Interaction method and device based on large model, medium and product thereof

Abstract

The invention provides an interaction method and device based on a large model, a medium and a product thereof, and the method comprises the steps: carrying out the interaction based on at least one first virtual model and a plurality of second virtual models, and generating a plurality of interaction records; scoring the interaction popularity of the plurality of interaction records based on a preset large model, and determining a target exchange record which is the interaction record with the highest interaction popularity in the plurality of interaction records; and pushing target account information to the target user, wherein the target account information is account information of other users corresponding to the target interaction record. According to the method and the device, the virtual model interacts with the virtual models of other users, so that the interaction record with the highest popularity is screened out, the target account information most suitable for the target user is determined, the target account information is pushed to the target user, and the interaction effect is improved.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of artificial intelligence, and in particular to a large model-based interaction method, device, medium and product thereof. Background Art

[0002] With the continuous advancement of artificial intelligence technology, the application of agents in the social field is becoming more and more extensive. Due to the limitations of traditional dating platforms in matching users' emotional needs, they mainly rely on one-on-one communication with real people, which not only limits the scope of dating, but also increases the time cost. In addition, real-person communication cannot determine whether the other party is suitable for oneself, resulting in poor interaction effects in existing technologies. Summary of the invention

[0003] The embodiments of the present application provide a large-model-based interaction method, device, medium, and product thereof to solve the problem of poor interaction effect in the prior art.

[0004] To solve the above problems, this application is implemented as follows:

[0005] In a first aspect, an embodiment of the present application provides an interaction method based on a large model, the method comprising:

[0006] Generate multiple interaction records based on interaction between at least one first virtual model and multiple second virtual models, wherein the at least one first virtual model is generated based on a target user, the multiple second virtual models are generated based on multiple other users and correspond one-to-one with the other users, and the multiple interaction records correspond one-to-one with the multiple second virtual models;

[0007] Scoring the interaction heat of the multiple interaction records based on a preset large model, and determining a target exchange record, wherein the target interaction record is an interaction record with the highest interaction heat among the multiple interaction records;

[0008] Push target account information to the target user, where the target account information is account information of other users corresponding to the target interaction record.

[0009] Optionally, before generating a plurality of interaction records based on the interaction between the at least one first virtual model and the plurality of second virtual models, the method further includes:

[0010] Acquire personal information of the target user, wherein the personal information includes at least one of the age, personality, hobbies and living habits of the target user;

[0011] generating at least one prompt instruction according to the personal information of the target user, wherein different prompt instructions correspond to different weights in the personal information of the target user;

[0012] The at least one first virtual model is generated according to the at least one prompt instruction and the preset template, and the at least one prompt instruction corresponds to the at least one first virtual model in a one-to-one manner.

[0013] Optionally, the generating a plurality of interaction records based on the interaction between the at least one first virtual model and the plurality of second virtual models includes:

[0014] When it is detected that the target user has posted a personal status, extracting information from the personal status to determine at least one keyword;

[0015] updating the at least one first virtual model according to the at least one keyword to obtain at least one third virtual model;

[0016] The plurality of interaction records are generated based on the interaction between the at least one third virtual model and the plurality of second virtual models.

[0017] Optionally, the generating a plurality of interaction records based on the interaction between the at least one first virtual model and the plurality of second virtual models includes:

[0018] Obtaining target interaction requirements of the target user;

[0019] According to the target interaction requirement, matching is performed among multiple candidate users to determine the multiple other users, whose personal information meets the target interaction requirement;

[0020] The plurality of interaction records are generated based on the interaction between the at least one third virtual model and the plurality of second virtual models corresponding to the plurality of other users.

[0021] Optionally, before scoring the interaction heat of the plurality of interaction records based on a preset large model and determining the target exchange record, the method further includes:

[0022] Obtaining an initial large model and a training data set, wherein the initial large model is used to score text content, and the training data set includes a plurality of training texts and a plurality of training labels corresponding to the plurality of training texts;

[0023] Using the training data set to train the initial large model, outputting multiple prediction values;

[0024] Calculate a loss function based on the multiple predicted values ​​and the multiple training labels to obtain a loss value;

[0025] The initial large model is updated according to the loss value to obtain the preset large model.

[0026] Optionally, the pushing the target account information to the target user includes:

[0027] Determine a target other user according to the target exchange record, where the target other user is the other user corresponding to the target exchange record;

[0028] Query the account information of the target other user to obtain the target account information;

[0029] The target account information is pushed to the target user.

[0030] In a second aspect, an embodiment of the present application further provides an interactive device based on a large model, the device comprising:

[0031] An interaction module, configured to interact with a plurality of second virtual models based on at least one first virtual model to generate a plurality of interaction records, wherein the at least one first virtual model is generated based on a target user, the plurality of second virtual models are generated based on a plurality of other users and correspond one-to-one with the other users, and the plurality of interaction records correspond one-to-one with the plurality of second virtual models;

[0032] A scoring module, configured to score the interaction heat of the plurality of interaction records based on a preset large model, and determine a target exchange record, wherein the target interaction record is the interaction record with the highest interaction heat among the plurality of interaction records;

[0033] The push module is used to push target account information to the target user, where the target account information is the account information of other users corresponding to the target interaction record.

[0034] In a third aspect, an embodiment of the present application further provides a communication device, comprising: a transceiver, a memory, a processor, and a program stored in the memory and executable on the processor; the processor is used to read the program in the memory to implement the steps in the method described in the first aspect above.

[0035] In a fourth aspect, an embodiment of the present application further provides a readable storage medium for storing a program, which, when executed by a processor, implements the steps in the method described in the first aspect above.

[0036] In a fifth aspect, an embodiment of the present application further provides a computer program product, which is stored in a storage medium and is executed by at least one processor to implement the steps in the method described in the first aspect.

[0037] The present application provides an interaction method, device, medium and product based on a large model, the method comprising: generating multiple interaction records based on at least one first virtual model interacting with multiple second virtual models, wherein the at least one first virtual model is generated based on a target user, the multiple second virtual models are generated based on multiple other users and correspond one-to-one with the other users, and the multiple interaction records correspond one-to-one with the multiple second virtual models; scoring the interaction heat of the multiple interaction records based on a preset large model, determining a target exchange record, wherein the target interaction record is the interaction record with the highest interaction heat among the multiple interaction records; and pushing target account information to the target user, wherein the target account information is the account information of other users corresponding to the target interaction record. The present application uses a virtual model to interact with virtual models of other users, thereby screening out the interaction record with the highest heat, thereby determining the target account information most suitable for the target user, and pushing it to the target user, thereby improving the effect of the interaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the description of the embodiments of the present application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0039] Figure 1 It is a flowchart of the interaction method based on the large model provided in the embodiment of the present application;

[0040] Figure 2 It is a structural schematic diagram of an interactive device based on a large model provided in an embodiment of the present application;

[0041] Figure 3 It is a schematic diagram of the structure of an electronic device provided by the implementation of this application. DETAILED DESCRIPTION

[0042] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0043] The terms "first", "second" etc. in the embodiments of the present application are used to distinguish similar objects, and need not be used to describe a specific order or sequential order. In addition, the terms "include" and "have" and any variation thereof are intended to cover non-exclusive inclusions, for example, the process, method, system, product or equipment comprising a series of steps or units need not be limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or equipment. In addition, "and / or" is used in the present application to represent at least one of connected objects, such as A and / or B and / or C, representing the inclusion of separate A, separate B, separate C, and A and B all exist, B and C all exist, A and C all exist, and 7 situations that A, B and C all exist.

[0044] See also Figure 1 , Figure 3 is a flow chart of the interactive method based on a large model provided in the embodiment of the present application. Figure 1 As shown, the interactive method based on the large model may include the following steps:

[0045] Step 101: Generate multiple interaction records based on interaction between at least one first virtual model and multiple second virtual models, wherein the at least one first virtual model is generated based on a target user, the multiple second virtual models are generated based on multiple other users and correspond one-to-one with the other users, and the multiple interaction records correspond one-to-one with the multiple second virtual models.

[0046] In this embodiment, the virtual model is an agent. Specifically, in AI, an agent can be a software program or a robot that can perform tasks, learn, and adapt to the environment. For example, chatbots, self-driving cars, and NPCs (non-player characters) in games can all be considered agents. In this embodiment, the virtual model can chat and interact with other virtual models according to the user's settings, that is, simulate chatting between real users.

[0047] Specifically, the first virtual model is generated based on the target user, and the plurality of second virtual models are generated based on a plurality of different other users and correspond one to one. Exemplarily, the user may input a personal description, thereby generating a plurality of virtual models of different styles.

[0048] Step 102: Score the interaction heat of the multiple interaction records based on a preset large model to determine a target exchange record, where the target interaction record is the interaction record with the highest interaction heat among the multiple interaction records.

[0049] In this embodiment, the preset large model may be a semantic large model, which generally refers to a natural language processing model based on deep learning, which has a strong ability to understand and generate language. They are usually trained by processing a large amount of text data to capture the semantic information, contextual relationship and grammatical structure of the language.

[0050] By presetting a large model, the interaction record with the highest interaction heat among multiple interaction records can be accurately determined, and the most suitable other users for the target user can be determined through the interaction record with the highest interaction heat. Subsequently, the target user can directly chat with the most suitable other users through real people, thereby improving the efficiency of making friends.

[0051] Step 103: Push target account information to the target user, where the target account information is account information of other users corresponding to the target interaction record.

[0052] In this embodiment, after acquiring the most suitable other users, data analysis is performed based on the background to determine the most suitable target account information of the other users, thereby pushing the target account information to the target user, thereby further enriching the dating experience and helping users quickly screen out the most popular dating partners.

[0053] The present application provides an interaction method based on a large model, the method comprising: generating multiple interaction records based on at least one first virtual model interacting with multiple second virtual models, wherein the at least one first virtual model is generated based on a target user, the multiple second virtual models are generated based on multiple other users and correspond one-to-one with the other users, and the multiple interaction records correspond one-to-one with the multiple second virtual models; scoring the interaction heat of the multiple interaction records based on a preset large model, determining a target exchange record, wherein the target interaction record is the interaction record with the highest interaction heat among the multiple interaction records; and pushing target account information to the target user, wherein the target account information is the account information of other users corresponding to the target interaction record. The present application uses a virtual model to interact with virtual models of other users, thereby screening out the interaction record with the highest heat, thereby determining the target account information that is most suitable for the target user, and pushing it to the target user, thereby improving the effect of the interaction.

[0054] In some feasible implementations, optionally, before generating multiple interaction records based on the interaction between the at least one first virtual model and the multiple second virtual models, the method further includes:

[0055] Acquire personal information of the target user, wherein the personal information includes at least one of the age, personality, hobbies and living habits of the target user;

[0056] generating at least one prompt instruction according to the personal information of the target user, wherein different prompt instructions correspond to different weights in the personal information of the target user;

[0057] The at least one first virtual model is generated according to the at least one prompt instruction and the preset template, and the at least one prompt instruction corresponds to the at least one first virtual model in a one-to-one manner.

[0058] In this embodiment, at least one first virtual model is generated through the user's personal information, wherein the personal information includes age, personality, hobbies and living habits, etc., which are not specifically limited in this embodiment. At least one prompt instruction prompt is generated through the user's personal information, and then the at least one first virtual model is generated according to the prompt instruction prompt and the agent template, that is, multiple agents that meet the user's characteristics are created, and different styles can also be selected to obtain multiple agents.

[0059] It should be noted that different prompt instructions have different weights in the target user's personal information. For example, the first agent focuses on the user's age, and during the agent's chat, relevant conversation content will be generated based on the user's age. The second agent focuses on the user's personality, so during the agent's chat, the user's personality generates relevant conversation content.

[0060] In addition, during the interaction between the first virtual model and the second virtual model, the second virtual model will also focus on the performance based on the personal information of other users, thereby better screening out dating partners that are more compatible with the target user.

[0061] Optionally, the generating a plurality of interaction records based on the interaction between the at least one first virtual model and the plurality of second virtual models includes:

[0062] When it is detected that the target user has posted a personal status, extracting information from the personal status to determine at least one keyword;

[0063] updating the at least one first virtual model according to the at least one keyword to obtain at least one third virtual model;

[0064] The plurality of interaction records are generated based on the interaction between the at least one third virtual model and the plurality of second virtual models.

[0065] In this embodiment, when it is detected that the target user publishes personal status, such as weather, mood, location or pictures, this information will be synchronized to the user's agent in real time. Specifically, by extracting information from the personal status, at least one keyword is determined, which can be implemented by a feature extraction model, wherein the feature extraction model is a model used in machine learning and deep learning to extract useful information or features from raw data. These features are usually used for subsequent classification, regression or other analysis tasks. The purpose of feature extraction is to simplify high-dimensional and complex data into a form that is easier to process while retaining important information.

[0066] After extracting at least one keyword through the feature extraction model, the at least one first virtual model is updated through the at least one keyword to obtain at least one third virtual model, thereby interacting with the multiple second virtual models through the at least one third virtual model to generate the multiple interaction records.

[0067] In this embodiment, the user can update his / her personal status at any time, so that the agent can more realistically reflect the user's dynamic life information and enhance the authenticity and interactivity of friendship.

[0068] Optionally, the generating a plurality of interaction records based on the interaction between the at least one first virtual model and the plurality of second virtual models includes:

[0069] Obtaining target interaction requirements of the target user;

[0070] According to the target interaction requirement, matching is performed among multiple candidate users to determine the multiple other users, whose personal information meets the target interaction requirement;

[0071] The plurality of interaction records are generated based on the interaction between the at least one third virtual model and the plurality of second virtual models corresponding to the plurality of other users.

[0072] In this embodiment, the target user may generate a target interaction requirement in advance before interacting. Specifically, the target interaction requirement is a user portrait with which the target user wishes to interact, that is, the user's age, specific personality, etc. can all serve as interaction requirements.

[0073] After determining the target interaction requirements, multiple candidate users in the database are matched, and multiple other users that meet the target interaction requirements are selected, thereby generating multiple interaction records by interacting with multiple second virtual models corresponding to the multiple other users through at least one third virtual model. In this way, the interaction object that meets the user's requirements can be better selected.

[0074] Optionally, before scoring the interaction heat of the plurality of interaction records based on a preset large model and determining the target exchange record, the method further includes:

[0075] Obtaining an initial large model and a training data set, wherein the initial large model is used to score text content, and the training data set includes a plurality of training texts and a plurality of training labels corresponding to the plurality of training texts;

[0076] Using the training data set to train the initial large model, outputting multiple prediction values;

[0077] Calculate a loss function based on the multiple predicted values ​​and the multiple training labels to obtain a loss value;

[0078] The initial large model is updated according to the loss value to obtain the preset large model.

[0079] In this embodiment, the preset big model is a semantic big model, and the initial semantic big model needs to be trained through training data to finally obtain a trained semantic big model. Specifically, a training data set is first obtained, wherein the training data set includes a plurality of training texts, and each training text may include a text content. In addition, the training text also includes a plurality of training tags, wherein the content included in each training tag is a semantic representation of the corresponding text content, exemplarily, such as the meaning expressed by the text content or the main features of the text content, etc., which are not specifically limited in this embodiment and can be adaptively selected according to actual conditions.

[0080] The initial large model is trained using a training data set, thereby outputting multiple predicted values ​​corresponding to the initial large model, each predicted value being the output of the initial large model corresponding to the training data set. The loss function is an important tool for evaluating the difference between the model prediction results and the actual results in machine learning and deep learning. Its main function is to quantify the prediction error of the model in order to guide the optimization of the model during the training process. By calculating the loss function for multiple predicted values ​​and corresponding multiple training labels, a loss value is generated, and the relevant parameters in the initial large model are updated by the loss value, and finally a preset large model is obtained. By adopting a large model as the core technology, the present invention can more accurately match the emotional needs of users and improve the efficiency and success rate of making friends.

[0081] Optionally, the pushing the target account information to the target user includes:

[0082] Determine a target other user according to the target exchange record, where the target other user is the other user corresponding to the target exchange record;

[0083] Query the account information of the target other user to obtain the target account information;

[0084] The target account information is pushed to the target user.

[0085] In this embodiment, after determining the target other user according to the target exchange record, the account information of the target other user is determined by querying, thereby accurately determining the account used by the target other user. The user selects the real person mode to invite the other party to interact in person, or the system sends a real person interaction invitation to both parties. After the invitation is accepted, the two parties will conduct subsequent communication and exchanges as real people. Through the setting of the real person interaction mode, users are provided with a more flexible and diverse way of making friends, meeting the needs of different users.

[0086] This application uses a virtual model to interact with other users' virtual models, thereby screening out the most popular interaction records, thereby determining the most suitable target account information for the target user, and pushing it to the target user, thereby improving the effect of the interaction.

[0087] See also Figure 2 , Figure 2 is a structural diagram of an interactive device based on a large model provided in an embodiment of the present application. Figure 2 As shown, the interactive device 500 based on the large model includes:

[0088] An interaction module 210 is configured to interact with a plurality of second virtual models based on at least one first virtual model to generate a plurality of interaction records, wherein the at least one first virtual model is generated based on a target user, the plurality of second virtual models are generated based on a plurality of other users and correspond one-to-one with the other users, and the plurality of interaction records correspond one-to-one with the plurality of second virtual models;

[0089] A scoring module 220, configured to score the interaction heat of the plurality of interaction records based on a preset large model, and determine a target exchange record, wherein the target interaction record is the interaction record with the highest interaction heat among the plurality of interaction records;

[0090] The push module 230 is used to push target account information to the target user, where the target account information is account information of other users corresponding to the target interaction record.

[0091] Optionally, also include:

[0092] An acquisition module, used to acquire personal information of the target user, wherein the personal information includes at least one of the age, personality, hobbies and living habits of the target user;

[0093] A first generating module, configured to generate at least one prompt instruction according to the personal information of the target user, wherein different prompt instructions correspond to different weights in the personal information of the target user;

[0094] The second generating module is used to generate the at least one first virtual model according to the at least one prompt instruction and the preset template, and the at least one prompt instruction and the at least one first virtual model have a one-to-one correspondence.

[0095] Optionally, the interaction module 210 includes:

[0096] An extraction submodule, configured to extract information from the personal status and determine at least one keyword when detecting that the target user has posted a personal status;

[0097] an updating submodule, configured to update the at least one first virtual model according to the at least one keyword to obtain at least one third virtual model;

[0098] The interaction submodule is configured to generate the plurality of interaction records based on the interaction between the at least one third virtual model and the plurality of second virtual models.

[0099] Optionally, the interaction submodules include:

[0100] An acquisition unit, configured to acquire a target interaction requirement of the target user;

[0101] A matching unit, configured to match multiple candidate users according to the target interaction requirement, and determine the multiple other users, whose personal information meets the target interaction requirement;

[0102] The interaction unit is configured to generate the plurality of interaction records based on the interaction between the at least one third virtual model and the plurality of second virtual models corresponding to the plurality of other users.

[0103] Optionally, also include:

[0104] A model acquisition module, used to acquire an initial large model and a training data set, wherein the initial large model is used to score text content, and the training data set includes a plurality of training texts and a plurality of training labels corresponding to the plurality of training texts;

[0105] A model training module, used to train the initial large model using the training data set and output a plurality of prediction values;

[0106] A model calculation module, used to calculate a loss function according to the multiple predicted values ​​and the multiple training labels to obtain a loss value;

[0107] The model updating module is used to update the initial large model according to the loss value to obtain the preset large model.

[0108] Optionally, the push module 230 includes:

[0109] A determination submodule, configured to determine a target other user according to the target exchange record, wherein the target other user is another user corresponding to the target exchange record;

[0110] A query submodule, used to query the logged-in account information of the target other user to obtain the target account information;

[0111] The push submodule is used to push the target account information to the target user.

[0112] This application uses a virtual model to interact with other users' virtual models, thereby screening out the most popular interaction records, thereby determining the most suitable target account information for the target user, and pushing it to the target user, thereby improving the effect of the interaction.

[0113] The present application also provides an electronic device. Figure 3 , the electronic device may include a processor 301, a memory 302, and a program 3021 stored in the memory 302 and executable on the processor 301.

[0114] When the program 3021 is executed by the processor 301, it can achieve Figure 1 Any step in the corresponding method embodiment:

[0115] Generate multiple interaction records based on interaction between at least one first virtual model and multiple second virtual models, wherein the at least one first virtual model is generated based on a target user, the multiple second virtual models are generated based on multiple other users and correspond one-to-one with the other users, and the multiple interaction records correspond one-to-one with the multiple second virtual models;

[0116] Scoring the interaction heat of the multiple interaction records based on a preset large model, and determining a target exchange record, wherein the target interaction record is an interaction record with the highest interaction heat among the multiple interaction records;

[0117] Push target account information to the target user, where the target account information is account information of other users corresponding to the target interaction record.

[0118] Optionally, before generating a plurality of interaction records based on the interaction between the at least one first virtual model and the plurality of second virtual models, the method further includes:

[0119] Acquire personal information of the target user, wherein the personal information includes at least one of the age, personality, hobbies and living habits of the target user;

[0120] generating at least one prompt instruction according to the personal information of the target user, wherein different prompt instructions correspond to different weights in the personal information of the target user;

[0121] The at least one first virtual model is generated according to the at least one prompt instruction and the preset template, and the at least one prompt instruction corresponds to the at least one first virtual model in a one-to-one manner.

[0122] Optionally, the generating a plurality of interaction records based on the interaction between the at least one first virtual model and the plurality of second virtual models includes:

[0123] When it is detected that the target user has posted a personal status, extracting information from the personal status to determine at least one keyword;

[0124] updating the at least one first virtual model according to the at least one keyword to obtain at least one third virtual model;

[0125] The plurality of interaction records are generated based on the interaction between the at least one third virtual model and the plurality of second virtual models.

[0126] Optionally, the generating a plurality of interaction records based on the interaction between the at least one first virtual model and the plurality of second virtual models includes:

[0127] Obtaining target interaction requirements of the target user;

[0128] According to the target interaction requirement, matching is performed among multiple candidate users to determine the multiple other users, whose personal information meets the target interaction requirement;

[0129] The plurality of interaction records are generated based on the interaction between the at least one third virtual model and the plurality of second virtual models corresponding to the plurality of other users.

[0130] Optionally, before scoring the interaction heat of the plurality of interaction records based on a preset large model and determining the target exchange record, the method further includes:

[0131] Obtaining an initial large model and a training data set, wherein the initial large model is used to score text content, and the training data set includes a plurality of training texts and a plurality of training labels corresponding to the plurality of training texts;

[0132] Using the training data set to train the initial large model, outputting multiple prediction values;

[0133] Calculate a loss function based on the multiple predicted values ​​and the multiple training labels to obtain a loss value;

[0134] The initial large model is updated according to the loss value to obtain the preset large model.

[0135] Optionally, the pushing the target account information to the target user includes:

[0136] Determine a target other user according to the target exchange record, where the target other user is the other user corresponding to the target exchange record;

[0137] Query the account information of the target other user to obtain the target account information;

[0138] The target account information is pushed to the target user.

[0139] This application uses a virtual model to interact with other users' virtual models, thereby screening out the most popular interaction records, thereby determining the most suitable target account information for the target user, and pushing it to the target user, thereby improving the effect of the interaction.

[0140] The embodiment of the present application also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, each process of the above-mentioned large-model-based interactive method embodiment is implemented, and the same technical effect can be achieved. To avoid repetition, it is not repeated here. The computer-readable storage medium is, for example, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0141] An embodiment of the present application further provides a computer program product, which is stored in a storage medium. The computer program product is executed by at least one processor to implement the various processes of the above-mentioned large model-based interaction method embodiment, and can achieve the same technical effect. To avoid repetition, it will not be repeated here.

[0142] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0143] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, a magnetic disk, or an optical disk), and includes a number of instructions for a terminal (which can be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to execute the methods described in each embodiment of the present application.

[0144] The embodiments of the present application are described above in conjunction with the accompanying drawings, but the present application is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present application, ordinary technicians in this field can also make many forms without departing from the purpose of the present application and the scope of protection of the claims, all of which are within the protection of the present application.

Claims

1. An interactive method based on a large model, characterized in that: The method comprises: Generate multiple interaction records based on interaction between at least one first virtual model and multiple second virtual models, wherein the at least one first virtual model is generated based on a target user, the multiple second virtual models are generated based on multiple other users and correspond one-to-one with the other users, and the multiple interaction records correspond one-to-one with the multiple second virtual models; Scoring the interaction heat of the multiple interaction records based on a preset large model, and determining a target exchange record, wherein the target interaction record is an interaction record with the highest interaction heat among the multiple interaction records; Push target account information to the target user, where the target account information is account information of other users corresponding to the target interaction record.

2. The method according to claim 1, characterized in that: Before generating a plurality of interaction records based on the interaction between the at least one first virtual model and the plurality of second virtual models, the method further includes: Acquire personal information of the target user, wherein the personal information includes at least one of the age, personality, hobbies and living habits of the target user; generating at least one prompt instruction according to the personal information of the target user, wherein different prompt instructions correspond to different weights in the personal information of the target user; The at least one first virtual model is generated according to the at least one prompt instruction and the preset template, and the at least one prompt instruction corresponds to the at least one first virtual model in a one-to-one manner.

3. The method according to claim 1, characterized in that The generating a plurality of interaction records based on the interaction between at least one first virtual model and a plurality of second virtual models comprises: When it is detected that the target user has posted a personal status, extracting information from the personal status to determine at least one keyword; updating the at least one first virtual model according to the at least one keyword to obtain at least one third virtual model; The plurality of interaction records are generated based on the interaction between the at least one third virtual model and the plurality of second virtual models.

4. The method according to claim 3, characterized in that: The generating a plurality of interaction records based on the interaction between at least one first virtual model and a plurality of second virtual models comprises: Obtaining target interaction requirements of the target user; According to the target interaction requirement, matching is performed among multiple candidate users to determine the multiple other users, whose personal information meets the target interaction requirement; The plurality of interaction records are generated based on the interaction between the at least one third virtual model and the plurality of second virtual models corresponding to the plurality of other users.

5. The method according to claim 1, characterized in that Before scoring the interaction heat of the plurality of interaction records based on the preset large model and determining the target exchange record, the method further includes: Obtaining an initial large model and a training data set, wherein the initial large model is used to score text content, and the training data set includes a plurality of training texts and a plurality of training labels corresponding to the plurality of training texts; Using the training data set to train the initial large model, outputting multiple prediction values; Calculate a loss function based on the multiple predicted values ​​and the multiple training labels to obtain a loss value; The initial large model is updated according to the loss value to obtain the preset large model.

6. The method according to claim 1, characterized in that The pushing the target account information to the target user includes: Determine a target other user according to the target exchange record, where the target other user is the other user corresponding to the target exchange record; Query the account information of the target other user to obtain the target account information; The target account information is pushed to the target user.

7. An interactive device based on a large model, characterized in that: The device comprises: An interaction module, configured to interact with a plurality of second virtual models based on at least one first virtual model to generate a plurality of interaction records, wherein the at least one first virtual model is generated based on a target user, the plurality of second virtual models are generated based on a plurality of other users and correspond one-to-one with the other users, and the plurality of interaction records correspond one-to-one with the plurality of second virtual models; A scoring module, configured to score the interaction heat of the plurality of interaction records based on a preset large model, and determine a target exchange record, wherein the target interaction record is the interaction record with the highest interaction heat among the plurality of interaction records; The push module is used to push target account information to the target user, where the target account information is the account information of other users corresponding to the target interaction record.

8. An electronic device comprising: A memory, a processor, and a program stored in the memory and executable on the processor; wherein the processor is used to read the program in the memory to implement the steps in the large model-based interaction method as described in any one of claims 1 to 6.

9. A readable storage medium for storing a program, characterized in that: When the program is executed by a processor, the steps of the large model-based interaction method according to any one of claims 1 to 6 are implemented.

10. A computer program product, characterized in that The computer program product is stored in a storage medium, and the computer program product is executed by at least one processor to implement the steps in the large model-based interaction method according to any one of claims 1 to 6.

Images