output by the collaborative intention prediction model, wherein the collaborative intention is used to indicate the intention generated by at least two users due to related purposes.

13. A server, characterized in that: The method comprises a memory and a processor, wherein the memory is coupled to the processor, the memory stores instructions, and when the instructions are executed by the processor, the processor executes the method according to any one of claims 1 to 10.

14. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores program codes, which can be called by a processor to execute the method according to any one of claims 1 to 10.

Images