Method for intelligently generating dialogue content and electronic equipment

Abstract

The embodiment of the invention discloses a method for intelligently generating dialogue content and electronic equipment. The method comprises the following steps: receiving inquiry information that a first user inquires a commodity published by a second user; if the second user is in the offline state, price related information is extracted from the inquiry information by using a first artificial intelligence AI large-scale parameter model; reasoning the inquiry information and the price information extraction result by using a second AI large-scale parameter model, and determining an action which needs to be executed currently from a plurality of predefined action types; and utilizing a third AI large-scale parameter model to generate dialogue content aiming at the inquiry information based on the current action needing to be executed. According to the embodiment of the invention, seller users can be helped to realize message hosting service more intelligently.

Description

Technical Field

[0001] This application relates to the field of content generation technology, and particularly to a method for intelligently generating conversation content and an electronic device. Background Art

[0002] In a commodity information service system (also known as an e-commerce platform), there are often some individual sellers. Especially in an e-commerce platform mainly engaged in second-hand idle item transactions, such individual sellers are more common. Since most of these individual sellers are not professional sellers, they have limited energy invested in their operations and are often offline. Such sellers lead to a situation where buyers are very interested in a commodity and initiate an inquiry, but the seller is offline, resulting in the transaction being unable to proceed. And when the seller sees the message and replies to the buyer, the buyer is often offline, thus missing many trading opportunities. This phenomenon is usually referred to as the problem of temporal and spatial dislocation between buyers and sellers.

[0003] In response to the above problem of temporal and spatial dislocation, some e-commerce platforms provide a preset automatic reply function for sellers. However, this function only supports preset fixed reply statements and cannot respond to diverse demands of buyers such as inquiries and bargaining for commodities according to the actual situation. Summary of the Invention

[0004] This application provides a method for intelligently generating conversation content and an electronic device, which can more intelligently help seller users implement a message trusteeship service.

[0005] This application provides the following solutions:

[0006] A method for intelligently generating conversation content, comprising:

[0007] Receiving inquiry information from a first user for inquiring about a commodity posted by a second user;

[0008] If the second user is in an offline state, extracting price-related information from the inquiry information by using a first artificial intelligence (AI) large-scale parameter model;

[0009] Using a second AI large-scale parameter model to reason about the inquiry information and the result of price information extraction, and determining an action to be performed currently from a predefined plurality of action types;

[0010] Using a third AI large-scale parameter model to generate conversation content for the inquiry information based on the currently required action.

[0011] Wherein, when determining the currently required action, the second AI large-scale parameter model is further used for:

[0012] Predict the actions that the first user may perform in the future based on the context information of the conversation, so that the second large-scale AI parameter model determines the actions that need to be performed currently or reflects on the rationality of the actions that need to be performed currently through two-way thinking.

[0013] Among them, if it is predicted that the action that the first user may perform in the future is to continue to make an offer to test the bottom price of the second user, then determine the action that needs to be performed currently as rejecting the current offer of the first user.

[0014] Among them, if the action that needs to be performed currently determined by the second large-scale AI parameter model is to make a new offer, the method further includes:

[0015] Determine the marked price and bottom price information set by the second user for the commodity;

[0016] Determine the offer information of the first user for the commodity according to the price information extraction result of the first large-scale AI parameter model;

[0017] Calculate new offer information according to the marked price, bottom price, offer information of the first user and the preset rule information of the commodity;

[0018] Input the new offer information into the third large-scale AI parameter model, so that the third large-scale AI parameter model generates the conversation content based on the action that needs to be performed currently and the new offer information.

[0019] Among them, the actions that need to be performed currently determined by the second large-scale AI parameter model include: not replying.

[0020] Among them, it further includes:

[0021] Provide the predefined common language skill information to the third large-scale AI parameter model, so that the third large-scale AI parameter model generates the conversation content based on the action that needs to be performed currently and the common language skill information.

[0022] Among them, when the third large-scale AI parameter model outputs the conversation content, it also outputs the content related to the thinking process.

[0023] A computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, the steps of the method described in any one of the foregoing are implemented.

[0024] An electronic device, including:

[0025] One or more processors; and

[0026] A memory associated with the one or more processors, the memory for storing program instructions that, when read and executed by the one or more processors, perform the steps of the method according to any one of the foregoing.

[0027] A computer program product, including a computer program / computer-executable instructions that, when executed by a processor in an electronic device, implement the steps of the method according to any one of the foregoing.

[0028] According to the specific embodiments provided in this application, the following technical effects are disclosed in this application:

[0029] Through the embodiments of this application, after the first user initiates an inquiry message to the second user, if the second user is not online, the intelligent hosting service system can be used to reply on behalf of the second user. Specifically, the intelligent hosting service system can first extract price-related information from the inquiry message through the first AI large model. After that, it can reason based on the inquiry message and the price extraction result through the second AI large model, and select the action to be executed currently from a plurality of predefined action types. Then, it can generate conversation content based on the selected action through the third AI large model for replying to the inquiry message of the first user. Among them, since the system is divided into two parts: policy selection and content generation, the entire system is no longer passively generating responses based on the user's input, but can determine whether to generate specific conversation content based on the specifically selected policy (action), thereby solving the hallucination problem existing in the end-to-end AI large model. In addition, for the problem that the AI large model is not sensitive to numbers, a dedicated price extraction module is also provided, so that in the conversation process mainly involving bargaining, etc., the system can more sensitively obtain the price information therein, and then make more accurate and reasonable action selections and content generations.

[0030] In a preferred implementation, during the process of action selection by the second AI large model, the possible actions that the first user may take in the future can also be predicted through a two-way thinking method, and then the policy selection can be made according to the prediction result, or the selected policy can be reflected on, and so on. In this way, since the large model can select its own policy by thinking about the opponent's policy, the large model can better achieve the negotiation goal.

[0031] In addition, when the policy selected by the second AI large model is to make a new offer, the price calculation can be specifically performed through a separate offer module, and the offer module can calculate the price in a rule-based manner to realize the linkage between the AI large model and the rules, so as to ensure the smooth progress of the bargaining process. Moreover, during the process of content generation by the third AI large model, multiple optional language skills can also be provided to make the generated statements more rich.

[0032] Of course, it is not necessary for any product implementing this application to achieve all the above-mentioned advantages simultaneously. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0034] Figure 1 is a schematic diagram of the system architecture provided by the embodiment of the present application;

[0035] Figure 2 is a flowchart of the method provided by the embodiment of the present application;

[0036] Figure 3 is a schematic diagram of the device provided by the embodiment of the present application;

[0037] Figure 4 is a schematic diagram of the electronic device provided by the embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0038] The following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, rather than all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art belong to the scope of protection of the present application.

[0039] First of all, it should be noted that with the emergence of AI (Artificial Intelligence) models, especially large-scale AI parameter models (abbreviated as large AI models), they have been applied in multiple scenarios and also provided new ideas for the intelligent generation of dialogue content. The so-called large AI model refers to a deep learning model containing a large number of parameters. Due to its large scale, this large AI model can store and process a large amount of information, thus achieving higher performance in various tasks. Therefore, it is possible to rely on the powerful natural language text understanding, logical thinking, and content generation capabilities of the pre-trained large AI model to generate dialogue content. In this way, if the seller user is not online when the buyer sends an inquiry, the large AI model can replace the seller to reply to the buyer's inquiry message.

[0040] The embodiments of the present application provide corresponding solutions based on the above ideas. However, the inventors of the present application found that in general, large AI models are generated in an end-to-end manner, that is, after inputting prompt information (including generation targets, control conditions, etc. described in natural language form) to the large AI model, the large AI model outputs specific content generation results. For example, in the relevant application scenarios of the embodiments of the present application, after the buyer sends an inquiry message, the conversation context information can be input to the large AI model, and the large AI model can output the conversation content for replying to the current inquiry message. In the above situation, if this end-to-end large AI model is directly used for generating conversation content, there are at least the following problems:

[0041] First, the entire system is too much of a black box and cannot solve the hallucination problem that may occur in large AI models. For example, in an e-commerce platform mainly for second-hand idle item transactions, assume that the item posted by a certain seller is clothing. Before the buyer decides to purchase, they may need to refer to the seller's height and weight information, so they ask about the seller's height and weight information in the inquiry message; or, some buyer users may want to pick up the item in person, so they ask where the seller lives, etc. However, the seller's height and weight, address, etc. information may not exist in the system. At this time, if an end-to-end large AI model is used, since the large AI model has to generate specific conversation content anyway, some seemingly reasonable but actually incorrect outputs may be generated.

[0042] Second, large AI models usually have a problem of being insensitive to numbers. However, in the conversation process between buyers and sellers, there are often contents related to prices, etc. Especially in a second-hand idle item transaction platform, the price of the item is usually not set fixedly, but can be communicated through the message system provided by the platform to determine a price acceptable to both parties, and then the transaction is completed at this price. Therefore, in the inquiry messages sent by buyers, there are even more contents related to price information. And price is closely related to specific numbers. Therefore, in the case where large AI models have the disadvantage of being insensitive to numbers, if the large AI model directly replies to the price-related inquiry messages initiated by buyer users, inaccurate or unreasonable reply contents may also be generated.

[0043] In view of the above situation, in the embodiments of the present application, refer to Figure 1, a service system for intelligently generating conversation content is provided, and the service system can be an agent system. First, the system is not implemented in an end-to-end form, but is split into two parts: policy selection and content generation (including dialogue generation, tone selection, etc.), and each part can correspond to an AI large model. Among them, the specific agent system can take the chat context as input, and can also include specific product information, etc. The policy selection module can be used to select the type of action to be executed currently. Here, the type of action can include both the specific type of response content (for example, closing a deal, initially proposing a price, proposing a price, proposing a price without a price, answering, polite greeting, etc.) when a response needs to be made, and can also include the situation of "not replying". After the policy selection module selects the specific type of action, the content generation module generates specific conversation content based on the selected type of action, and finally outputs a response statement. In this way, when the system cannot give an answer, it has the opportunity to "not reply" instead of forcing an answer. Among them, in the preferred manner, when the specific policy selection module is making a policy selection, it can also adopt a "two-way thinking" mode, that is, not only thinking about the actions that the seller can take next, but also thinking about the actions that the buyer may execute in the future, so as to make a more reasonable policy selection, or verify the rationality of the selected policy, etc.

[0044] In addition, in view of the situation that the AI large model is not sensitive to numbers, the embodiment of the present application also provides a separate price extraction module in the system. That is to say, after receiving the inquiry message from the buyer user, the price-related information can be extracted from it first, for example, including the quotation information of the buyer user, etc. Among them, the price extraction module can also be implemented by an AI large model. After extracting the price information, it can be provided to the policy selection module to enhance the system's sensitivity to digital information.

[0045] The following details the specific implementation solutions provided by the embodiments of the present application.

[0046] First, the embodiment of the present application provides a method for intelligently generating conversation content. Refer to Figure 2 , and the method can specifically include:

[0047] S201: Receive the inquiry information of the first user regarding the product posted by the second user.

[0048] In the embodiment of the present application, the buyer user in the commodity information service system may be referred to as the first user, and the seller user may be referred to as the second user, wherein the commodity information service system may be an ordinary e-commerce platform, or may be a trading service platform mainly for second-hand idle commodities (which may be referred to as a second-hand idle commodity trading platform), etc. After the second user publishes the information of a certain commodity, the first user may browse it in the platform, and if interested in the commodity, he may send an inquiry to the second user through the message system provided in the system. In particular, in the aforementioned second-hand idle commodity trading platform, since the commodity price is usually negotiable, the first user often sends a bargaining message to the second user through the message system.

[0049] S202: If the second user is offline, extract price-related information from the inquiry information using a first artificial intelligence (AI) large-scale parameter model.

[0050] After the first user sends an inquiry message to the second user, the system can determine whether the second user is online. If the second user is not online, the system can communicate with the first user on behalf of the second user. Of course, in actual applications, the hosting service can be provided with the consent of the second user, that is, before the second user uses the hosting service for the first time, the consent of the second user can be obtained first.

[0051] In the case where the second user agrees to use the hosting service, if the second user is not online when receiving the inquiry message from the first user, the first AI large-scale parameter model can be used to extract price-related information from the inquiry information. Among them, the reason for extracting price information through the AI ​​large model is that the description of price information in the inquiry message input by the first user may be varied, and some of these descriptions may require full natural language understanding, etc., in order to accurately extract price information. For example, the first user may ask: "Can I give you a 20% discount?", "Can I get a 20% discount?", or "Can I round it off?", etc. In the case where the first user has a high degree of freedom in inquiries, if a simple small model is used to extract price information, it may be impossible to accurately extract price information. Therefore, in the embodiment of the present application, the price extraction module can also be implemented by the AI ​​big model. For example, assuming that the first user says "Can I remove the decimal point?", the AI ​​big model can determine through context analysis and other methods that the first user may want the second user to remove the "decimal point" in the price. For example, assuming that the second user's price is 105 yuan, the second user may want to ask whether it is possible to remove the decimal point "5" in the price and take the integer "100", and then determine the actual quotation of the first user as "100" yuan, and so on.

[0052] S203: Use the second AI large-scale parameter model to infer the query information and price information extraction results, and determine the action that needs to be performed currently from multiple predefined action types.

[0053] After extracting the price information, a prompt message (Prompt) can be constructed based on the query information of the first user and the extracted price information. In addition, the prompt message can also include predefined multiple action type information. Afterwards, the prompt message can be input into the second AI big model so that the second AI big model determines the action that needs to be executed currently from multiple action types.

[0054] The specific action types may be predefined according to the needs of the actual system. For example, in a specific implementation scheme, the specifically defined action types and corresponding meanings may include:

[0055] Transaction: Agree with the buyer's offer and conclude the transaction;

[0056] Initial price: Propose a price that the system considers reasonable;

[0057] Propose a price: Propose a new price based on the existing quotation;

[0058] Propose a price without a price: Based on the existing quotation, hope to increase it a little bit, but do not specify a specific price;

[0059] Reject: reject the buyer's offer;

[0060] Greetings: A polite greeting;

[0061] Answer: Respond to product related questions.

[0062] The second AI large model can determine the action that needs to be executed currently from the above various action types according to the specific inquiry information and the result of price information extraction. Among them, if the first user makes an offer in the inquiry message, it is determined that the action that needs to be executed may be for the second user to make a new offer. At this time, it is also necessary to recalculate the specific price of the new offer. However, as mentioned above, the AI large model is not sensitive to numbers. If price calculations are directly performed in the AI large model, it may be difficult to obtain reasonable offer results. If the AI large model's ability in this regard is trained, it may consume more resources. Therefore, in the embodiments of the present application, a method of combining the AI large model with specific rules is also provided, that is, for the part of making a new offer, an offer module can be provided, and the offer module can be implemented by code, and the preset calculation logic is implemented in the code. In this way, if the action selected by the second AI large model is to make a new offer, the calculation can be performed through this offer module. Specifically, when calculating, the price information marked by the second user for the commodity can be obtained (that is, the commodity price set when the commodity is released, and this price can be displayed on the front-end page and can be viewed by the first user), the floor price information (that is, the lowest price that the second user can accept, and this information can be set in the system background and will not be displayed on the front-end page). In addition, since the offer information of the first user can also be extracted through the price extraction module (that is, the price proposed by the first user during the negotiation process), therefore, the above various price information and the preset calculation rules can be combined, and the foregoing offer module can calculate the appropriate new offer information. Through this method, the linkage between the AI large model algorithm and the rules can be realized, so as to ensure the smooth progress of the negotiation process.

[0063] In addition, during the process of the second AI large model making strategy selection, the "two-way thinking" method can also be adopted, that is, not only does it need to think about what strategy the second user will adopt to reply to the inquiry message of the first user, but it can also think about what actions the first user may take next, and then select a more suitable strategy, or reflect on the selected strategy. If it is not reasonable enough, the strategy selection can be re-performed, and so on. Among them, regarding two-way thinking, it can be realized by presenting the corresponding task to the second AI large model in the prompt information. For example, the following content can be included in the prompt information: "You are the second user, the inquiry message of the first user is *** and the context includes ***. You can first predict what the first user may ask in the future and select what actions the second user can perform currently", and so on. By using two-way thinking, the AI large model no longer generates responses passively based on user input, but can enable the AI large model to better achieve the negotiation goal.

[0064] For example, in the actual bargaining process, the first user may continuously lower the offer to test the bottom price of the second user. Specifically, for example, assume the second user's asking price is 100 yuan. A certain first user asks for the first time: "Is 90 okay?" At this time, if the second user's bottom price is 80 yuan, the strategy selected by the second large AI model may be to agree to the first user's offer. However, the new inquiry message sent by the first user later may be to continue the offer. For example: "Is 85 okay?" At this time, if the new offer is directly compared with the bottom price set by the second user, it is actually still higher than the bottom price. That is to say, the second user can continue to agree to the first user's new round of offer. However, if "two-way thinking" is added, the following prediction results may be obtained: If the second user continues to agree to the first user's new round of offer, the first user may still continue to lower the price next time, and at the same time, the system may seem not "smart" enough. Therefore, in the case of predicting that the first user may continue to test the bottom price in the future, the strategy selected by the second large AI model can be to reject the first user's offer, or not reply, etc.

[0065] S204: Use the third large-scale AI parameter model to generate the conversation content for the inquiry message based on the currently required action.

[0066] After determining the type of the currently required action, it can be input to the third large AI model, which can generate the conversation content for the inquiry message based on the currently required action. Among them, if the specific action is "proposing a price", that is, based on the existing offer, proposing a new price, the new offer can also be calculated by the offer module and input to the third large AI model to generate the conversation content. For example, the specific prompt information input to the third large AI model can include: "You are the second user. The first user initiated the following inquiry message ***. Now you need to make a new offer. The price is *** yuan. Please generate the reply content.", etc.

[0067] In specific implementation, to avoid the singularity of the generated conversation, some commonly used language skills can be additionally defined to make the generated statements richer. That is, the pre-defined commonly used language skill information can be provided to the third large AI model so that the third large AI model can generate the conversation content based on the currently required action and the commonly used language skill information.

[0068] For example, in a second-hand idle item trading service system, the specifically defined language skills can include:

[0069] Highlighting value: Highlighting the cost value, quality or bottom price of the product to show the reasonableness of the pricing.

[0070] Value-added services: In addition to the product itself, additional value is provided, such as free gifts, free delivery, etc.

[0071] Emotional strategy: Use friendly expressions, complaints, and empathy to resonate with the other party.

[0072] Comparison with the market: Compare the product with other products in the market to highlight the advantages of its own product.

[0073] Transaction guarantee: Promise to ensure the security and reliability of the transaction by providing after-sales service.

[0074] Create a sense of urgency: Remind that the product may soon be sold out or the price may increase.

[0075] Chat: Reply to the other party simply without using any techniques.

[0076] It should be noted that in the embodiments of this application, since the system is divided into two parts, namely strategy selection and conversation generation, and the AI large model is used to implement them respectively. Therefore, for inquiry messages such as asking the height, weight, and home address of the second user, the selected strategy can include "not reply". At this time, the subsequent third AI large model does not need to generate specific conversation content. That is to say, through the implementation method provided by the embodiments of this application, when the AI large model cannot give an accurate answer, it can not reply, so as to avoid the hallucination problem.

[0077] In addition, it should be noted that in the specific implementation, the embodiments of this application can output the reply statement in the form of a chain of thought. That is to say, while outputting the answer, the thinking process of the large model outputting the answer is also intuitively displayed, so as to guide the model to generate more accurate and logical answers, thereby improving the reasoning effect of the large model.

[0078] In summary, through the embodiments of the present application, after the first user sends an inquiry message to the second user, if the second user is offline, the intelligent hosting service system can reply on behalf of the second user. Specifically, the intelligent hosting service system can first extract price-related information from the inquiry message through the first AI large model. Then, it can reason based on the inquiry message and the price extraction result through the second AI large model, and select the action to be executed currently from multiple predefined action types. Then, it can generate conversation content based on the selected action through the third AI large model to reply to the inquiry message of the first user. Among them, since the system is divided into two parts: policy selection and content generation, the entire system is no longer passive in generating responses based on user input, but can determine whether to generate specific conversation content based on the specifically selected policy (action), thereby solving the hallucination problem existing in the end-to-end AI large model. In addition, for the problem that the AI large model is not sensitive to numbers, a dedicated price extraction module is also provided, so that in the conversation process mainly involving bargaining, the system can more sensitively obtain the price information therein, and then make more accurate and reasonable action selections and content generations.

[0079] In a preferred implementation manner, during the process of action selection by the second AI large model, the actions that the first user may take in the future can also be predicted through a two-way thinking method, and then the strategy can be selected according to the prediction result, or the selected strategy can be reflected on, and so on. In this way, since the strategy of the opponent can be considered to select its own strategy, the large model can better achieve the negotiation goal.

[0080] In addition, when the strategy selected by the second AI large model is to re-bid, the price can be specifically calculated through a separate bidding module. The bidding module can calculate the price through rules to realize the linkage between the AI large model and the rules, so as to ensure the smooth progress of the bargaining process. Furthermore, during the process of content generation by the third AI large model, multiple optional language skills can also be provided to make the generated sentences more rich.

[0081] It should be noted that the embodiments of the present application may involve the use of user data. In actual applications, user-specific personal data can be used in the solutions described herein within the scope permitted by applicable laws and regulations (for example, when the user clearly consents and is effectively notified to the user, etc.) in compliance with the applicable laws and regulations of the country where it is located.

[0082] Corresponding to the foregoing method embodiments, the embodiments of the present application also provide a device for intelligently generating conversation content. See Figure 3 and this device may include:

[0083] An inquiry message receiving unit 301, configured to receive inquiry information in which a first user inquires about a product posted by a second user;

[0084] A price extraction unit 302, configured to, if the second user is in an offline state, extract price-related information from the inquiry information by using a first large-scale artificial intelligence (AI) parameter model;

[0085] A policy selection unit 303, configured to perform inference on the inquiry information and the price information extraction result by using a second large-scale AI parameter model, and determine an action that needs to be performed currently from a plurality of predefined action types;

[0086] A conversation statement generation unit 304, configured to generate conversation content for the inquiry information based on the action that needs to be performed currently by using a third large-scale AI parameter model.

[0087] When specifically implemented, the second large-scale AI parameter model can also be used for:

[0088] Predicting actions that the first user may perform in the future according to conversation context information, so that the second large-scale AI parameter model determines the action that needs to be performed currently in a two-way thinking manner, or reflecting on the rationality of the action that needs to be performed currently.

[0089] Among them, if it is predicted that an action that the first user may perform in the future is to continue to make an offer to probe the floor price of the second user, the second large-scale AI model can determine the action that needs to be performed currently as rejecting the current offer of the first user.

[0090] If the action that needs to be performed currently determined by the second large-scale AI parameter model is to make a new offer, the device may further include:

[0091] A marked price and floor price determination unit, configured to determine the marked price and floor price information set by the second user for the product;

[0092] An offer information determination unit, configured to determine the offer information of the first user for the product according to the price information extraction result of the first large-scale AI parameter model;

[0093] A bid calculation unit, configured to calculate new bid information according to the marked price, floor price, offer information of the first user, and preset rule information of the product;

[0094] A bid information input unit, configured to input the new bid information to the third large-scale AI parameter model, so that the third large-scale AI parameter model generates the conversation content based on the action that needs to be performed currently and the new bid information.

[0095] Among them, the actions that need to be executed currently determined by the second large-scale AI parameter model include: not replying.

[0096] In addition, the device may further include:

[0097] A language skill providing unit, configured to provide predefined common language skill information to the third large-scale AI parameter model, so that the third large-scale AI parameter model generates conversation content based on the currently required action and the common language skill information.

[0098] Among them, when outputting the conversation content, the third large-scale AI parameter model also outputs content related to the thinking process.

[0099] In addition, an embodiment of the present application also provides a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, it implements the steps of the method described in any one of the foregoing method embodiments.

[0100] And an electronic device, including:

[0101] One or more processors; and

[0102] A memory associated with the one or more processors, where the memory is used to store program instructions, and when the program instructions are read and executed by the one or more processors, they implement the steps of the method described in any one of the foregoing method embodiments.

[0103] A computer program product, including a computer program / computer executable instructions, and when the computer program / computer executable instructions are executed by a processor in an electronic device, they implement the steps of the method described in the foregoing method embodiments.

[0104] Among them, Figure 4 An exemplary architecture of the electronic device is shown, which may specifically include a processor 410, a video display adapter 411, a disk drive 412, an input / output interface 413, a network interface 414, and a memory 420. The above-mentioned processor 410, video display adapter 411, disk drive 412, input / output interface 413, network interface 414, and the memory 420 can be communicatively connected through a communication bus 430.

[0105] Among them, the processor 410 can be implemented in the form of a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, etc., and is used to execute relevant programs to implement the technical solutions provided in this application.

[0106] The memory 420 can be implemented in the form of a ROM (Read Only Memory), a RAM (Random Access Memory), a static storage device, a dynamic storage device, etc. The memory 420 can store an operating system 421 for controlling the operation of the electronic device 400, and a basic input / output system (BIOS) for controlling the low-level operations of the electronic device 400. In addition, a web browser 423, a data storage management system 424, and an intelligent dialogue content generation processing system 425, etc. can also be stored. The above intelligent dialogue content generation processing system 425 can be the application program that specifically implements the operations of the foregoing steps in the embodiments of this application. In short, when implementing the technical solutions provided in this application through software or firmware, the relevant program codes are stored in the memory 420 and called and executed by the processor 410.

[0107] The input / output interface 413 is used to connect to an input / output module to implement information input and output. The input / output module can be configured as a component in the device (not shown in the figure) or externally connected to the device to provide corresponding functions. Among them, the input device can include a keyboard, a mouse, a touch screen, a microphone, various sensors, etc., and the output device can include a display, a speaker, a vibrator, an indicator light, etc.

[0108] The network interface 414 is used to connect to a communication module (not shown in the figure) to implement communication interaction between this device and other devices. Among them, the communication module can implement communication in a wired manner (such as USB, network cable, etc.) or in a wireless manner (such as mobile network, WIFI, Bluetooth, etc.).

[0109] The bus 430 includes a path for transmitting information between various components of the device (such as the processor 410, the video display adapter 411, the disk drive 412, the input / output interface 413, the network interface 414, and the memory 420).

[0110] It should be noted that although the above device only shows the processor 410, video display adapter 411, disk drive 412, input / output interface 413, network interface 414, memory 420, bus 430, etc., in the specific implementation process, the device may also include other components necessary for normal operation. In addition, those skilled in the art can understand that the above device may also only include the components necessary for implementing the solution of the present application, and does not necessarily include all the components shown in the figure.

[0111] From the description of the above embodiments, those skilled in the art can clearly understand that the present application can be implemented by means of software plus a necessary general hardware platform. Based on this understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) to execute the methods described in various embodiments or some parts of the embodiments of the present application.

[0112] Each embodiment in this specification is described in a progressive manner. The same or similar parts among the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for a system or system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment. The systems and system embodiments described above are only illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they can be located in one place, or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative efforts.

[0113] The above has introduced in detail the method and electronic device for intelligently generating conversation content provided by the present application. Specific examples are used in this article to elaborate on the principle and implementation manner of the present application. The description of the above embodiments is only used to help understand the method and its core idea of the present application; at the same time, for those of ordinary skill in the art, according to the idea of the present application, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation to the present application.

Claims

1. A method for intelligently generating conversation content, characterized in that, Including: Receiving inquiry information from a first user regarding a product posted by a second user; If the second user is offline, using a first large-scale artificial intelligence (AI) parameter model to extract price-related information from the inquiry information; Using a second large-scale AI parameter model to reason about the inquiry information and the price information extraction result, and determining an action that needs to be performed currently from a predefined plurality of action types; Using a third large-scale AI parameter model to generate conversation content for the inquiry information based on the currently required action.

2. The method according to claim 1, wherein: When determining the action that needs to be performed currently, the second large-scale AI parameter model is further configured to: Predict actions that the first user may perform in the future based on conversation context information, so that the second large-scale AI parameter model determines the action that needs to be performed currently or reflects on the rationality of the currently required action in a two-way thinking manner.

3. The method according to claim 2, wherein: If it is predicted that the action that the first user may perform in the future is to continue to quote to test the bottom price of the second user, then determine the currently required action as rejecting the current quote of the first user.

4. The method according to claim 1, wherein: If the action that needs to be performed currently determined by the second large-scale AI parameter model is to make a new bid, the method further includes: Determining the marked price and bottom price information set by the second user for the product; Determining the quote information of the first user for the product according to the price information extraction result of the first large-scale AI parameter model; Calculating new bid information according to the marked price, bottom price, the first user's quote information of the product and preset rule information; Inputting the new bid information into the third large-scale AI parameter model, so that the third large-scale AI parameter model generates the conversation content based on the currently required action and the new bid information.

5. The method according to claim 1, wherein: The action that needs to be performed currently determined by the second large-scale AI parameter model includes: not replying.

6. The method according to claim 1, characterized in that It further includes: Providing predefined common language skill information to the third large-scale AI parameter model, so that the third large-scale AI parameter model generates conversation content based on the currently required action and the common language skill information.

7. The method according to claim 1, wherein: When outputting the conversation content, the third large-scale AI parameter model also outputs content related to the thinking process.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

9. An electronic device, characterized in that, Including: One or more processors; And A memory associated with the one or more processors, the memory being used to store program instructions, and when the program instructions are read and executed by the one or more processors, they execute the steps of the method according to any one of claims 1 to 7.

10. A computer program product, comprising a computer program / computer-executable instructions, characterized in that, When the computer program / computer-executable instructions are executed by a processor in an electronic device, the steps of the method according to any one of claims 1 to 7 are implemented.

Images