Interaction method, model training method, device, equipment and storage medium

Abstract

The embodiment of the present invention provides an interaction method, a model training method, an apparatus, a device and a storage medium, the method comprising: first obtaining an interaction record for a target object and identification information of an interaction subject, wherein the interaction record is used to describe the interaction action generated by the interaction subject on the target object. Then, based on the identification information of the interaction subject and the interaction record, the expected interaction action is determined, and finally a response record describing the expected interaction action is generated to achieve interaction. It can be seen that the above is an automatic interaction method that can improve transaction efficiency. At the same time, when generating a response record, the current interaction action of the interaction subject and the identification information used to reflect the interaction habit of the interaction subject are taken into account at the same time, so that the generated response record is more in line with the interaction habit of the interaction subject, thereby improving the transaction success rate.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to an interaction method, a model training method, a device, a equipment and a storage medium. Background Art

[0002] Online transactions can be considered as an interactive behavior between people. In this interactive process, the two parties, the buyer and the seller, can communicate about the details and prices of the goods. Generally speaking, the buyer and the seller need to keep their online status at the same time and communicate for multiple rounds before they can reach a deal.

[0003] However, during the transaction, the seller may need to deal with multiple buyers, which puts the seller under great pressure to communicate, which can easily lead to low transaction efficiency and success rate. Therefore, how to ensure the success rate of transactions has become an urgent problem to be solved. Summary of the invention

[0004] In view of this, an embodiment of the present invention provides an interaction method, a model training method, an apparatus, a device and a storage medium to ensure the success rate of a transaction.

[0005] In a first aspect, an embodiment of the present invention provides an interaction method, including:

[0006] Acquire identification information of an interaction subject and an interaction record for a target object, wherein the interaction record is used to describe the interaction action;

[0007] Determining an expected interaction action according to the identification information and the interaction record;

[0008] A response record corresponding to the interaction record is generated according to the expected interaction action.

[0009] In a second aspect, an embodiment of the present invention provides an interactive device, including:

[0010] An acquisition module, used to acquire identification information of an interaction subject and an interaction record for a target object, wherein the interaction record is used to describe an interaction action;

[0011] an action determination module, configured to determine an expected interaction action according to the identification information and the interaction record;

[0012] A generating module is used to generate a response record corresponding to the interaction record according to the expected interaction action.

[0013] In a third aspect, an embodiment of the present invention provides an electronic device, comprising a processor and a memory, wherein the memory is used to store one or more computer instructions, wherein the one or more computer instructions, when executed by the processor, implement the interaction method in the first aspect. The electronic device may also include a communication interface for communicating with other devices or a communication network.

[0014] In a fourth aspect, an embodiment of the present invention provides a non-temporary machine-readable storage medium having executable code stored thereon. When the executable code is executed by a processor of an electronic device, the processor can at least implement the interaction method described in the first aspect.

[0015] In a fifth aspect, an embodiment of the present invention provides an interaction method, including:

[0016] Display an interactive interface containing interactive buttons;

[0017] In response to a click operation triggered by the interaction subject on the interaction button, acquiring an interaction record for the target object;

[0018] Determining an expected interaction action according to the interaction record and the identification information of the interaction subject;

[0019] A response record for describing the expected interaction action is displayed in the interaction interface.

[0020] In a sixth aspect, an embodiment of the present invention provides an interactive device, including:

[0021] A display module, used to display an interactive interface including interactive buttons;

[0022] An acquisition module, configured to acquire an interaction record for a target object in response to a click operation triggered by an interaction subject on the interaction button;

[0023] An action determination module, used to determine an expected interaction action according to the interaction record and the identification information of the interaction subject;

[0024] The display module is further used to display a response record describing the expected interactive action in the interactive interface.

[0025] In a seventh aspect, an embodiment of the present invention provides an electronic device, comprising a processor and a memory, wherein the memory is used to store one or more computer instructions, wherein the one or more computer instructions, when executed by the processor, implement the interaction method in the fifth aspect. The electronic device may also include a communication interface for communicating with other devices or a communication network.

[0026] In an eighth aspect, an embodiment of the present invention provides a non-temporary machine-readable storage medium having executable code stored thereon. When the executable code is executed by a processor of an electronic device, the processor can at least implement the interaction method described in the fifth aspect.

[0027] In a ninth aspect, an embodiment of the present invention provides an interaction method, including:

[0028] Display an interactive interface containing interactive buttons;

[0029] In response to a click operation triggered by the first interaction party on the interaction button, obtaining identification information of each of the first interaction party and the second interaction party and a first interaction record for a target object;

[0030] determining an expected interaction action of the first interaction party according to the first interaction record and the identification information of the second interaction party;

[0031] displaying a first response record for describing the expected interaction action of the first interaction party;

[0032] In response to a click operation triggered by the second interaction party on an interaction button, determining an expected interaction action of the second interaction party according to the first interaction record, the first response record, and identification information of the first interaction party;

[0033] A second response record describing the expected interaction action of the second interaction party is displayed.

[0034] In a tenth aspect, an embodiment of the present invention provides an interactive device, including:

[0035] A display module, used to display an interactive interface including interactive buttons;

[0036] an acquisition module, configured to acquire identification information of each of the first interaction party and the second interaction party and a first interaction record for a target object in response to a click operation triggered by the first interaction party on the interaction button;

[0037] A first action determination module, configured to determine an expected interaction action of the first interaction party according to the first interaction record and the identification information of the second interaction party;

[0038] The display module is further used to display a first response record describing the expected interaction action of the first interaction party;

[0039] a second action determination module, configured to determine, in response to a click operation triggered by the second interaction party on the interaction button, an expected interaction action of the second interaction party according to the first interaction record, the first response record and the identification information of the first interaction party;

[0040] The display module is further used to display a second response record for describing the expected interaction action of the second interaction party.

[0041] In an eleventh aspect, an embodiment of the present invention provides an electronic device, comprising a processor and a memory, wherein the memory is used to store one or more computer instructions, wherein the one or more computer instructions, when executed by the processor, implement the interaction method in the ninth aspect. The electronic device may also include a communication interface for communicating with other devices or a communication network.

[0042] In the twelfth aspect, an embodiment of the present invention provides a non-temporary machine-readable storage medium, on which executable code is stored. When the executable code is executed by a processor of an electronic device, the processor can at least implement the interaction method described in the ninth aspect.

[0043] In a thirteenth aspect, an embodiment of the present invention provides a model training method, including:

[0044] receiving an interaction record corresponding to a historical object by a first interaction party, wherein the interaction record includes a plurality of records generated by the first interaction party and a second interaction party, and the interaction record is input by the second interaction party;

[0045] determining the interaction characteristics of the first interaction party according to the interaction record;

[0046] determining action prediction parameters according to the interaction characteristics;

[0047] The action prediction parameters are outputted so that the second interactive party obtains prediction model parameters for the first interactive party according to the action prediction parameters.

[0048] In a fourteenth aspect, an embodiment of the present invention provides a model training device, including:

[0049] A receiving module, configured to receive an interaction record of a first interaction party corresponding to a historical object, wherein the interaction record includes a plurality of records generated by the first interaction party and a second interaction party, and the interaction record is input by the second interaction party;

[0050] a characteristic determination module, configured to determine the interaction characteristic of the first interaction party according to the interaction record;

[0051] A parameter determination module, used to determine action prediction parameters according to the interaction characteristics;

[0052] An output module is used to output the action prediction parameters so that the second interactive party obtains the prediction model parameters for the first interactive party according to the action prediction parameters.

[0053] In a fifteenth aspect, an embodiment of the present invention provides an electronic device, comprising a processor and a memory, wherein the memory is used to store one or more computer instructions, wherein the one or more computer instructions are executed by the processor to implement the model training method in the thirteenth aspect. The electronic device may also include a communication interface for communicating with other devices or communication networks.

[0054] In the sixteenth aspect, an embodiment of the present invention provides a non-temporary machine-readable storage medium having executable code stored thereon. When the executable code is executed by a processor of an electronic device, the processor can at least implement the model training method as described in the thirteenth aspect.

[0055] In a seventeenth aspect, an embodiment of the present invention provides a model training method, including:

[0056] Acquire an interaction record corresponding to the historical object, wherein the interaction record includes a plurality of records generated by the first interaction party and the second interaction party;

[0057] determining the interaction characteristics of the first interaction party according to the interaction record;

[0058] An action prediction model of the second interaction policy for the first interaction party is determined according to the interaction characteristics.

[0059] In an eighteenth aspect, an embodiment of the present invention provides a model training device, including:

[0060] An acquisition module, configured to acquire an interaction record corresponding to a historical object, wherein the interaction record includes a plurality of records generated by a first interaction party and a second interaction party;

[0061] a characteristic determination module, configured to determine the interaction characteristic of the first interaction party according to the interaction record;

[0062] A model determination module is used to determine an action prediction model of the second interaction policy for the first interaction party according to the interaction characteristics.

[0063] In a nineteenth aspect, an embodiment of the present invention provides an electronic device, comprising a processor and a memory, wherein the memory is used to store one or more computer instructions, wherein the one or more computer instructions are executed by the processor to implement the model training method in the seventeenth aspect. The electronic device may also include a communication interface for communicating with other devices or communication networks.

[0064] In the twentieth aspect, an embodiment of the present invention provides a non-temporary machine-readable storage medium having executable code stored thereon. When the executable code is executed by a processor of an electronic device, the processor can at least implement the model training method described in the seventeenth aspect.

[0065] The interactive method provided by the embodiment of the present invention first obtains the interactive record for the target object and the identification information of the interactive subject, wherein the interactive record is used to describe the interactive action generated by the interactive subject on the target object. Then, according to the identification information of the interactive subject and the interactive record, the expected interactive action is determined, and finally a record describing the expected interactive action is generated, which is also the response record corresponding to the interactive record, so as to realize the interaction.

[0066] According to the above description, on the one hand, the present invention provides an automatic interaction method, that is, automatically generating the next interaction action of the interaction subject, that is, the expected interaction action, according to the action described in the interaction record. On the other hand, when generating the response record, the current interaction action of the interaction subject and the identification information of the interaction subject are taken into account at the same time. Since the identification information can reflect the interaction habits of the interaction subject, the generated response record is more in line with the interaction habits of the interaction party and more targeted.

[0067] When the above interaction is specifically a commodity transaction in the background technology, the above automatic transaction method can improve the transaction efficiency. At the same time, since the generated response record is in line with the transaction habits of the first interactive party, the transaction success rate can also be greatly improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0069] Figure 1 A flowchart of an interaction method provided by an embodiment of the present invention;

[0070] Figure 2 A schematic diagram of an interaction record provided by an embodiment of the present invention;

[0071] Figure 3a A flowchart of an optional action prediction method provided by an embodiment of the present invention;

[0072] Figure 3b A flowchart of another optional action prediction method provided by an embodiment of the present invention;

[0073] Figure 4 A flowchart of another interactive method provided by an embodiment of the present invention;

[0074] Figure 5 A flowchart of another interactive method provided by an embodiment of the present invention;

[0075] Figure 6 A schematic diagram of an interactive method provided by an embodiment of the present invention applied to a transaction scenario;

[0076] Figure 7 A flowchart of a model training method provided by an embodiment of the present invention;

[0077] Figure 8 A flowchart of another model training method provided by an embodiment of the present invention;

[0078] Fig. 9 A flowchart of a method for determining an action prediction model provided by an embodiment of the present invention;

[0079] Fig.10 A flowchart of another method for determining an action prediction model provided by an embodiment of the present invention;

[0080] Fig.11 A flowchart of another method for determining an action prediction model provided by an embodiment of the present invention;

[0081] Fig.12 A schematic diagram of the structure of an interactive device provided by an embodiment of the present invention;

[0082] Fig.13 For Fig.12 A schematic diagram of the structure of an electronic device corresponding to the interactive device provided in the illustrated embodiment;

[0083] Fig.14 A schematic diagram of the structure of another interactive device provided by an embodiment of the present invention;

[0084] Fig.15 For Fig.14 A schematic diagram of the structure of an electronic device corresponding to the interactive device provided in the illustrated embodiment;

[0085] Fig.16 A schematic diagram of the structure of another interactive device provided by an embodiment of the present invention;

[0086] Fig.17 For Fig.16 A schematic diagram of the structure of an electronic device corresponding to the interactive device provided in the illustrated embodiment;

[0087] Fig.18 A schematic diagram of the structure of a model training device provided by an embodiment of the present invention;

[0088] Fig.19 For Fig.18 A schematic diagram of the structure of an electronic device corresponding to the model training device provided in the illustrated embodiment;

[0089] Fig. 20 A schematic diagram of the structure of another model training device provided by an embodiment of the present invention;

[0090] Fig.21 For Fig. 20 A schematic diagram of the structure of an electronic device corresponding to the model training device provided in the illustrated embodiment. DETAILED DESCRIPTION

[0091] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0092] The terms used in the embodiments of the present invention are only for the purpose of describing specific embodiments, and are not intended to limit the present invention. The singular forms "a", "said" and "the" used in the embodiments of the present invention and the appended claims are also intended to include plural forms, unless the context clearly indicates other meanings, and "multiple" generally includes at least two, but does not exclude the inclusion of at least one.

[0093] It should be understood that the term "and / or" used in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship.

[0094] As used herein, the words "if" and "if" may be interpreted as "at the time of" or "when" or "in response to determining" or "in response to identifying", depending on the context. Similarly, the phrases "if it is determined" or "if (stated condition or event) is identified" may be interpreted as "when determining" or "in response to determining" or "when identifying (stated condition or event)" or "in response to identifying (stated condition or event)", depending on the context.

[0095] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a product or system including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such a product or system. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the product or system including the elements.

[0096] Before describing in detail the interactive method provided by the embodiment of the present invention, the practical significance of the interactive method may be described exemplarily:

[0097] There may be a variety of interaction scenarios in daily life. As mentioned in the background technology, a common scenario may be that the two parties to the interaction use terminal devices to conduct online transactions. And this transaction can be a second-hand transaction of goods or a non-second-hand transaction. At this time, the two parties to the interaction can directly communicate the detailed information, price and other contents of the transaction object through the application installed on the terminal device. However, this original interaction method requires both parties to be online at the same time to complete. If one party is offline or one party cannot reply in time, the efficiency and success rate of the transaction will be greatly reduced. At this time, the interaction method provided by the present invention can be used to avoid the above problems.

[0098] In another interactive scenario, user A can interact with an intelligent robot. However, the intelligent robot often outputs response statements without emotion. In order to improve the interactive experience of user A, the interactive method provided by the present invention can be used. In this case, the tone and expression of the response statement output by the intelligent robot can be the same as that of another user B who is familiar to the user, so that user A feels like he is talking to user B.

[0099] Of course, the present invention is not limited to the usage scenarios. In addition to the above scenarios, any scenario that requires automatic interaction can use the interaction method provided by the present invention.

[0100] Some embodiments of the present invention are described in detail below in conjunction with the accompanying drawings. In the case where there is no conflict between the embodiments, the following embodiments and the features in the embodiments can be combined with each other. In addition, the step sequence in the following method embodiments is only an example and not a strict limitation.

[0101] The following embodiments are all described by taking the transaction scenario mentioned above as an example. In this scenario, the interaction action is a transaction action generated by the interaction subject on the target object, and the target object can be a commodity. The transaction action includes any one of a bargaining action on the target object, an inquiry action on the attribute information of the target object, a transaction success action on the target object, and a transaction failure action on the target object.

[0102] Based on the above, Figure 1 The flowchart of an interactive method provided by an embodiment of the present invention is a flowchart of an interactive method provided by an embodiment of the present invention. The interactive method provided by an embodiment of the present invention can be executed by an interactive device. It can be understood that the interactive device can be implemented as software, or a combination of software and hardware. The interactive device in this embodiment and the following embodiments can be a terminal device used by the interactive subject, such as a mobile phone, etc. Figure 1 As shown, the method comprises the following steps:

[0103] S101, obtaining identification information of an interaction subject and an interaction record for a target object, wherein the interaction record is used to describe an interaction action.

[0104] The interactive device used by the interactive subject may be installed with a transaction application (Application, referred to as APP). The interactive subject may initiate a transaction for any commodity to be traded on the transaction APP, i.e., the target object. After the transaction is initiated, the interactive subject may start a conversation with the target object, and the content of the conversation will be synchronously displayed on the interactive device used by the interactive subject.

[0105] The conversation content may include: the interaction record generated by the interaction subject for the target object. Optionally, the interaction record may be in the form of text, video, or picture. In terms of content, the interaction record may be asking for detailed information about the target object, such as model, size, etc., or negotiating price, etc.

[0106] During the conversation between the interactive subject and the target object, suppose that the interactive subject initiates a transaction with the target object at time T1. The conversation content generated by the interactive subject with respect to the target object during the period from time T1 to the current time T2 is the interaction record. The interactive device used by the interactive subject can obtain and display this interaction record. For example, at time T2, the interaction record generated by the interactive subject with respect to the target object can be Figure 2 shown.

[0107] At the same time, after the interactive subject initiates a transaction, the interactive device used by the interactive subject can also obtain the user identification of the interactive subject, and further query the identification information of the interactive subject based on the user identification. Among them, the user identification can be specifically expressed as a user ID, and the identification information is used to reflect the transaction habits of the first interactive party.

[0108] S102: Determine an expected interaction action according to the identification information and the interaction record.

[0109] For the interaction record that the interaction device has obtained, which has described the interaction actions generated by the interaction subject during the time period from time T1 to time T2, continue to take over Figure 2The interaction records shown can be understood as follows for the interaction actions:

[0110] During this time period, the interaction actions generated by the interaction subject may be: bargaining - disagreeing to bargaining - bargaining - disagreeing to bargaining - bargaining. At this time, the interaction device will determine the expected interaction action based on the above interaction actions and the identification information of the interaction subject. This expected interaction action can be considered as the interaction action that conforms to the transaction habits of the interaction subject, is most easily accepted by the interaction subject, and is also the most likely to make the transaction successful under the current transaction state.

[0111] S103: Generate a response record corresponding to the interaction record according to the expected interaction action.

[0112] After obtaining the expected interaction action, the interactive device will further generate a record describing the expected interaction action, that is, the response record corresponding to the interaction record. In actual applications, the interaction record and the response record are usually in the form of text sentences. For the generation of the response record, an optional way is that the interactive device already contains sentence templates corresponding to different interaction actions, and the generation of the response record can be completed by querying. The generated response record will be displayed on the interactive device used by the interactive subject at the same time, which means that the automatic response to the interaction record is completed.

[0113] In this embodiment, the interaction record for the target object and the identification information of the interaction subject are first obtained, wherein the interaction record is used to describe the interaction action generated by the interaction subject on the target object. Then, based on the identification information of the interaction subject and the interaction record, the expected interaction action is determined, and finally a response record describing the expected interaction action is generated to achieve interaction. It can be seen that the above is an automatic interaction method that can improve transaction efficiency. At the same time, when generating the response record, the current interaction action of the interaction subject and the identification information used to reflect the interaction habit of the interaction subject are taken into account at the same time, so that the generated response record is more in line with the interaction habit of the interaction subject, thereby improving the transaction success rate.

[0114] In practical applications, the interactive subjects mentioned in the above embodiments may specifically include the first interactive party and the second interactive party. At this time, the determination of the expected interactive action is actually to determine the expected action of the second interactive party according to the identification information of the first interactive party and the interactive record describing the current interactive action of the first interactive party. The expected interactive action obtained is the interactive action that conforms to the trading habits of the first interactive party, is most easily accepted by the first interactive party, and is also the most likely to make the transaction successful under the current transaction state. In the following embodiments, the first interactive party can be considered as the buyer and the second interactive party as the seller. And the interactive devices mentioned in the following embodiments are all interactive devices used by the second interactive party unless otherwise specified.

[0115] Based on the above description, optionally, the determination of the expected interactive action can be achieved with the help of a prediction model. Specifically, obtaining the identification information of the interactive subject is to obtain the identification information of the first interactive party. At this time, the interactive device can first determine the action prediction model of the second interactive policy for the first interactive party based on this identification information. Among them, the identification information can be considered as a model parameter of the action prediction model. At this time, the action prediction model corresponding to the first interactive party can be directly generated according to the identification information of the first interactive party. Then, the interaction records generated by the two interactive parties for the target object in the time period from time T1 to time T2 are input into the action prediction model to output the expected interaction action of the second interactive party, that is, the expected transaction action, through the action prediction model.

[0116] In practical applications, the action prediction models can be varied, and the process of using the action prediction models to predict the expected interaction action of the second interaction party is also different. An optional action prediction method, such as Figure 3a As shown:

[0117] S201, input the interaction record, the attribute information of the target object, and the transaction characteristics of the first interaction party and the second interaction party into the action prediction model, so as to extract features of the interaction record, the attribute information, and the transaction characteristics through the feature extraction layer of the action prediction model.

[0118] S202: Input the extracted features into the output layer of the action prediction model to output the expected transaction action of the second interacting party through the output layer.

[0119] The interaction record, the attribute information of the target object, and the transaction characteristics of the first and second interaction parties are input into the action prediction model, so as to output the expected transaction action of the second interaction party on the target object through the action prediction model. The action prediction can be a convolutional neural network (RNN) model. The training process of the action prediction model can be referred to as follows Fig.11 Detailed description of the illustrated embodiments.

[0120] Specifically, the feature extraction layer in the action prediction model can extract features from the interaction record, the attribute information of the target object, and the transaction features of the two interacting parties. The interaction record describes the current interaction state of the target object and the current transaction action of the first interacting party. Then, the extracted features are input into the output layer of the action prediction model to output the expected transaction action through the output layer.

[0121] The current interaction state of the target object may be an action sequence consisting of transaction actions described in each record in the interaction record. Figure 2In the example shown, at the current time T2, the current interaction status described in the interaction record may be: bargaining (first interaction party) - disagreeing with bargaining (second interaction party) - bargaining (first interaction party) - disagreeing with bargaining (second interaction party) - bargaining (first interaction party). The current transaction action of the first interaction party is bargaining, and the current transaction action of the second interaction party is disagreeing with bargaining.

[0122] The attribute information of the target object may include: the attribute information of the target object includes detailed information of the target transaction object and / or the transaction competitiveness of the target transaction object. The detailed information may be, for example, the size, model, type, etc. of the commodity. It is easy to understand that on the transaction platform, for the same target transaction object, there may be multiple sellers, and the second interactive party is one of the multiple sellers, then the transaction competitiveness may be expressed as: the middle price of the target transaction object by the multiple sellers on the transaction platform at time T2 / the price of the target transaction object by the second interactive party. The higher the ratio, the higher the transaction competitiveness.

[0123] It should be noted that after the first interactive party initiates a transaction with the target transaction object, the interactive device may start to collect the above data, thereby obtaining attribute information of the target transaction object.

[0124] As for the middle price, one optional method is that the interactive device can determine how many transaction objects on the trading platform are the same as the target transaction object by identifying the pictures of each transaction object on the trading APP. If the transaction object is purchased on the trading APP, another optional method is that the second interactive party can directly use the secondary transaction function provided on the APP to put the transaction object back online, and the interactive device can directly obtain the detailed information of this transaction object, so as to determine whether it is the same as the target transaction object. After determining the transaction object on the trading platform that is the same as the target transaction object, the middle price of the target transaction object can be further obtained.

[0125] The interaction characteristics of the first interactive party may include: at least one of the bargaining power, transaction frequency, transaction efficiency, and transaction success rate of the first interactive party within a preset time period, such as one year. The interaction characteristics of the second interactive party are also the same, which will not be repeated here.

[0126] Transaction efficiency can be considered as transaction patience, that is, the average number of interaction rounds of the first interaction party within a preset time period. A question and answer between the first interaction party and the second interaction party can be considered as one round of interaction.

[0127] Bargaining power refers to the average transaction price / average bid price of the transaction objects successfully traded by the first interactive party within a preset time period. When the first interactive party is the buyer, the larger the ratio, the higher the bargaining power; when the first interactive party is the seller, the smaller the ratio, the higher the bargaining power.

[0128] The transaction frequency can be specifically considered as the transaction activity, that is, the number of transactions initiated by the first interacting party on the transaction object on the transaction platform within a preset time period, that is, the total transaction volume, which includes both successful transactions and failed transactions.

[0129] The transaction success rate can be specifically considered as transaction proficiency, that is, the number of successful transactions / total transactions of the first interacting party within a preset time period.

[0130] It should be noted that the above transaction characteristics can usually be updated regularly, and after the first interactive party initiates a transaction with the target transaction object, the interactive device can directly obtain it.

[0131] In this embodiment, in the process of predicting the expected transaction action of the second interactive party, the attribute information of the target object and the transaction characteristics that can reflect the transaction habits of the first interactive party are used, so that the predicted expected transaction action can be more in line with the transaction habits of the first interactive party, thereby improving the transaction success rate.

[0132] Another optional action prediction method is to input the interaction record and the attribute information of the target object into the action prediction model, so as to predict the expected transaction action of the second interaction party through the action prediction model. Optionally, the action prediction model can be a Markov Decision Process (MDP) model. The model parameters of this action prediction model can be determined according to the transaction characteristics of the first interaction party and the attribute information of the historical transaction objects of the first interaction party. The parameter determination process can be referred to in Fig.10 Related description in the illustrated embodiment.

[0133] The attribute information of the historical transaction object includes the detailed information of the historical transaction object and / or the transaction competitiveness of the historical transaction object. The historical transaction object is all the transaction objects with which the first interactive party has successfully or failed in a preset time period. The specific content of the attribute information and transaction characteristics can be found in Figure 3a Description of the illustrated embodiment.

[0134] In this embodiment, Figure 3a Similar to the illustrated embodiment, the attribute information of the target object and the transaction characteristics that can reflect the transaction habits of the first interactive party are also used in the action prediction process, so that the predicted expected transaction action is more in line with the transaction habits of the first interactive party, thereby improving the transaction success rate.

[0135] In addition to the above two methods, there is another optional action prediction method, such as Figure 3b As shown:

[0136] S203: Obtain the current transaction action of the first interacting party described in the interaction record.

[0137] S204, determining the expected transaction action of the second interactive party according to the current transaction action and the prediction equation in the action prediction model, wherein the prediction equation is determined according to the transaction characteristics of the first interactive party.

[0138] After the interaction record is obtained, since the interaction record describes the current interaction action of the first interaction party, that is, the current transaction action, the interaction device also obtains the current transaction action of the first interaction party. Then, optionally, the alternative transaction action of the second interaction party can be determined based on the current transaction action and the prediction equation in the action prediction model. Among them, the action prediction model can be a negotiation decision function (NDF) model.

[0139] If the current transaction action and the alternative transaction action meet the preset conditions, indicating that if the current bid of the first interactive party is used, the first interactive party is less satisfied, then the interactive device determines the alternative transaction action as the expected interactive action of the second interactive party. The alternative transaction action is usually any other transaction action other than a successful or failed transaction.

[0140] If the current transaction action and the alternative transaction action do not meet the preset conditions, indicating that if the transaction is completed with the current bid of the first interactive party, the first interactive party is more satisfied, then the interactive device can directly generate a response record indicating that the transaction is successful, and the target object is completed with the current bid of the first interactive party.

[0141] For determining whether the current transaction action and the alternative transaction action meet the preset conditions, optionally,

[0142] First, the utility values ​​of the current transaction action and the alternative transaction action are determined according to the utility equation. Among them, the utility value of the current transaction action reflects the satisfaction of the first interactive party after the target object is traded with the current transaction action. For example, when the current transaction action of the first interactive party is bargaining, the fewer the number of interaction rounds to which this bargaining action belongs, and the greater the difference between the current price and the bid price of the second interactive party, the higher the utility value corresponding to this bargaining action.

[0143] Then, according to the utility value and the utility loss parameter, determine whether the current transaction action and the alternative transaction action meet the preset conditions. The utility loss parameter can be determined according to the transaction characteristics of the first interactive party. Among them, the current transaction action can be expressed as σ1, the utility value of the current transaction action can be expressed as U(σ1), the alternative transaction action is expressed as σ2, the utility value of the alternative transaction action is expressed as U(σ2), and the utility loss parameter can be expressed as δ. If U(σ1)≤δU(σ2), the preset conditions are met between the two actions, otherwise the preset conditions are not met.

[0144] In the transaction process, bargaining is the most common transaction action. On the one hand, for the utility equation used above, the simplest case can be considered as the difference between the historical average transaction price of the first interactive party corresponding to the historical transaction object and the bottom line price of the second interactive party. The bottom line price can be a preset value.

[0145] On the other hand, for the prediction equation used in the above process, a simple way is that the parameters that the prediction equation can include can be the handicap ratio and the cut value. The above parameters can be understood as follows:

[0146] During the transaction process of each historical transaction object, after the first bargaining action, the first interacting party has an initial bid for each historical transaction object, and the average of multiple initial bids is the bargaining value in the prediction equation.

[0147] First, a transaction curve is obtained based on the bargaining process of the two parties on the historical transaction object, and the curve can be optionally smoothed. The parameter value of the transaction curve is the concession ratio. For example, the transaction curve is usually a power function, and the exponent of the power function is the concession ratio.

[0148] In this embodiment, in the process of predicting the expected transaction action of the second interactive party, the transaction characteristics that can reflect the transaction habits of the first interactive party are used, so that the predicted expected transaction action is more consistent with the transaction habits of the first interactive party, thereby improving the transaction success rate.

[0149] In summary, the three prediction models have different advantages and disadvantages, such as Figure 3a The model applicable to the embodiment shown has high prediction accuracy, but requires many training samples, converges slowly, and has high extractability. Taking the above transaction as an example, the higher the extractability, the higher the profit value of the seller. In practical applications, the extractability should be as low as possible. The MDP model has low prediction accuracy, few training samples, fast convergence, and low extractability. Figure 3b The model applicable to the embodiment shown has moderate prediction accuracy and moderate extractability. In practical applications, one can be selected according to actual needs.

[0150] In addition, the above provides a variety of prediction methods for expected transaction actions. After obtaining the expected interactive actions, you can also follow Figure 1 In the embodiment shown, the query statement template is used to generate the response record. In another optional manner, the expected transaction action can be input into the generation model with the help of the generation model, so as to output the response record through the generation model.

[0151] It should be noted that in the above embodiments, the first interactive party can be considered as the buyer, and the second interactive party can be considered as the seller. However, in actual applications, the two can also be exchanged, the first interactive party can be considered as the seller, and the second interactive party can be considered as the buyer, so that the buyer can achieve differentiated and personalized transactions with different sellers, thereby improving the success rate and efficiency of transactions.

[0152] In addition to the above-mentioned transaction scenarios, in daily interaction scenarios, the first interactive party can be a user and the second interactive party can be an intelligent robot.

[0153] The above embodiments describe the internal working process of the interactive device. Of course, the interactive process can also be described from the perspective of the interface of the interactive device, that is, from the perspective of the interactive party. Figure 4 Flow chart of another interactive method provided by an embodiment of the present invention. The execution subject of this embodiment can also be an interactive device used by the interactive subject. Figure 4 As shown, the method may include the following steps:

[0154] S301, displaying an interactive interface including interactive buttons.

[0155] S302: In response to a click operation triggered by an interaction subject on an interaction button, an interaction record for a target object is obtained.

[0156] S303: Determine the expected interaction action according to the interaction record and the identification information of the interaction subject.

[0157] S304, displaying a response record describing the expected interactive action in the interactive interface. Specifically, the interactive subject can initiate a transaction with the target object through the interactive button provided by the interactive device in the interactive interface. Optionally, the interactive button corresponds to the interactive subject. In response to this click operation, the interactive device obtains the interactive record of the interactive subject for the target object. At the same time, the identification information of the interactive subject can also be obtained, and this identification information reflects the transaction habits of the interactive subject. The expected interactive action of the interactive subject is determined based on the interaction record and the identification information. Finally, a response record is generated based on the expected interactive action, and this response record is also displayed on the interactive device of the interactive subject, which means that automatic response to the interaction record is realized.

[0158] For details not described in this embodiment, please refer to the above Figure 1 to Figure 3bThe relevant description in the illustrated embodiment will not be repeated here.

[0159] In this embodiment, an interactive interface including an interactive button is displayed, and in response to a click operation triggered by the interactive subject on the button, an interactive record of the interactive subject on the target object is obtained. Then, based on the identification information of the interactive subject and the interactive record, the expected interactive action is determined. Finally, the sentence to be recorded describing the expected interactive action is displayed on the interactive device, and this response record is also a response to the interactive record. It can be seen that the above is an efficient automatic interaction method. At the same time, when generating the response record, the interactive device takes into account the identification information used to reflect the interactive subject's interactive habits, so that the generated response record is more in line with the interactive subject's interactive habits, thereby improving the transaction success rate.

[0160] It should be noted that the interaction subject may specifically include a first interaction party and a second interaction party, wherein the first interaction party is the buyer of the target object and the second interaction party may be the seller of the target object. Figure 4 The illustrated embodiments can be understood as follows:

[0161] The first interactive party can initiate a transaction with the target object through the interactive device, and then input the interaction record for the target object. Then, the interactive device used by the second interactive party can receive the identification information of the first interactive party and the interaction record input by the first interactive party, and determine the expected interaction action of the second interactive party based on the two. As the interaction proceeds, the number of records generated by the two interactive parties for the target object will gradually increase. The two interactive parties can generate at least one record for the target object, and this at least one record includes the interaction record input by the first interactive party.

[0162] Finally, the interactive device can generate a response record based on the expected interactive action, and this response record will also be sent to the interactive device used by the first interactive party, so that the response record is displayed on the interactive devices of both interactive parties, that is, the second interactive party automatically responds to the interaction record generated by the first interactive party.

[0163] Above Figure 1 to Figure 4 In the embodiment shown, it can be considered that one of the two parties in the interaction is manually answering online, and the other party is automatically answered by the interactive device. However, in actual applications, especially in transaction scenarios, both parties in the interaction can answer automatically, which can maximize the transaction efficiency. Figure 5 A flowchart of another interactive method provided by an embodiment of the present invention is as follows: Figure 5 As shown, the method may include the following steps:

[0164] S401, displaying an interactive interface including interactive buttons.

[0165] S402: In response to a click operation on an interaction button triggered by the first interaction party, obtaining identification information of each of the first interaction party and the second interaction party and a first interaction record for a target object.

[0166] Taking a transaction scenario as an example, a transaction APP may be installed on the interactive device used by the first interactive party, and an interactive button is also provided on the interactive interface of the transaction APP, and this interactive button corresponds to the first interactive party. The first interactive party may trigger a click operation on this interactive button, and the interactive device used by the first interactive party responds to this operation, obtains the interaction identifiers of the first interactive party and the second interactive party, and the first interaction record of the two parties for the target object at the current moment.

[0167] The interaction identifier is used to reflect the interaction habits of the interacting parties. The content contained in the first interaction record is at least one interaction record generated by the interacting parties on the target object after the first interacting party triggers the click operation. The first interaction record describes the current interaction action of the second interacting party, and the interaction action may be, for example, asking for detailed information of the target object, or making a bid, etc. And this first interaction record will be displayed on the interaction devices used by the interacting parties at the same time.

[0168] S403: Determine an expected interaction action of the first interaction party according to the first interaction record and the identification information of the second interaction party.

[0169] Next, the interaction device used by the first interaction party may predict the expected interaction action of the first interaction party according to the interaction identifier of the second interaction party and the first interaction record.

[0170] In an optional action prediction method, the interaction device used in the first interaction method can first determine the first action prediction model of the first interaction policy for the second interaction party based on the identification information of the second interaction party. Then, the first interaction record is input into the first action prediction model to predict the expected interaction action of the first interaction party through the first prediction model. Among them, the first prediction model can adopt the multiple prediction models provided in the above embodiments.

[0171] S404: Display a first response record for describing the expected interaction action of the first interaction party.

[0172] Then, the interactive device used in the first interactive mode can also generate a first response record according to the expected interactive action, and the first response record is used to respond to the first interactive record in step 402. The generation of the first response record can be achieved by means of a generation model, and the first response record will also be displayed on the interactive devices used by the two parties in the interaction.

[0173] S405: In response to a click operation on the interaction button triggered by the second interaction party, an expected interaction action of the second interaction party is determined according to the first interaction record, the first response record and the identification information of the first interaction party.

[0174] Furthermore, the interactive device used by the second interactive party can also trigger a click operation on the interactive button provided on the interactive interface, and the interactive button at this time corresponds to the second interactive party. The interactive device used by the second interactive party can predict the expected interactive action of the second interactive party based on the identification information of the first interactive party, the first interaction record, and the first response record.

[0175] In an optional action prediction mode, the interaction device used in the second interaction mode can determine the second action prediction model of the second interaction policy for the first interaction party based on the identification information of the first interaction party. Then the first interaction record and the first response record are input into the second action prediction model to predict the expected interaction action of the second interaction party through the second prediction model. Among them, the second prediction model can adopt the multiple prediction models provided in the above embodiments. The interaction devices used by the two interaction parties can be provided with the same or different prediction models.

[0176] S406: Display a second response record for describing the expected interaction action of the second interaction party.

[0177] Finally, the interactive device used in the second interactive mode can also generate a second response record according to the expected interactive action, and this record is used to respond to the first response record in step 404. The generation of the second response record can be achieved by means of a generation model. And the generated second response record will also be displayed on the interactive device used by the two parties in the interaction.

[0178] In addition, for details not described in this embodiment, please refer to the above Figure 1 to Figure 3b The relevant description in the illustrated embodiment will not be repeated here.

[0179] In this embodiment, after the first interactive party starts the interaction, both parties can automatically respond to each other's interaction records, thereby realizing automatic transactions between the two parties and ensuring transaction efficiency. In addition, the response records output by the interactive devices used by the two parties in the automatic reply process are consistent with the transaction habits of the other party, thereby ensuring the success rate of the transaction.

[0180] The following uses a transaction scenario as an example to illustrate the interaction method provided above. Figure 6 understand.

[0181] Assume that the first interactive party is a buyer, and the interactive device used by the first interactive party is a first terminal device; the second interactive party is a seller, and the interactive device used by the second interactive party is a second terminal device; the target object is novel M, and the price is 30 yuan.

[0182] The buyer can trigger the transaction start operation on the transaction interface provided by the first terminal device, and enter an interaction record describing the transaction action on the transaction interface: "How new is novel M?" The interaction action described in this interaction record is: asking for attribute information. This interaction record can be displayed on the terminal devices of both the buyer and the seller at the same time.

[0183] The second terminal device used by the seller responds to the start operation triggered by the buyer, thereby obtaining the buyer's identification information, such as the user ID and the attribute information of the novel M, and then determines the seller's action prediction model for the buyer based on this identification information, and then uses this action prediction model to output the seller's expected transaction action: output attribute information. Finally, the generation model automatically generates an interaction record describing this expected transaction action: "Novel M is 80% new." This generated interaction record can also be displayed on the terminal devices of the buyer and seller. At this point, the seller's automatic reply to the buyer is realized, and the above two records can constitute a round of dialogue for the novel M.

[0184] Then, the buyer can continue to input the interaction record describing the bargaining action on the transaction interface: "Is 20 yuan OK?" At this time, the second terminal device can input the interaction records: "How new is novel M?", "Is novel M 80% new?" and "Is 20 yuan OK?" into the action prediction model, so that the action prediction model outputs the seller's expected transaction action as: not accepting bargaining. Then, the generation model can generate the interaction record "Sincerely offer 25 yuan, no lower price", which is an automatic response to the record "Is 20 yuan OK?". In other words, another round of dialogue for novel M is realized.

[0185] After at least one round of dialogue, the seller can automatically trade with the buyer. Automatic trading can improve trading efficiency. Since the action prediction model is trained based on the buyer's historical trading objects and the buyer's trading habits, the output is in line with the buyer's trading habits, thus ensuring the success rate of the transaction.

[0186] The following is a schematic illustration of the above-mentioned interaction method using a daily chat scenario as an example.

[0187] Assume that the first interactive party is user A and the second interactive party is an intelligent robot. The target object can be the content discussed by the two parties, such as a movie.

[0188] User A can issue a voice command to the intelligent robot, i.e., an interaction record: "What good movies are there recently?" The intelligent robot can, for example, determine the identification information of user A through voiceprint recognition, and then determine the action prediction model corresponding to the user based on the identification information of user A. The above interaction record is input into the model, and the model outputs a response record based on the action prediction type: "There are many good movies recently. You like comedy movies the most. XXX is good." At this point, the intelligent robot has realized its reply to user A.

[0189] Since the prediction model is trained based on the speaking habits of user B, the output response record has a tone and thinking similar to that of user B, so that user A has a feeling of communicating with user B, ensuring the user's interactive experience.

[0190] The following embodiments are all described by taking the transaction scenario as an example. In this scenario, the interaction characteristics are the transaction characteristics. The historical objects can be commodities traded by the first interactive party within a preset time period. In the following embodiments, the first interactive party can be considered as the buyer and the second interactive party as the seller.

[0191] Based on the above, Figure 7 The flowchart of a model training method provided by an embodiment of the present invention is a flowchart of a model training method provided by an embodiment of the present invention. The model training method provided by an embodiment of the present invention can be executed by a training device. It is understandable that the training device can be implemented as software, or a combination of software and hardware, such as a server. Figure 7 As shown, the method comprises the following steps:

[0192] S501, receiving an interaction record corresponding to a historical object by a first interaction party, wherein the interaction record includes a plurality of records generated by the first interaction party and a second interaction party, and the interaction record is input by the second interaction party.

[0193] S502: Determine the interaction characteristics of the first interaction party according to the interaction record.

[0194] The second interactive party can determine which historical objects the first interactive party has traded with within a preset time period, including successful and failed transactions. At the same time, the second interactive party can also collect at least one record generated by the first interactive party and the second interactive party for these historical objects, that is, the interaction record. In actual applications, different historical objects can correspond to different second interactive parties, but the second interactive party for which the interaction record is collected can be any second interactive party on the transaction platform, not limited to the second interactive party corresponding to the historical object.

[0195] The second interactive party can input the collected interactive records into the server. When receiving the interactive records, the server can also obtain the generation time of each record in the interactive records, and obtain the interactive features of the first interactive party by analyzing and processing the interactive records.

[0196] In the transaction scenario, the interaction characteristics are transaction characteristics, which may include at least one of the bargaining power, transaction frequency, transaction efficiency, and transaction success rate of the first interacting party. Figure 3a Related description in the illustrated embodiment.

[0197] S503: Determine action prediction parameters according to the interaction characteristics.

[0198] S504: Outputting action prediction parameters so that the second interactive party can obtain prediction model parameters for the first interactive party according to the action prediction parameters.

[0199] The server can further determine the action prediction parameters according to the interaction characteristics of the first interaction party, and finally output the action prediction parameters. For the second interaction party that needs model training, it can directly input the model parameters into the built model structure, so as to obtain the action prediction model of the second interaction policy for the first interaction party. The parameter determination process can be referred to as follows Figures 9 to 11 Related description in the illustrated embodiment.

[0200] This embodiment can be considered to provide an online platform for model training. For the second interactive party that has model training needs, it can input the collected data into the platform, so that the platform automatically outputs model parameters for it. The second interactive party can directly input the model parameters into the model to obtain the model, so as to respond to the transaction records generated by the first interactive party.

[0201] Figure 8 This is a flow chart of another model training method provided by an embodiment of the present invention. The model training method provided by an embodiment of the present invention can be executed by a training device. It can be understood that the training device can be implemented as software, or a combination of software and hardware. Figure 8 As shown, the method comprises the following steps:

[0202] S601: Acquire an interaction record corresponding to a historical object, where the interaction record includes a plurality of records generated by a first interaction party and a second interaction party.

[0203] S602: Determine the interaction characteristics of the first interaction party according to the interaction record.

[0204] S603: Determine an action prediction model of the second interaction policy for the first interaction party according to the interaction characteristics.

[0205] Taking the transaction scenario as an example, the interaction includes the transaction of the historical object between the first interaction party and the second interaction party within a preset time period. The interaction characteristics of the first interaction party include the transaction characteristics of the first interaction party. The transaction characteristics include at least one of the bargaining power, transaction frequency, transaction efficiency, and transaction success rate of the first interaction party within the preset time period.

[0206] The steps of this embodiment are Figure 7 The specific implementation process can be referred to in Figure 7 The relevant description in the illustrated embodiment will not be repeated here.

[0207] In this embodiment, the interaction characteristics of the first interactive party are taken into consideration during the model training process, so that the trained action prediction model of the second interactive party is specifically targeted at the first interactive party. The obtained model can automatically output a response record that conforms to the interaction habits of the first interactive party, thereby ensuring the efficiency and success rate of the transaction.

[0208] In practical applications, the action prediction model can be varied. The action model can be specifically expressed as an NDF model. Fig. 9 As shown, a method for determining an action prediction model, that is, an optional implementation method of step 603:

[0209] S701: Determine a utility loss parameter corresponding to the first interactive party according to a transaction characteristic of the first interactive party.

[0210] After obtaining the transaction characteristics, we can perform maximum likelihood estimation on data such as transaction frequency, transaction success rate, transaction efficiency, bargaining power, etc., so as to determine the utility loss parameters.

[0211] S702, determining a prediction equation and a utility equation according to transaction characteristics.

[0212] For a description of the prediction equation and the utility equation, see Figure 3b Relevant description in the illustrated embodiment. Two equations can be determined by the method provided in the above embodiment, and in the process of determining the equation parameters, the bargaining power in the transaction characteristics is mainly used.

[0213] S703, constructing an action prediction model by using the utility loss parameter, the utility equation and the prediction equation.

[0214] Finally, the utility loss parameter, utility equation and prediction equation can form a formula for action prediction, namely the action prediction model, such as Figure 3b The embodiment shown provides that the expected interaction action of the second interaction party is predicted by determining whether the formula is established.

[0215] In the training process provided in this embodiment, the transaction characteristics of the first interactive party are taken into consideration, thereby ensuring that the trained action prediction model is specifically designed for the first interactive party by the second interactive party. Figure 3b Related description in the illustrated embodiment.

[0216] like Fig.10 As shown, another method for determining the action prediction model, that is, an optional implementation method of step 603:

[0217] S801, obtaining attribute information of a historical transaction object of a first interactive party and a plurality of transaction states described in an interaction record, where the transaction state includes at least one transaction action described in the interaction record.

[0218] Acquire the attribute information of the historical transaction object of the first interactive party and the multiple transaction states described in the interaction record. The attribute information of the historical transaction object includes the detailed information of the historical transaction object and / or the transaction competitiveness of the historical transaction object. The transaction state can be considered as an action sequence composed of the interaction actions described in each record in the interaction record.

[0219] Undertake Figure 2 The transaction record shown may include multiple transaction states:

[0220] State 1: Bargaining (first party)

[0221] State 2: Bargaining (first party) - Disagreeing to bargain (second party)

[0222] State 3: Bargaining (first party) - Disagreeing to bargain (second party) - Bargaining (first party)

[0223] State 4: Bargaining (first party) - Disagreeing to bargain (second party) - Bargaining (first party) - Disagreeing to bargain (second party)

[0224] State 5: Bargaining (first party) - Disagreeing to bargain (second party) - Bargaining (first party) - Disagreeing to bargain (second party) - Bargaining (first party)

[0225] Therefore, for the first interacting party, each historical transaction object that the first interacting party has traded with may have multiple interaction states.

[0226] S802, combining attribute information of historical transaction objects, determining, in a target transaction state, probability values ​​of different transaction actions expected by the first interactive party, where the target transaction state is any one of a plurality of transaction states.

[0227] After obtaining the multiple transaction states of the first interactive party for different historical transaction objects, the expected transaction actions of the first interactive party under the target transaction state are obtained by statistical means, and the probability values ​​of different transaction actions, such as asking for attribute information, bargaining, and agreeing to the transaction, etc. The expected transaction action corresponding to the highest probability value is also the transaction action most often performed by the first interactive party under the target transaction state, reflecting the transaction habits of the first interactive party. The above-mentioned target transaction state is any one of the multiple transaction states in step 801.

[0228] S803: Generate an action prediction model according to the transaction characteristics of the first interactive party, the probability values ​​corresponding to the plurality of transaction states, and the transaction characteristics of the first interactive party.

[0229] The mapping relationship between different transaction states and probability values ​​can be reflected as a Markov chain matrix, which can be considered as the parameters of the action prediction model, that is, the action prediction model is obtained. This model can actually be an MDP model.

[0230] In the training process provided by this embodiment, the transaction characteristics of the first interactive party and the attribute information of the historical transaction objects are taken into consideration at the same time, thereby ensuring that the trained action prediction model is specifically targeted at the first interactive party by the second interactive party.

[0231] It should be noted that the action prediction model generated above is specifically for the first interactive party. However, in actual applications, it is unrealistic for the second interactive party to train a prediction model specifically for each first interactive party. Therefore, after obtaining the prediction model in the above manner, the obtained interaction records of the first interactive party with the historical transaction object can be further analyzed to determine the transaction characteristics of the first interactive party. Then the prediction model obtained above is not only applicable to the first interactive party, but also to all interactive parties with the same or similar transaction characteristics as the first interactive party.

[0232] like Fig.11 As shown, another method for determining the action prediction model, that is, an optional implementation method of step 603:

[0233] S901, obtaining attribute information of a historical transaction object of a first interaction party and an interaction record describing multiple transaction states, where the transaction state includes at least one transaction action described in the interaction record.

[0234] S902, inputting the interaction records describing the target transaction state, the expected transaction action of the second interactive party in the target state, the transaction characteristics of the first interactive party, and the attribute information of the historical transaction objects into the action prediction model to train the action prediction model, wherein the target transaction state is any transaction state among the multiple transaction states.

[0235] Acquire the attribute information of the historical transaction object of the first interactive party and the interaction record generated for the historical transaction object. The interaction record describes multiple transaction states for the historical transaction object. Specifically, the interaction record may describe the transaction action of the first interactive party in the target state and the expected transaction action of the second interactive party in the target state. The attribute information of the historical transaction object includes detailed information of the historical transaction object and / or the transaction competitiveness of the historical transaction object.

[0236] Then, the training device performs training based on the acquired interaction records, the transaction characteristics of the first interaction party, and the attribute information of the historical transaction object as training samples, thereby obtaining an action prediction model. The action prediction model may be an RNN model.

[0237] The action prediction model can output an expected transaction action of the second interactive party for the input interaction record that describes the current interaction action of the first interactive party. This action can be considered as a predicted transaction action. And because the interaction record input to the model already contains the expected action of the second interactive party, this action is a real transaction action. Therefore, the semantic similarity between the predicted transaction action and the real transaction action can be calculated optionally. The higher the semantic similarity between the two, the better the training effect of the model. The similarity value can be indirectly considered as the loss value of the model. Therefore, the model parameters can be adjusted according to this similarity.

[0238] Another optional model parameter adjustment method is that the predicted transaction action output by the action prediction model is the expected transaction action of the second interactive party under the target state. If the predicted transaction action does not meet the preset transaction rules with at least one transaction action included in the target transaction state described in the interaction record, it indicates that the output predicted transaction action is wrong and does not conform to the transaction logic. At this time, a larger loss value can be set for this output to adjust the model parameters of the action prediction model according to the loss value.

[0239] The preset transaction rules include: Rule 1: The bid of the first interactive party increases monotonically, and the bid of the second interactive party decreases monotonically.

[0240] Rule 2: The bid of the second interacting party is higher than the bid of the first interacting party.

[0241] Rule 3: The transaction price is greater than or equal to the last bid of the first interactive party, and less than or equal to the last bid of the second interactive party.

[0242] Based on the above rules, for example, the target transaction status is: original price 100 yuan (second interactive party) - bargaining to 80 yuan (first interactive party). At this time, according to the interaction records that have been obtained, the second interactive party's actual expected transaction action is: do not agree to bargain, and re-bid 90 yuan. If the predicted action input by the prediction model is to disagree to bargain, and re-bid 60 yuan, this is obviously inconsistent with the transaction logic.

[0243] The above process actually means that if the model outputs an expected trading action that does not conform to the trading logic, it will be punished by setting a larger loss value.

[0244] In addition, if the expected transaction action output by the model conforms to the transaction logic and can maximize the benefits of the first interacting party, it will be rewarded and the reward value will be used to offset the loss value.

[0245] Alternatively, the reward value can be calculated by the following equation: s =p T -υ s +ω d *II d ,γ b =υ b -p T +ω d *II d Among them, γ s is the reward value of the second interacting party, γ b is the reward value of the first interacting party, p T is the final transaction ratio, ω d is the proportion of transaction revenue, and if the transaction is successful, d =1, otherwise ω d =0, II d is an indicator function indicating whether there is a transaction, υ s is the psychological price of the second interacting party, υ b The psychological price is the psychological price of the first interacting party. The psychological price can be considered as the lowest price that the interacting party can accept, which can be set in advance.

[0246] In the training process provided by this embodiment, the transaction characteristics of the first interactive party and the attribute information of the historical transaction objects are taken into consideration at the same time, thereby ensuring that the trained action prediction model is specifically targeted at the first interactive party by the second interactive party.

[0247] In addition, the above Figures 9 to 11The embodiment shown actually trains the first action prediction model of the second interaction policy for the first interaction party. In practical applications, the second action prediction model of the first interaction policy for the second interaction party can also be trained in the above manner. At this time, in order to further ensure the training effect of the model, optionally, the above two action prediction models can also be trained unsupervised.

[0248] Specifically, the first action prediction model is deployed in the interactive device used by the second interactive party, and the second action prediction model is deployed in the interactive device used by the first interactive party. For the interaction record output by one of the action prediction models, the other action prediction model can automatically respond to it, that is, generate an interaction record. The two models can further complete their own training based on the interaction records automatically output by themselves and the interaction records automatically output by the other party, that is, to ensure the effect of model training.

[0249] The interactive device of one or more embodiments of the present invention will be described in detail below. Those skilled in the art will appreciate that these image replacement devices can be configured using commercially available hardware components through the steps taught in this solution.

[0250] Fig.12 A schematic diagram of the structure of an interactive device provided by an embodiment of the present invention is shown in FIG. Fig.12 As shown, the device comprises:

[0251] The acquisition module 11 is used to acquire identification information of an interaction subject and an interaction record for a target object, wherein the interaction record is used to describe an interaction action.

[0252] The action determination module 12 is used to determine the expected interaction action according to the identification information and the interaction record.

[0253] The generating module 13 is used to generate a response record corresponding to the interaction record according to the expected interaction action.

[0254] Optionally, the interaction action includes a transaction action of the interaction subject on the target object. The transaction action includes any one of a bargaining action on the target object, an inquiry action on the attribute information of the target object, a transaction success action on the target object, and a transaction failure action on the target object. The interaction subject includes a first interaction party and a second interaction party.

[0255] The action determination module 12 is specifically configured to determine the expected transaction action of the second interactive party according to the identification information of the first interactive party and the interaction record describing the current interaction action of the first interactive party.

[0256] Optionally, the action determination module 12 specifically includes:

[0257] The determining unit 121 is configured to determine an action prediction model of the second interaction policy for the first interaction party according to the identification information of the first interaction party.

[0258] The input module 122 is used to input the interaction record describing the current transaction action of the first interaction party into the action prediction model, so as to output the expected transaction action of the second interaction party through the action prediction model.

[0259] Optionally, the input module 122 is specifically used to: input the interaction record, the attribute information of the target object, and the transaction characteristics of the first interaction party and the second interaction party into the action prediction model, so as to perform feature extraction on the interaction record, the attribute information, and the transaction characteristics through the feature extraction layer of the action prediction model;

[0260] The extracted features are input into the output layer of the action prediction model to output the expected transaction action of the second interacting party through the output layer.

[0261] Among them, the transaction characteristics include at least one of the bargaining power, transaction frequency, transaction efficiency, and transaction success rate of the first interactive party within the preset time period; the attribute information of the historical transaction object includes the detailed information of the historical transaction object and / or the transaction competitiveness of the historical transaction object; the attribute information of the target object includes the detailed information of the target transaction object and / or the transaction competitiveness of the target transaction object.

[0262] Optionally, the input module 122 is specifically used to: input the interaction record and the attribute information of the target object into the action prediction model, so as to output the expected transaction action of the second interactive party through the action prediction model, wherein the model parameters of the action prediction model are determined according to the transaction characteristics of the first interactive party and the attribute information of the historical transaction objects of the first interactive party.

[0263] Optionally, the input module 122 is specifically used to: obtain the current transaction action of the first interaction party described in the interaction record;

[0264] The expected transaction action of the second interactive party is determined according to the current transaction action and a prediction equation in the action prediction model, wherein the prediction equation is determined according to the transaction characteristics of the first interactive party.

[0265] Optionally, the input module 122 is specifically configured to: if a preset condition is satisfied between the current transaction action of the first interacting party and the candidate transaction action of the second interacting party output by the action prediction model, determine the candidate transaction action as the expected transaction action.

[0266] Optionally, the generating module 13 is further configured to generate a response record indicating that the transaction is successful if the current transaction action and the alternative transaction action do not satisfy a preset condition.

[0267] Optionally, the device further comprises:

[0268] The utility determination module 14 is used to determine the utility values ​​of the current transaction action and the alternative transaction action according to the utility equation, wherein the utility value of the current transaction action reflects the satisfaction of the first interacting party after the target object completes the transaction with the current transaction action.

[0269] The judgment module 15 is used to determine whether the current transaction action and the alternative transaction action meet the preset condition according to the utility value and the utility loss parameter, and the utility loss parameter is determined according to the transaction characteristics of the first interactive party.

[0270] Optionally, the generating module 13 is further used to: input the expected transaction action into the generating model, so as to output the response record through the generating model.

[0271] Fig.12 The device shown can perform Figures 1 to 3b For the method of the embodiment shown in the figure, the part not described in detail in this embodiment can be referred to Figures 1 to 3b The implementation process and technical effects of this technical solution refer to Figures 1 to 3b The description in the illustrated embodiment will not be repeated here.

[0272] The above describes the internal functions and structure of the interactive device. In a possible design, the structure of the interactive device can be implemented as an electronic device, such as Fig.13 As shown, the electronic device may include: a processor 21 and a memory 22. The memory 22 is used to store the electronic device to perform the above Figures 1 to 3b The program of the interactive method provided in the illustrated embodiment, the processor 21 is configured to execute the program stored in the memory 22 .

[0273] The program includes one or more computer instructions, wherein the one or more computer instructions can implement the following steps when executed by the processor 21:

[0274] Acquire identification information of an interaction subject and an interaction record for a target object, wherein the interaction record is used to describe the interaction action;

[0275] Determining an expected interaction action according to the identification information and the interaction record;

[0276] A response record corresponding to the interaction record is generated according to the expected interaction action.

[0277] Optionally, the processor 21 is further configured to execute the aforementioned Figures 1 to 3b All or part of the steps in the illustrated embodiments.

[0278] The structure of the electronic device may further include a communication interface 23 for the electronic device to communicate with other devices or a communication network.

[0279] In addition, an embodiment of the present invention provides a computer storage medium for storing computer software instructions used by the above electronic device, which includes instructions for executing the above Figures 1 to 3b The procedures involved in the interactive method in the method embodiment shown.

[0280] Fig.14 A schematic diagram of the structure of another interactive device provided by an embodiment of the present invention, such as Fig.14 As shown, the device comprises:

[0281] The display module 31 is used to display an interactive interface including interactive buttons, wherein the interactive buttons correspond to interactive subjects.

[0282] The acquisition module 32 is used to acquire the interaction record for the target object in response to the click operation triggered by the interaction subject on the interaction button.

[0283] The action determination module 32 is used to determine the expected interaction action according to the interaction record and the identification information of the interaction subject.

[0284] The display module 31 is further used to display a response record describing the expected interaction action in the interaction interface.

[0285] Fig.14 The device shown can perform Figure 4 For the method of the embodiment shown in the figure, the part not described in detail in this embodiment can be referred to Figure 4 The implementation process and technical effects of this technical solution refer to Figure 4 The description in the illustrated embodiment will not be repeated here.

[0286] The above describes the internal functions and structure of the interactive device. In a possible design, the structure of the interactive device can be implemented as an electronic device, such as Fig.15 As shown, the electronic device may include: a processor 41 and a memory 42. The memory 42 is used to store the electronic device to perform the above Figure 4 The program of the interactive method provided in the illustrated embodiment, the processor 41 is configured to execute the program stored in the memory 42 .

[0287] The program includes one or more computer instructions, wherein the one or more computer instructions can implement the following steps when executed by the processor 41:

[0288] Display an interactive interface containing interactive buttons;

[0289] In response to a click operation triggered by the interaction subject on the interaction button, acquiring an interaction record for the target object;

[0290] Determining an expected interaction action according to the interaction record and the identification information of the interaction subject;

[0291] A response record for describing the expected interaction action is displayed in the interaction interface.

[0292] Optionally, the processor 41 is further configured to execute the aforementioned Figure 4 All or part of the steps in the illustrated embodiments.

[0293] The structure of the electronic device may further include a communication interface 43 for the electronic device to communicate with other devices or a communication network.

[0294] In addition, an embodiment of the present invention provides a computer storage medium for storing computer software instructions used by the above electronic device, which includes instructions for executing the above Figure 4 The procedures involved in the interactive method in the method embodiment shown.

[0295] Fig.16 A structural diagram of another interactive device provided by an embodiment of the present invention is shown in FIG. Fig.16 As shown, the device comprises:

[0296] The display module 51 is used to display an interactive interface including interactive buttons.

[0297] The acquisition module 52 is configured to acquire identification information of each of the first and second interaction parties and a first interaction record for a target object in response to a click operation triggered by the first interaction party on the interaction button.

[0298] The first action determination module 53 is configured to determine the expected interaction action of the first interaction party according to the first interaction record and the identification information of the second interaction party.

[0299] The display module 51 is further configured to display a first response record describing the expected interaction action of the first interaction party.

[0300] The second action determination module 54 is used to determine the expected interaction action of the second interaction party according to the first interaction record, the first response record and the identification information of the first interaction party in response to the click operation triggered by the second interaction party on the interaction button.

[0301] The display module 51 is further configured to display a second response record describing the expected interaction action of the second interaction party.

[0302] Optionally, the first action determination module 52 is specifically configured to: determine a first action prediction model of the first interaction policy for the second interaction party according to the identification information of the second interaction party;

[0303] A first interaction record describing a current interaction action of the second interaction party is input into the first action prediction model, so as to output an expected interaction action of the first interaction party through the first action prediction model.

[0304] Optionally, the second action determination module 54 is specifically configured to: determine a second action prediction model of the second interaction policy for the first interaction party according to the identification information of the first interaction party;

[0305] The first interaction record describing the current interaction action of the second interaction party and the first response record describing the expected interaction action of the first interaction party are input into the second action prediction model to output the expected interaction action of the second interaction party through the second action prediction model.

[0306] Fig.16 The device shown can perform Figure 5 For the method of the embodiment shown in the figure, the part not described in detail in this embodiment can be referred to Figure 5 The implementation process and technical effects of this technical solution refer to Figure 5 The description in the illustrated embodiment will not be repeated here.

[0307] The above describes the internal functions and structure of the interactive device. In a possible design, the structure of the interactive device can be implemented as an electronic device, such as Fig.17 As shown, the electronic device may include: a processor 61 and a memory 62. The memory 62 is used to store the electronic device to perform the above Figure 5 The program of the interactive method provided in the illustrated embodiment, the processor 61 is configured to execute the program stored in the memory 62 .

[0308] The program includes one or more computer instructions, wherein the one or more computer instructions can implement the following steps when executed by the processor 61:

[0309] Display an interactive interface containing interactive buttons;

[0310] In response to a click operation triggered by the first interaction party on the interaction button, obtaining identification information of each of the first interaction party and the second interaction party and a first interaction record for a target object;

[0311] determining an expected interaction action of the first interaction party according to the first interaction record and the identification information of the second interaction party;

[0312] displaying a first response record for describing the expected interaction action of the first interaction party;

[0313] In response to a click operation triggered by the second interaction party on an interaction button, determining an expected interaction action of the second interaction party according to the first interaction record, the first response record, and identification information of the first interaction party;

[0314] A second response record describing the expected interaction action of the second interaction party is displayed.

[0315] Optionally, the processor 61 is further configured to execute the aforementioned Figure 5 All or part of the steps in the illustrated embodiments.

[0316] The structure of the electronic device may further include a communication interface 63 for the electronic device to communicate with other devices or a communication network.

[0317] In addition, an embodiment of the present invention provides a computer storage medium for storing computer software instructions used by the above electronic device, which includes instructions for executing the above Figure 5 The procedures involved in the interactive method in the method embodiment shown.

[0318] Fig.18 A schematic diagram of the structure of a model training device provided by an embodiment of the present invention is shown in FIG. Fig.18 As shown, the device comprises:

[0319] The receiving module 71 is configured to receive an interaction record of a first interaction party corresponding to a historical object, wherein the interaction record includes a plurality of records generated by the first interaction party and a second interaction party, and the interaction record is input by the second interaction party.

[0320] The characteristic determination module 72 is used to determine the interaction characteristics of the first interaction party according to the interaction record.

[0321] The parameter determination module 73 is used to determine the action prediction parameters according to the interaction characteristics.

[0322] The output module 74 is used to output the action prediction parameters so that the second interactive party can obtain the prediction model parameters for the first interactive party according to the action prediction parameters.

[0323] Fig.18 The device shown can perform Figure 7 For the method of the embodiment shown in the figure, the part not described in detail in this embodiment can be referred to Figure 7 The implementation process and technical effects of this technical solution refer to Figure 7 The description in the illustrated embodiment will not be repeated here.

[0324] The above describes the internal functions and structure of the model training device. In a possible design, the structure of the interactive device can be implemented as an electronic device, such as Fig.19 As shown, the electronic device may include: a processor 81 and a memory 82. The memory 82 is used to store the electronic device to perform the above Figure 7 The program of the model training method provided in the illustrated embodiment, the processor 81 is configured to execute the program stored in the memory 82.

[0325] The program includes one or more computer instructions, wherein the one or more computer instructions can implement the following steps when executed by the processor 81:

[0326] receiving an interaction record corresponding to a historical object by a first interaction party, wherein the interaction record includes a plurality of records generated by the first interaction party and a second interaction party, and the interaction record is input by the second interaction party;

[0327] determining the interaction characteristics of the first interaction party according to the interaction record;

[0328] determining action prediction parameters according to the interaction characteristics;

[0329] The action prediction parameters are outputted so that the second interactive party obtains prediction model parameters for the first interactive party according to the action prediction parameters.

[0330] Optionally, the processor 81 is further configured to execute the aforementioned Figure 7 All or part of the steps in the illustrated embodiments.

[0331] The structure of the electronic device may further include a communication interface 53 for the electronic device to communicate with other devices or a communication network.

[0332] In addition, an embodiment of the present invention provides a computer storage medium for storing computer software instructions used by the above electronic device, which includes instructions for executing the above Figure 7The procedures involved in the model training method in the method embodiment shown.

[0333] Fig. 20 A schematic diagram of the structure of another model training device provided by an embodiment of the present invention, such as Fig. 20 As shown, the device comprises:

[0334] The acquisition module 91 is used to acquire the interaction record corresponding to the historical object, where the interaction record includes multiple records generated by the first interaction party and the second interaction party.

[0335] The characteristic determination module 92 is used to determine the interaction characteristics of the first interaction party according to the interaction record.

[0336] The model determination module 93 is used to determine an action prediction model of the second interaction policy for the first interaction party according to the interaction characteristics.

[0337] Optionally, the interaction includes a transaction between the first interactive party and the second interactive party on the historical object within a preset time period; the interaction characteristics of the first interactive party include transaction characteristics of the first interactive party; the transaction characteristics include at least one of the bargaining power, transaction frequency, transaction efficiency, and transaction success rate of the first interactive party within the preset time period.

[0338] The model determination module 93 is used to determine the utility loss parameter corresponding to the first interactive party according to the transaction characteristics of the first interactive party;

[0339] Determining a prediction equation and a utility equation according to the transaction characteristics;

[0340] The action prediction model is composed of the utility loss parameter, the utility equation and the prediction equation.

[0341] Optionally, the attribute information of the historical transaction object includes detailed information of the historical transaction object and / or transaction competitiveness of the historical transaction object.

[0342] The model determination module 93 is used to obtain the attribute information of the historical transaction objects of the first interaction party and multiple transaction states described in the interaction record, where the transaction state includes at least one transaction action described in the interaction record;

[0343] combining the attribute information of the historical transaction object, determining the probability value of the expected transaction action of the first interactive party being different transaction actions under the target transaction state, the target transaction state being any transaction state among the multiple transaction states;

[0344] The action prediction model is generated according to the transaction characteristics of the first interacting party and the probability values ​​corresponding to each of the plurality of transaction states.

[0345] Optionally, the model determination module 93 is used to obtain attribute information of the historical transaction object of the first interaction party and the interaction record describing multiple transaction states, where the transaction state includes at least one transaction action described in the interaction record;

[0346] An interaction record describing the target transaction state, the expected transaction action of the second interactive party under the target state, the transaction characteristics of the first interactive party, and the attribute information of the historical transaction object are input into an action prediction model to train the action prediction model, wherein the target transaction state is any one of the multiple transaction states.

[0347] Optionally, the device further comprises: an adjustment module 94, configured to adjust model parameters of the prediction model according to the semantic similarity between the predicted transaction action output by the action prediction model and the expected transaction action of the second interactive party.

[0348] Optionally, the adjustment module 94 is also used to adjust the model parameters of the prediction model if the predicted transaction action output by the action prediction model and at least one transaction action included in the target transaction state do not satisfy a preset transaction rule, and the at least one transaction action is generated by the first transaction party and / or the second transaction party.

[0349] Optionally, the device also includes: an output module 95, which is used to respond to the transaction start operation so that the first interaction policy's first action prediction model for the second interaction party, and the second interaction policy's second action prediction model for the first interaction party, alternately output response records corresponding to the interaction records generated by the other party.

[0350] The adjustment module 94 is further configured to adjust the model parameters of the first action prediction model and the second action prediction model according to the response record.

[0351] Fig. 20 The device shown can perform Figure 8 For the method of the embodiment shown in the figure, the part not described in detail in this embodiment can be referred to Figure 8 The implementation process and technical effects of this technical solution refer to Figure 8 The description in the illustrated embodiment will not be repeated here.

[0352] The above describes the internal functions and structure of the model training device. In a possible design, the structure of the interactive device can be implemented as an electronic device, such as Fig.21 As shown, the electronic device may include: a processor 101 and a memory 102. The memory 102 is used to store information that supports the electronic device to execute the above Figure 8 The program of the model training method provided in the illustrated embodiment, the processor 101 is configured to execute the program stored in the memory 52.

[0353] The program includes one or more computer instructions, wherein the one or more computer instructions can implement the following steps when executed by the processor 101:

[0354] Acquire an interaction record corresponding to the historical object, wherein the interaction record includes a plurality of records generated by the first interaction party and the second interaction party;

[0355] determining the interaction characteristics of the first interaction party according to the interaction record;

[0356] An action prediction model of the second interaction policy for the first interaction party is determined according to the interaction characteristics.

[0357] Optionally, the processor 101 is further configured to execute the aforementioned Figure 8 All or part of the steps in the illustrated embodiments.

[0358] The structure of the electronic device may further include a communication interface 103 for the electronic device to communicate with other devices or a communication network.

[0359] In addition, an embodiment of the present invention provides a computer storage medium for storing computer software instructions used by the above electronic device, which includes instructions for executing the above Figure 8 The procedures involved in the model training method in the method embodiment shown.

[0360] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention 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 deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An interactive method, characterized in that: include: Acquire identification information of an interaction subject and an interaction record for a target object, wherein the interaction record is used to describe an interaction action, the interaction action includes a transaction action of the interaction subject on the target object, and the interaction subject includes a first interaction party and a second interaction party; Determine an action prediction model of the second interaction policy for the first interaction party based on the identification information of the first interaction party; input the interaction record describing the current transaction action of the first interaction party into the action prediction model, so as to output the expected transaction action of the second interaction party through the action prediction model; wherein, the interaction record, the attribute information of the target object, and the transaction characteristics of the first interaction party and the second interaction party are input into the action prediction model to obtain the expected transaction action of the second interaction party; or, input the interaction record and the attribute information of the target object into the action prediction model to obtain the expected transaction action of the second interaction party; and generate a response record corresponding to the interaction record based on the expected transaction action.

2. The method according to claim 1, characterized in that The transaction action includes any one of a bargaining action on the target object, an inquiry action on the attribute information of the target object, a transaction success action on the target object, and a transaction failure action on the target object.

3. The method according to claim 1, characterized in that: The step of inputting the interaction record describing the current interaction action of the first interaction party into the action prediction model, so as to output the expected transaction action of the second interaction party through the action prediction model, comprises: Inputting the interaction record, the attribute information of the target object, and the transaction characteristics of the first interaction party and the second interaction party into the action prediction model, so as to perform feature extraction on the interaction record, the attribute information, and the transaction characteristics through the feature extraction layer of the action prediction model; The extracted features are input into the output layer of the action prediction model to output the expected transaction action of the second interacting party through the output layer.

4. The method according to claim 1, characterized in that: The step of inputting the interaction record describing the current interaction action of the first interaction party into the action prediction model, so as to output the expected transaction action of the second interaction party through the action prediction model, comprises: The interaction record and the attribute information of the target object are input into the action prediction model to output the expected transaction action of the second interactive party through the action prediction model, wherein the model parameters of the action prediction model are determined according to the transaction characteristics of the first interactive party and the attribute information of the historical transaction objects of the first interactive party.

5. The method according to claim 1, characterized in that The step of inputting the interaction record describing the current interaction action of the first interaction party into the action prediction model, so as to output the expected transaction action of the second interaction party through the action prediction model, comprises: Acquire the current transaction action of the first interaction party described in the interaction record; The expected transaction action of the second interactive party is determined according to the current transaction action and a prediction equation in the action prediction model, wherein the prediction equation is determined according to the transaction characteristics of the first interactive party.

6. The method according to claim 5, characterized in that The determining the expected transaction action of the second interacting party according to the current transaction action and the prediction equation in the action prediction model includes: If the current transaction action of the first interacting party and the candidate transaction action of the second interacting party output by the action prediction model meet a preset condition, the candidate transaction action is determined to be the expected transaction action.

7. The method according to claim 6, characterized in that The method further comprises: If the current transaction action and the alternative transaction action do not satisfy the preset condition, a response record indicating that the transaction is successful is generated.

8. The method according to claim 6, characterized in that The method further comprises: Determining the utility values ​​of the current transaction action and the alternative transaction action according to the utility equation, respectively, wherein the utility value of the current transaction action reflects the satisfaction degree of the first interacting party after the target object completes the transaction with the current transaction action; It is determined whether the current transaction action and the candidate transaction action satisfy the preset condition according to the utility value and the utility loss parameter, wherein the utility loss parameter is determined according to the transaction characteristics of the first interactive party.

9. The method according to any one of claims 1 to 8, characterized in that The step of generating a response record corresponding to the interaction record according to the expected transaction action includes: The expected transaction action is input into a generation model to output the response record through the generation model.

10. The method according to any one of claims 3 to 8, characterized in that The transaction characteristics include at least one of the bargaining power, transaction frequency, transaction efficiency, and transaction success rate of the first interactive party within a preset time period; the attribute information of the historical transaction object includes the detailed information of the historical transaction object and / or the transaction competitiveness of the historical transaction object; the attribute information of the target object includes the detailed information of the target transaction object and / or the transaction competitiveness of the target transaction object.

11. An interactive method, characterized in that: include: Display an interactive interface containing interactive buttons; In response to a click operation triggered by an interaction subject on the interaction button, an interaction record for a target object is acquired; the interaction action includes a transaction action of the interaction subject on the target object, and the interaction subject includes a first interaction party and a second interaction party; Determine an action prediction model of the second interaction policy for the first interaction party according to the identification information of the first interaction party; input the interaction record describing the current transaction action of the first interaction party into the action prediction model, so as to output the expected transaction action of the second interaction party through the action prediction model; wherein the interaction record, the attribute information of the target object, and the transaction characteristics of the first interaction party and the second interaction party are input into the action prediction model to obtain the expected transaction action of the second interaction party; or, input the interaction record and the attribute information of the target object into the action prediction model to obtain the expected transaction action of the second interaction party; A response record for describing the expected transaction action is displayed in the interactive interface.

12. The method according to claim 11, characterized in that The interaction button corresponds to the interaction subject.

13. An interactive method, characterized in that: include: Display an interactive interface containing interactive buttons; In response to a click operation triggered by the first interaction party on the interaction button, obtaining identification information of each of the first interaction party and the second interaction party and a first interaction record for a target object; determining an expected interaction action of the first interaction party according to the first interaction record and the identification information of the second interaction party; displaying a first response record for describing the expected interaction action of the first interaction party; In response to a click operation triggered by the second interaction party on an interaction button, determining an expected interaction action of the second interaction party according to the first interaction record, the first response record, and identification information of the first interaction party; A second response record describing the expected interaction action of the second interaction party is displayed.

14. The method according to claim 13, characterized in that The determining, according to the first interaction record and the identification information of the second interaction party, the expected interaction action of the first interaction party includes: determining, according to the identification information of the second interaction party, a first action prediction model of the first interaction policy for the second interaction party; A first interaction record describing a current interaction action of the second interaction party is input into the first action prediction model, so as to output an expected interaction action of the first interaction party through the first action prediction model.

15. The method according to claim 13, characterized in that The determining, according to the first interaction record, the first response record, and the identification information of the first interaction party, the expected interaction action of the second interaction party includes: determining, according to the identification information of the first interaction party, a second action prediction model of the second interaction policy for the first interaction party; The first interaction record describing the current interaction action of the second interaction party and the first response record describing the expected interaction action of the first interaction party are input into the second action prediction model to output the expected interaction action of the second interaction party through the second action prediction model.

16. A model training method, characterized in that: include; receiving an interaction record corresponding to a historical object by a first interaction party, wherein the interaction record includes a plurality of records generated by the first interaction party and a second interaction party, and the interaction record is input by the second interaction party; determining the interaction characteristics of the first interaction party according to the interaction record; determining action prediction parameters according to the interaction characteristics; The action prediction parameters are outputted so that the second interactive party obtains prediction model parameters for the first interactive party according to the action prediction parameters.

17. A model training method, characterized in that: include: Acquire an interaction record corresponding to the historical object, wherein the interaction record includes a plurality of records generated by the first interaction party and the second interaction party; determining the interaction characteristics of the first interaction party according to the interaction record; Determine an action prediction model of the second interaction policy for the first interaction party according to the interaction characteristics; wherein the action prediction model is used to: output the expected transaction action of the second interaction party according to the interaction record describing the current transaction action of the first interaction party, the attribute information of the target object, and the respective transaction characteristics of the first interaction party and the second interaction party, or, the action prediction model is used to: output the expected transaction action of the second interaction party according to the interaction record describing the current transaction action of the first interaction party and the attribute information of the target object.

18. The method according to claim 17, characterized in that The interaction includes transactions between the first interactive party and the second interactive party on the historical object within a preset time period; the interaction characteristics of the first interactive party include transaction characteristics of the first interactive party; the transaction characteristics include at least one of the bargaining power, transaction frequency, transaction efficiency, and transaction success rate of the first interactive party within the preset time period.

19. The method according to claim 18, characterized in that The step of determining the action prediction model of the second interaction policy for the first interaction party according to the interaction characteristic includes: determining a utility loss parameter corresponding to the first interactive party according to the transaction characteristics of the first interactive party; Determining a prediction equation and a utility equation according to the transaction characteristics; The action prediction model is composed of the utility loss parameter, the utility equation and the prediction equation.

20. The method according to claim 18, characterized in that The step of determining the action prediction model of the second interaction policy for the first interaction party according to the interaction characteristic includes: Acquire attribute information of historical transaction objects of the first interactive party and a plurality of transaction states described in the interaction record, where the transaction state includes at least one transaction action described in the interaction record; combining the attribute information of the historical transaction object, determining, in a target transaction state, a probability value that the expected transaction action of the first interactive party is different transaction actions, wherein the target transaction state is any one of the multiple transaction states; The action prediction model is generated according to the transaction characteristics of the first interacting party and the probability values ​​corresponding to each of the plurality of transaction states.

21. The method according to claim 18, characterized in that The step of determining the action prediction model of the second interaction policy for the first interaction party according to the interaction characteristic includes: Acquire attribute information of historical transaction objects of the first interaction party and the interaction record describing a plurality of transaction states, where the transaction state includes at least one transaction action described in the interaction record; A description of a target transaction state, an interaction record of the expected transaction action of the second interactive party under the target transaction state, the transaction characteristics of the first interactive party, and attribute information of the historical transaction object are input into an action prediction model to train the action prediction model, wherein the target transaction state is any one of the multiple transaction states.

22. The method according to claim 20 or 21, characterized in that The attribute information of the historical transaction object includes detailed information of the historical transaction object and / or transaction competitiveness of the historical transaction object.

23. The method according to claim 21, characterized in that The method further comprises: According to the semantic similarity between the predicted transaction action output by the action prediction model and the expected transaction action of the second interactive party, the model parameters of the prediction model are adjusted.

24. The method according to claim 21, characterized in that The method further comprises: If the predicted transaction action output by the action prediction model and at least one transaction action included in the target transaction state do not satisfy the preset transaction rules, the model parameters of the prediction model are adjusted, and the at least one transaction action is generated by the first transaction party and / or the second transaction party.

25. The method according to claim 21, characterized in that The method further comprises: In response to the transaction start operation, the first interaction policy's first action prediction model for the second interaction party and the second interaction policy's second action prediction model for the first interaction party alternately output response records corresponding to the interaction records generated by the other party; According to the response record, model parameters of the first action prediction model and the second action prediction model are adjusted respectively.

26. An interactive device, characterized in that: include: an acquisition module, configured to acquire identification information of an interaction subject and an interaction record for a target object, wherein the interaction record is used to describe an interaction action, the interaction action includes a transaction action of the interaction subject on the target object, and the interaction subject includes a first interaction party and a second interaction party; An action determination module is used to determine an action prediction model of the second interaction policy for the first interaction party based on the identification information of the first interaction party; the interaction record describing the current transaction action of the first interaction party is input into the action prediction model, so as to output the expected transaction action of the second interaction party through the action prediction model; wherein the interaction record, the attribute information of the target object, and the transaction characteristics of the first interaction party and the second interaction party are input into the action prediction model to obtain the expected transaction action of the second interaction party; or, the interaction record and the attribute information of the target object are input into the action prediction model to obtain the expected transaction action of the second interaction party; a generation module is used to generate a response record corresponding to the interaction record according to the expected transaction action.

27. An electronic device, characterized in that: include: A memory and a processor; wherein the memory stores executable code, and when the executable code is executed by the processor, the processor executes the interaction method as described in any one of claims 1 to 10.

28. A non-transitory machine-readable storage medium, characterized in that The non-transitory machine-readable storage medium stores executable code, and when the executable code is executed by a processor of an electronic device, the processor is caused to execute the interaction method according to any one of claims 1 to 10.

29. An interactive device, characterized in that: include: A display module, used to display an interactive interface including interactive buttons; an acquisition module, configured to acquire an interaction record for a target object in response to a click operation triggered by an interaction subject on the interaction button; the interaction action includes a transaction action of the interaction subject on the target object, and the interaction subject includes a first interaction party and a second interaction party; an action determination module, for determining an action prediction model of the second interaction policy for the first interaction party according to the identification information of the first interaction party; inputting the interaction record describing the current transaction action of the first interaction party into the action prediction model, so as to output the expected transaction action of the second interaction party through the action prediction model; wherein the interaction record, the attribute information of the target object, and the transaction characteristics of the first interaction party and the second interaction party are input into the action prediction model to obtain the expected transaction action of the second interaction party; or, the interaction record and the attribute information of the target object are input into the action prediction model to obtain the expected transaction action of the second interaction party; The display module is also used to display a response record describing the expected transaction action in the interactive interface.

30. An electronic device, characterized in that: include: A memory and a processor; wherein the memory stores executable code, and when the executable code is executed by the processor, the processor executes the interaction method as described in claim 11 or 12.

31. A non-transitory machine-readable storage medium, characterized in that The non-transitory machine-readable storage medium stores executable code, and when the executable code is executed by a processor of an electronic device, the processor is caused to execute the interaction method as claimed in claim 11 or 12.

32. An interactive device, characterized in that: include: A display module, used to display an interactive interface including interactive buttons; an acquisition module, configured to acquire identification information of the first and second interactive parties and a first interaction record for a target object in response to a click operation triggered by the first interactive party on the interaction button; A first action determination module, configured to determine an expected interaction action of the first interaction party according to the first interaction record and the identification information of the second interaction party; The display module is further used to display a first response record describing the expected interaction action of the first interaction party; a second action determination module, configured to determine, in response to a click operation triggered by the second interaction party on the interaction button, an expected interaction action of the second interaction party according to the first interaction record, the first response record and the identification information of the first interaction party; The display module is further used to display a second response record for describing the expected interaction action of the second interaction party.

33. An electronic device, characterized in that: include: A memory and a processor; wherein the memory stores executable code, and when the executable code is executed by the processor, the processor executes the interactive method as described in any one of claims 13 to 15.

34. A non-transitory machine-readable storage medium, characterized in that The non-transitory machine-readable storage medium stores executable code, and when the executable code is executed by a processor of an electronic device, the processor is caused to execute the interaction method as described in any one of claims 13 to 15.

35. A model training device, characterized in that: include: A receiving module, configured to receive an interaction record of a first interaction party corresponding to a historical object, wherein the interaction record includes a plurality of records generated by the first interaction party and a second interaction party, and the interaction record is input by the second interaction party; a characteristic determination module, configured to determine the interaction characteristic of the first interaction party according to the interaction record; A parameter determination module, used to determine action prediction parameters according to the interaction characteristics; An output module is used to output the action prediction parameters so that the second interactive party obtains the prediction model parameters for the first interactive party according to the action prediction parameters.

36. An electronic device, characterized in that: include: A memory and a processor; wherein the memory stores executable code, and when the executable code is executed by the processor, the processor executes the model training method as described in claim 16.

37. A non-transitory machine-readable storage medium, characterized in that The non-temporary machine-readable storage medium stores executable code, and when the executable code is executed by a processor of an electronic device, the processor executes the model training method as described in claim 16.

38. A model training device, characterized in that: include: An acquisition module, configured to acquire an interaction record corresponding to a historical object, wherein the interaction record includes a plurality of records generated by a first interaction party and a second interaction party; a characteristic determination module, configured to determine the interaction characteristic of the first interaction party according to the interaction record; A model determination module, configured to determine an action prediction model of the second interaction policy for the first interaction party according to the interaction characteristics; Among them, the action prediction model is used to: output the expected transaction action of the second interactive party based on the interaction record describing the current transaction action of the first interactive party, the attribute information of the target object, and the respective transaction characteristics of the first interactive party and the second interactive party, or, the action prediction model is used to: output the expected transaction action of the second interactive party based on the interaction record describing the current transaction action of the first interactive party and the attribute information of the target object.

39. An electronic device, characterized in that: include: A memory and a processor; wherein the memory stores executable code, and when the executable code is executed by the processor, the processor executes the model training method as described in any one of claims 17 to 25.

40. A non-transitory machine-readable storage medium, characterized in that The non-temporary machine-readable storage medium stores executable code, and when the executable code is executed by a processor of an electronic device, the processor executes the model training method as described in any one of claims 17 to 25.

Images