Large model dialogue method, device and equipment based on multiple media and medium
By establishing outbound call and private domain communication channels during the enterprise sales process, and combining multimedia dialogue record generation model, accurate customer communication and subsequent business operations are achieved, the problems of low customer conversion rates and high costs are solved, and customer conversion rates and business benefits are improved.
Patent Information
- Application Number
- CN202510748492.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-06
- Publication Date
- 2025-08-08
AI Technical Summary
The business knowledge of corporate sales personnel is limited, which makes it difficult for customers to answer customer questions accurately during the call. The customer conversion rate is low, and the telephone sales method can only achieve single communication, which increases customer follow-up costs and causes dissatisfaction, and has low business benefits.
By establishing an out-of-call communication channel with the client, sending a channel establishment request message for the private domain communication channel. After successfully establishing a private domain communication channel, the client message is received and the current dialogue record is generated. The pre-established message generation model is used to generate the sub-end message, and combined with the history records of the out-of-call and private domain communication channel, accurate reply is achieved and subsequent business operations are continued based on the private domain communication channel.
It improves customer conversion rate, saves business operation costs, provides multiple business operation channels, and improves business revenue.
Smart Images
Figure CN120455593A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of artificial intelligence technology, and more specifically, to a large-scale model dialogue method, apparatus, device, and medium based on multi-media. Background Art
[0002] Currently, corporate sales personnel primarily promote and sell corporate businesses, such as financial insurance, to customers over the phone. On the one hand, these sales personnel have limited business knowledge and struggle to accurately answer customer questions during calls, leaving customers unclear about the company's services and other information, leading to low customer conversion rates. On the other hand, this telephone sales method only allows for a single conversation. After a single call, sales personnel often need to continue with potential customers multiple times to ensure continued business operations. This increases customer follow-up costs over the long term and can easily lead to customer dissatisfaction with frequent calls, resulting in lower business returns. Summary of the Invention
[0003] The purpose of the embodiments of the present application is to provide a large-scale model dialogue method, device, equipment and medium based on multimedia to achieve the technical effect of improving customer conversion rate and business revenue.
[0004] In a first aspect, an embodiment of the present application provides a large model dialogue method based on multimedia, including: Establishing an outbound communication channel with the client, and sending a channel establishment request message for a private domain communication channel to the client through the outbound communication channel; When the private domain communication channel is successfully established, receiving a current round of client messages sent by the client through the outbound communication channel or the private domain communication channel; Generate a current conversation record based on the current round of client messages, a first historical conversation record based on the outbound communication channel, and a second historical conversation record based on the private domain communication channel; The current conversation record is input into a pre-established message generation model, and the obtained current turn-side message is returned to the client.
[0005] In the above implementation process, an outbound communication channel is established with the client, and a channel establishment request message for the private domain communication channel is sent to the client through the outbound communication channel. When the private domain communication channel is successfully established, the current round client message sent by the client through the outbound communication channel or the private domain communication channel is received. According to the current round client message, the first historical conversation record based on the outbound communication channel, and the second historical conversation record based on the private domain communication channel, a current conversation record is generated. The current conversation record is input into a pre-established message generation model, and the current round end message obtained is returned to the client. On the one hand, during an outbound call process with the client, the message generation model can be used to combine the current conversation records on multiple communication channels such as the outbound communication channel and the private domain communication channel to generate the current round end message, thereby ensuring the accurate generation of the current round end message and improving the customer conversion rate. On the other hand, after the outbound call with the client ends, the private domain communication channel can continue to interact with the client in conversation to carry out subsequent business operations, effectively saving business operation costs and increasing business revenue.
[0006] Furthermore, before sending a channel establishment request message for a private domain communication channel to the client through the outbound communication channel, the method further includes: Determining that the private domain communication channel is not established with the client; A channel establishment inquiry message for the private domain communication channel is sent to the client through the outbound communication channel, and it is determined that a channel establishment consent message returned by the client is received.
[0007] In the above implementation process, by first sending a channel establishment inquiry message for the private domain communication channel to the client through the outbound communication channel when it is determined that a private domain communication channel has not been established with the client, and then sending a channel establishment request message for the private domain communication channel to the client through the outbound communication channel after confirming that a channel establishment consent message returned by the client has been received, the client can be asked in advance whether it agrees to establish the private domain communication channel, thereby effectively improving the success rate of establishing the private domain communication channel.
[0008] Furthermore, the current round of client messages is a multimodal message sent by the client through the private domain communication channel; Before generating the current conversation record according to the current round of client messages, the first historical conversation record based on the outbound communication channel, and the second historical conversation record based on the private domain communication channel, the method further includes: Determining a target modality message recognition model corresponding to the message modality according to the message modality of the current round of client messages; Inputting the current round of client messages into the target modal message recognition model to obtain a recognition result; Concatenate the current round client message and the recognition result to obtain a target current round client message; The current round client message is updated to the target current round client message.
[0009] In the above implementation process, when the current round client message is a multimodal message sent by the client through the private domain communication channel, the target modal message recognition model corresponding to the message modality of the current round client message is used to identify the current round client message, and the obtained recognition result domain is spliced with the current round client message as a new current round client message. This can ensure that the subsequent message generation model accurately understands the content of the current round client message, which is conducive to further ensuring that the message generation model accurately generates the current round local message, thereby better improving the customer conversion rate.
[0010] Furthermore, generating a current conversation record based on the current round of client messages, a first historical conversation record based on the outbound communication channel, and a second historical conversation record based on the private domain communication channel includes: The current round of client messages, all messages in the first historical conversation record, and all messages in the second historical conversation record are concatenated in order of message sending and receiving times from earliest to latest to obtain the current conversation record.
[0011] In the above implementation process, the current round of client messages, all messages in the first historical conversation record, and all messages in the second historical conversation record are spliced in the order of message sending and receiving time from early to late to obtain the current conversation record. This can completely and orderly merge the current conversation records on multiple communication channels, ensuring that the message generation model accurately generates the current round of local-side messages.
[0012] Furthermore, the message generation model includes a question extraction agent, an intent recognition agent, and a message generation agent; The step of inputting the current conversation record into a pre-established message generation model and returning the obtained current turn-side message to the client includes: Extracting at least one customer question from the current conversation record by the question extraction agent; identifying the customer's intention based on the at least one customer question by the intention identification agent; An intelligent agent is generated through the message to determine an answer that meets the customer's intention, and the current turn-end message is generated based on the answer.
[0013] In the above implementation process, by pre-establishing a message generation model including a question extraction agent, an intent recognition agent and a message generation agent, the question extraction agent extracts at least one customer question from the current conversation record, the intent recognition agent identifies the customer intention based on at least one customer question, the message generation agent determines the answer that meets the customer intention, and generates the current round-end message based on the answer. This can collaborate with multiple agents to accurately and quickly identify the user intention in the current conversation record and determine the answer that meets the user intention to generate the current round-end message, which is conducive to further improving the processing efficiency of the message generation model.
[0014] Furthermore, the method further comprises: In the case that the current round of the client's message fails to pass the user's review, the target agent terminal is triggered to interact with the client through the outbound communication channel.
[0015] In the above implementation process, by triggering the target agent end to interact with the client through the outbound communication channel when the current round-end message fails to pass the user review, the abnormal current round-end message can be discovered in time, and the outbound call can be transferred to the target agent end for manual processing in time, which is conducive to improving customer experience and better improving customer conversion rate.
[0016] Furthermore, the number of the outbound communication channel is one, and the number of the private domain communication channel is at least one.
[0017] In the above implementation process, by conducting dialogue and interaction with the client based on multiple communication channels including an outbound call communication channel and at least one private domain communication channel, the normal execution of an outbound call process with the client can be guaranteed, thereby improving the customer conversion rate. After an outbound call with the client is completed, a variety of business operation channels are provided to the user, making it convenient for the user to flexibly choose business operation channels and increasing business revenue.
[0018] In a second aspect, an embodiment of the present application provides a large-scale model dialogue device based on multimedia, including: A channel establishment module is configured to establish an outbound communication channel with a client, and send a channel establishment request message for a private domain communication channel to the client via the outbound communication channel; a message receiving module, configured to receive a current round of client messages sent by the client through the outbound communication channel or the private communication channel when the private communication channel is successfully established; a message processing module, configured to generate a current conversation record based on the current round of client messages, a first historical conversation record based on the outbound communication channel, and a second historical conversation record based on the private domain communication channel; The message sending module is used to input the current conversation record into a pre-established message generation model and return the obtained current turn-side message to the client.
[0019] In a third aspect, an embodiment of the present application provides an electronic device, comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor; when the processor executes the computer program, the method described above is implemented.
[0020] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, which includes a stored computer program; wherein, when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the method described above. BRIEF DESCRIPTION OF THE DRAWINGS
[0021] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments of the present application. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without creative work.
[0022] Figure 1 A flowchart of a large model dialogue method based on multimedia provided in the first embodiment of the present application; Figure 2 A schematic structural diagram of a large-scale multimedia dialogue device provided in accordance with the second embodiment of the present application; Figure 3 A schematic structural diagram of an electronic device provided in the third embodiment of the present application. DETAILED DESCRIPTION
[0023] The technical solutions in the embodiments of the present application will be described below in conjunction with the drawings in the embodiments of the present application.
[0024] It should be noted that in the description of this application, the terms "first" and "second" are used only to distinguish descriptions and should not be understood to indicate or imply relative importance. Furthermore, the step numbers herein are used only to facilitate the explanation of the embodiments of this application and do not limit the order in which the steps are to be executed.
[0025] In related technologies, sales personnel primarily promote and sell corporate businesses, such as financial insurance, to customers via telephone. On the one hand, these sales personnel have limited business knowledge and struggle to accurately answer customer questions during calls, leaving customers unclear about the company's service offerings and other information, leading to low customer conversion rates. Furthermore, this telephone sales method only allows for a single conversation. After a single call, sales personnel often need to continue with multiple calls with potential customers to ensure continued business operations. This increases customer follow-up costs over the long term and can easily lead to customer dissatisfaction with frequent calls, resulting in lower business returns.
[0026] To this end, the present application proposes a large-model dialogue method based on multi-media, which establishes an outbound communication channel with the client, sends a channel establishment request message for the private domain communication channel to the client through the outbound communication channel, and receives the current round client message sent by the client through the outbound communication channel or the private domain communication channel when the private domain communication channel is successfully established. According to the current round client message, the first historical dialogue record based on the outbound communication channel and the second historical dialogue record based on the private domain communication channel, the current dialogue record is generated, the current dialogue record is input into a pre-established message generation model, and the current round end message obtained is returned to the client. On the one hand, during an outbound call with the client, the message generation model can be used to combine the current dialogue records on multiple communication channels of the outbound communication channel and the private domain communication channel to generate the current round end message, ensuring the accurate generation of the current round end message and improving the customer conversion rate. On the other hand, after an outbound call with the client is completed, the private domain communication channel can continue to interact with the client for subsequent business operations, effectively saving business operation costs and increasing business revenue.
[0027] The technical solutions in the embodiments of the present application will be described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments.
[0028] The method provided in the embodiment of the present application can be executed by a relevant terminal device, and the following description will be given using a user terminal as an example of the execution subject.
[0029] It should be noted that user terminals include terminal devices such as mobile phones, tablets or computers held by enterprise staff.
[0030] Please see Figure 1 , Figure 1 The first embodiment of the present application provides a large model dialogue method based on multimedia, including steps S101 to S104: S101: Establish an outbound communication channel with a client, and send a channel establishment request message for a private domain communication channel to the client through the outbound communication channel.
[0031] As an example, the types of outbound communication channels and private domain communication channels are predetermined according to the business operation requirements of the enterprise, and any client is selected according to actual business sales needs.
[0032] It should be noted that outbound call communication channels refer to communication channels established based on outbound call management systems, including traditional telephone communication channels and VoIP (Voice over Internet Protocol) communication channels, and private domain communication channels refer to communication channels established based on private domain business systems, including WeChat communication channels and DingTalk communication channels, among other types of communication channels.
[0033] The user terminal and the client establish an outbound call communication channel. After the outbound call communication channel is successfully established, the user terminal and the client can conduct a dialogue interaction through the outbound call communication channel.
[0034] The user terminal sends a channel establishment request message for the private domain communication channel to the client through the outbound call communication channel, so that the client can receive the channel establishment request message for the private domain communication channel and choose whether to establish the private domain communication channel with the user terminal.
[0035] For example, assuming that the private domain communication channel is a WeChat communication channel, the user terminal can use the enterprise WeChat software system on which the WeChat communication channel is established to obtain an access link for adding the user's enterprise WeChat account, and then generate a channel establishment request message including this access link, and send the channel establishment request message to the client through the outbound communication channel, so that the client can receive the channel establishment request message and choose whether to click on this access link to add the user's enterprise WeChat account, so as to choose whether to establish a private domain communication channel with the user terminal.
[0036] S102: When the private domain communication channel is successfully established, a current round of client messages sent by the client through the outbound communication channel or the private domain communication channel is received.
[0037] As an example, after the user terminal sends a channel establishment request message for the private communication channel to the client via the outbound communication channel, the user terminal detects in real time whether the private communication channel has been successfully established. If the private communication channel is successfully established, the user terminal and the client can interact with each other via the private communication channel. In other words, the client can now interact with the user terminal via both the outbound communication channel and the private communication channel.
[0038] The user terminal monitors the traffic on the outbound communication channel in real time, and after monitoring the current round of client messages sent by the client via the outbound communication channel, receives the current round of client messages sent by the client via the outbound communication channel. The user terminal also monitors the traffic on the private communication channel in real time, and after monitoring the current round of client messages sent by the client via the private communication channel, receives the current round of client messages sent by the client via the private communication channel.
[0039] S103: Generate a current conversation record based on the current round of client messages, a first historical conversation record based on the outbound communication channel, and a second historical conversation record based on the private domain communication channel.
[0040] As an example, after receiving the current round of client messages, the user terminal obtains a first historical conversation record based on the outbound communication channel and a second historical conversation record based on the private domain communication channel.
[0041] It should be noted that the first historical conversation record refers to the conversation record obtained by recording all messages exchanged between the user terminal and the client through the outbound communication channel before the current round of client messages are received, and the second historical conversation record refers to the conversation record obtained by recording all messages exchanged between the user terminal and the client through the private domain communication channel before the current round of client messages are received.
[0042] The user terminal generates a current conversation record based on the current round of client messages, the first historical conversation record based on the outbound communication channel, and the second historical conversation record based on the private domain communication channel.
[0043] S104: Input the current conversation record into a pre-established message generation model, and return the obtained current turn-side message to the client.
[0044] As an example, after obtaining the current conversation record, the user terminal inputs the current conversation record into a pre-established message generation model to obtain the current turn-side message, and returns the current turn-side message to the client.
[0045] Among them, if the current round of client messages is sent by the client to the user terminal through the outbound communication channel, the user terminal returns the current round of local end messages to the client through the outbound communication channel, and updates and stores the first historical conversation record; if the current round of client messages is sent by the client to the user terminal through the private domain communication channel, the current round of local end messages is returned to the client through the private domain communication channel, and the second historical conversation record is updated and stored.
[0046] In practical applications, large models can be used as message generation models. Large models refer to machine learning models with large-scale parameters and complex computing structures, which can process multiple different types of data in parallel on a large scale.
[0047] Since the user terminal and the client can interact with each other through the outbound communication channel after the outbound communication channel is successfully established, when only the outbound communication channel is established with the client, the user terminal will receive the current round client message sent by the client through the outbound communication channel, obtain the current conversation record based on the current round client message and the first historical conversation record based on the outbound communication channel, input the current conversation record into the message generation model, and return the obtained current round local end message to the client.
[0048] The embodiment of the present application establishes an outbound communication channel with the client, sends a channel establishment request message for the private domain communication channel to the client through the outbound communication channel, and when the private domain communication channel is successfully established, receives the current round client message sent by the client through the outbound communication channel or the private domain communication channel, generates a current conversation record based on the current round client message, the first historical conversation record based on the outbound communication channel, and the second historical conversation record based on the private domain communication channel, inputs the current conversation record into a pre-established message generation model, and returns the obtained current round end message to the client. On the one hand, during an outbound call with the client, the message generation model can be used to combine the current conversation records on multiple communication channels of the outbound communication channel and the private domain communication channel to generate the current round end message, thereby ensuring the accurate generation of the current round end message and improving the customer conversion rate. On the other hand, after an outbound call with the client ends, the private domain communication channel can continue to interact with the client in conversation based on the private domain communication channel to carry out subsequent business operations, effectively saving business operation costs and improving business revenue.
[0049] In an optional embodiment, before sending a channel establishment request message for a private domain communication channel to the client through the outbound communication channel, the method further includes: determining that a private domain communication channel has not been established with the client; sending a channel establishment inquiry message for the private domain communication channel to the client through the outbound communication channel, and determining that a channel establishment consent message returned by the client is received.
[0050] As an example, after the outbound communication channel is successfully established, the user terminal detects whether a private domain communication channel is established with the client. If it is determined that a private domain communication channel is not established with the client, a channel establishment inquiry message for the private domain communication channel is sent to the client through the outbound communication channel, so that the client can receive the channel establishment inquiry message for the private domain communication channel and choose whether to agree to establish a private domain communication channel with the user terminal.
[0051] After receiving the channel establishment approval message returned by the client, the user terminal sends a channel establishment request message for the private domain communication channel to the client through the outbound communication channel; after receiving the channel establishment rejection message returned by the client, the user terminal only interacts with the client through the outbound communication channel.
[0052] In actual applications, when the user terminal determines to establish a private domain communication channel with the client, it can directly monitor the traffic on the private domain communication channel in real time.
[0053] In an embodiment of the present application, when it is determined that a private domain communication channel has not been established with the client, the client is first sent a channel establishment inquiry message for the private domain communication channel through the outbound communication channel. After it is determined that a channel establishment consent message returned by the client is received, the client is sent a channel establishment request message for the private domain communication channel through the outbound communication channel. This can inquire the client in advance whether it agrees to establish the private domain communication channel, thereby effectively improving the success rate of establishing the private domain communication channel.
[0054] In an optional embodiment, the current round client message is a multimodal message sent by the client through the private domain communication channel; before generating the current conversation record based on the current round client message, the first historical conversation record based on the outbound communication channel and the second historical conversation record based on the private domain communication channel, it also includes: determining a target modality message recognition model corresponding to the message modality based on the message modality of the current round client message; inputting the current round client message into the target modality message recognition model to obtain a recognition result; splicing the current round client message and the recognition result to obtain a target current round client message; and updating the current round client message to the target current round client message.
[0055] As an example, considering the actual application scenario, the client can send multimodal messages such as text, documents, images, videos and audio to the user terminal through the private domain communication channel.
[0056] When the current round of client messages sent by the client to the user terminal through the private domain communication channel is a multimodal message, the user terminal determines the target modality message recognition model corresponding to the message modality according to the message modality of the current round of client messages.
[0057] In practical applications, multiple modal message recognition models can be pre-established, and according to the message modality of the current round of client messages, a target modal message recognition model corresponding to the message modality is selected from the multiple modal message recognition models.
[0058] For example, assuming that the message modality of the current round of client messages is text, the user terminal selects the text recognition model corresponding to the message modality from multiple modal message recognition models as the target modal message recognition model; assuming that the message modality of the current round of client messages is document, the user terminal selects the document recognition model corresponding to the message modality from multiple modal message recognition models as the target modal message recognition model; assuming that the message modality of the current round of client messages is image, the user terminal selects the image recognition model corresponding to the message modality from multiple modal message recognition models as the target modal message recognition model; assuming that the message modality of the current round of client messages is video, the user terminal selects the video recognition model corresponding to the message modality from multiple modal message recognition models as the target modal message recognition model; assuming that the message modality of the current round of client messages is audio, the user terminal selects the audio recognition model corresponding to the message modality from multiple modal message recognition models as the target modal message recognition model.
[0059] After determining the target modal message recognition model, the user terminal inputs the current round of client messages into the target modal message recognition model to obtain a recognition result.
[0060] Among them, the recognition process of the target modal message recognition models such as text recognition model, document recognition model, image recognition model, video recognition model and audio recognition model for the current round of client messages can refer to the recognition process of the existing corresponding models and will not be repeated here.
[0061] After obtaining the recognition result, the user terminal concatenates the current round client message and the recognition result to obtain the target current round client message, and updates the current round client message to the target current round client message.
[0062] In an embodiment of the present application, when the current round of client messages is a multimodal message sent by the client through a private domain communication channel, a target modal message recognition model corresponding to the message modality of the current round of client messages is used to identify the current round of client messages, and the obtained recognition result is spliced with the current round of client messages as a new current round of client messages. This can ensure that the subsequent message generation model accurately understands the content of the current round of client messages, which is conducive to further ensuring that the message generation model accurately generates the current round of local-end messages, thereby better improving customer conversion rates.
[0063] In an optional embodiment, the current conversation record is generated based on the current round of client messages, the first historical conversation record based on the outbound communication channel, and the second historical conversation record based on the private domain communication channel, including: splicing the current round of client messages, all messages in the first historical conversation record, and all messages in the second historical conversation record in order of message sending and receiving time from early to late to obtain the current conversation record.
[0064] As an example, after receiving the current round of client messages and obtaining the first historical conversation record based on the outbound communication channel and the second historical conversation record based on the private domain communication channel, the user terminal determines the receiving time of the current round of client messages, determines the sending and receiving times of all messages in the first historical conversation record, and determines the sending and receiving times of all messages in the second historical conversation record.
[0065] The user terminal concatenates the current round of client messages, all messages in the first historical conversation record, and all messages in the second historical conversation record in the order of message sending and receiving times from earliest to latest to obtain the current conversation record.
[0066] The embodiment of the present application obtains the current conversation record by splicing the current round of client messages, all messages in the first historical conversation record, and all messages in the second historical conversation record in the order of message sending and receiving time from early to late. This can completely and orderly merge the current conversation records on multiple communication channels, ensuring that the message generation model accurately generates the current round of local-side messages.
[0067] In an optional embodiment, the message generation model includes a question extraction agent, an intent recognition agent, and a message generation agent; the current conversation record is input into a pre-established message generation model, and the current turn-end message is returned to the client, including: extracting at least one customer question from the current conversation record through a question extraction agent; identifying the customer intent based on at least one customer question through an intent recognition agent; determining an answer that meets the customer intent through a message generation agent, and generating the current turn-end message based on the answer.
[0068] As an example, considering that the data composition of the current conversation record is relatively complex, in order to ensure accurate and rapid responses to customers, a message generation model is pre-built that includes a question extraction agent, an intent recognition agent, and a message generation agent.
[0069] In practical applications, a system model including a question extraction agent, an intent recognition agent, and a message generation agent can be constructed based on the large model as a message generation model.
[0070] An agent is an agent that can perceive its environment and take actions to achieve specific goals. It can be software, hardware, or a system, and possesses autonomy, adaptability, and interaction. An agent perceives changes in its environment (e.g., through sensors or data input), makes judgments and decisions based on its learned knowledge and algorithms, and then executes actions to influence the environment or achieve a predetermined goal.
[0071] The question extraction agent is an agent trained using a large amount of sample data, such as sample conversation records, and at least one customer question corresponding to the sample conversation records, and is used to achieve the goal of extracting at least one customer question from the current conversation record.
[0072] The intent recognition agent is an agent trained using a large amount of sample data, such as at least one customer question corresponding to sample conversation records, and the customer intent corresponding to these customer questions, to achieve the goal of identifying customer intent based on at least one customer question.
[0073] The message generation agent is an agent trained using a large amount of sample data, such as the customer intent corresponding to these customer questions, and the knowledge answers in the pre-built business knowledge base. It is used to achieve the goal of generating reply messages based on answers that meet the customer intent.
[0074] The training method of the above-mentioned intelligent agent can refer to the existing intelligent agent training method, which will not be described in detail here.
[0075] After obtaining the current conversation record, the user terminal inputs the current conversation record into a pre-built message generation model, extracts at least one customer question from the current conversation record through the question extraction agent, identifies the customer intention based on at least one customer question through the intent recognition agent, determines the answer that meets the customer intention through the message generation agent, and generates the current turn-end message based on the answer.
[0076] In practical applications, the question extraction agent can locate and extract the current round of client messages and the historical conversation records of this outbound call from the current conversation records, extract features from the current round of client messages and the historical conversation records of this outbound call respectively, cluster the multiple features obtained, and understand the internal logic of each type of feature to sort out each customer problem.
[0077] The intent recognition agent can extract keywords from each customer question, perform preprocessing such as deduplication on multiple keywords, match the preprocessed keywords with each candidate intent in the pre-built candidate intent library, and determine the candidate intent that matches the most keywords as the customer intent.
[0078] The message generation agent can query the answer that meets the customer's intention from the pre-built business knowledge base, and fill the answer into the message template corresponding to the customer's intention to generate the current round-end message.
[0079] The embodiment of the present application pre-establishes a message generation model including a question extraction agent, an intent recognition agent and a message generation agent. The question extraction agent extracts at least one customer question from the current conversation record, the intent recognition agent identifies the customer intention based on at least one customer question, the message generation agent determines the answer that meets the customer intention, and generates the current round-end message based on the answer. It can coordinate multiple agents to accurately and quickly identify the user intention in the current conversation record and determine the answer that meets the user intention to generate the current round-end message, which is conducive to further improving the processing efficiency of the message generation model.
[0080] In an optional embodiment, the method further includes step S105: S105: If the current round of the local end message fails to pass the user's review, the target agent end is triggered to interact with the client through the outbound communication channel.
[0081] As an example, after obtaining the current round-end message output by the message generation model, the user terminal waits to obtain the review result input by the user for the current round-end message. If it is determined that the current round-end message has passed the user review, the current round-end message is returned to the client; if it is determined that the current round-end message has not passed the user review, the target agent end is selected, such as an idle agent end, and the target agent end is triggered to interact with the client through the outbound communication channel.
[0082] In actual applications, users can review multiple review items, such as whether the content of the current round of local messages is clear, and whether it can solve the customer problems in the current round of client messages. If any review item fails, the target agent will be triggered to interact with the client through the outbound communication channel to transfer the outbound call to the target agent for manual processing.
[0083] The embodiment of the present application triggers the target agent end to interact with the client through the outbound communication channel when the current round-end message fails to pass the user review. It can promptly discover abnormal current round-end messages and promptly transfer the outbound call to the target agent end for manual processing, which is conducive to improving customer experience and better improving customer conversion rate.
[0084] In an optional embodiment, the number of the outbound communication channel is one, the number of the outbound communication channel is one, and the number of the private domain communication channel is at least one.
[0085] As an example, considering that the telephone communication resources of the client are limited, in order to avoid occupying the telephone communication resources of the client, the type of the outbound communication channel is predetermined, and the number of the outbound communication channel is determined to be one.
[0086] Taking into account the diversity of corporate business operations, it may support the use of multiple private domain business systems such as WeChat and DingTalk for business operations, pre-determine the type of private domain communication channels, and determine the number of private domain communication channels to be at least one.
[0087] The embodiment of the present application can ensure the normal execution of an outbound call process with the client by conducting dialogue interaction with the client based on multiple communication channels such as one outbound call communication channel and at least one private domain communication channel, thereby improving customer conversion rate. After an outbound call with the client is completed, it can provide users with multiple business operation channels, facilitate users to flexibly select business operation channels, and improve business revenue.
[0088] Please see Figure 2 , Figure 2 A schematic diagram of the structure of a large-scale model dialogue device based on multimedia provided in the second embodiment of the present application. The second embodiment of the present application provides a large-scale model dialogue device based on multimedia, comprising: a channel establishment module 201, for establishing an outbound communication channel with a client, and sending a channel establishment request message for a private domain communication channel to the client through the outbound communication channel; a message receiving module 202, for receiving the current round of client messages sent by the client through the outbound communication channel or the private domain communication channel when the private domain communication channel is successfully established; a message processing module 203, for generating a current conversation record based on the current round of client messages, a first historical conversation record based on the outbound communication channel, and a second historical conversation record based on the private domain communication channel; a message sending module 204, for inputting the current conversation record into a pre-established message generation model, and returning the obtained current round of local-end messages to the client.
[0089] In an optional embodiment, the channel establishment module 201 is further used to: before sending a channel establishment request message for the private domain communication channel to the client through the outbound communication channel, determine that a private domain communication channel has not been established with the client; send a channel establishment inquiry message for the private domain communication channel to the client through the outbound communication channel, and determine that a channel establishment consent message returned by the client is received.
[0090] In an optional embodiment, the current round client message is a multimodal message sent by the client through the private domain communication channel; the message processing module 203 is further used to determine the target modality message recognition model corresponding to the message modality according to the message modality of the current round client message before generating the current conversation record based on the current round client message, the first historical conversation record based on the outbound communication channel and the second historical conversation record based on the private domain communication channel; input the current round client message into the target modality message recognition model to obtain a recognition result; splice the current round client message and the recognition result to obtain the target current round client message; and update the current round client message to the target current round client message.
[0091] In an optional embodiment, the current conversation record is generated based on the current round of client messages, the first historical conversation record based on the outbound communication channel, and the second historical conversation record based on the private domain communication channel, including: splicing the current round of client messages, all messages in the first historical conversation record, and all messages in the second historical conversation record in order of message sending and receiving time from early to late to obtain the current conversation record.
[0092] In an optional embodiment, the message generation model includes a question extraction agent, an intent recognition agent, and a message generation agent; the current conversation record is input into a pre-established message generation model, and the current turn-end message is returned to the client, including: extracting at least one customer question from the current conversation record through a question extraction agent; identifying the customer intent based on at least one customer question through an intent recognition agent; determining an answer that meets the customer intent through a message generation agent, and generating the current turn-end message based on the answer.
[0093] In an optional embodiment, the device further includes: a dialogue switching module, which is used to trigger the target agent end to conduct a dialogue interaction with the client end through an outbound communication channel when the current turn-side message fails to pass the user review.
[0094] In an optional embodiment, the number of the outbound communication channel is one, the number of the outbound communication channel is one, and the number of the private domain communication channel is at least one.
[0095] The implementation process of the functions and effects of each module in the above-mentioned device is specifically described in the implementation process of the corresponding steps in the above-mentioned method, and will not be repeated here.
[0096] Please see Figure 3 , Figure 3 This is a schematic diagram of the structure of an electronic device provided in the third embodiment of the present application. The third embodiment of the present application provides an electronic device 30, comprising a processor 301, a memory 302, and a computer program stored in the memory 302 and configured to be executed by the processor 301; when the processor 301 executes the computer program, it implements the method described in the first embodiment of the present application and can achieve the same beneficial effects as described above.
[0097] In which, when the processor 301 reads the computer program from the memory 302 through the bus 303 and executes the computer program, it can implement the method of any embodiment included in the method described in the first embodiment of the present application.
[0098] Processor 301 can process digital signals and can include various computing architectures, such as a complex instruction set computer architecture, a reduced instruction set computer architecture, or an architecture that implements a combination of multiple instruction sets. In some examples, processor 301 can be a microprocessor.
[0099] The memory 302 can be used to store instructions executed by the processor 301 or data related to the execution of instructions. These instructions and / or data may include code for implementing some or all functions of one or more modules described in the embodiments of this application. The processor 301 of the embodiment of the present disclosure can be used to execute the instructions in the memory 302 to implement the method described in the first embodiment of this application. The memory 302 includes dynamic random access memory, static random access memory, flash memory, optical storage, or other memory known to those skilled in the art.
[0100] The fourth embodiment of the present application provides a computer-readable storage medium, which includes a stored computer program; wherein, when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the method described in the first embodiment of the present application, and can achieve the same beneficial effects as the method described in the first embodiment of the present application.
[0101] In summary, the embodiments of the present application provide a large-model dialogue method, apparatus, device and medium based on multimedia. The large-model dialogue method based on multimedia includes: establishing an outbound communication channel with the client, and sending a channel establishment request message for a private domain communication channel to the client through the outbound communication channel; when the private domain communication channel is successfully established, receiving the current round of client messages sent by the client through the outbound communication channel or the private domain communication channel; generating a current dialogue record based on the current round of client messages, a first historical dialogue record based on the outbound communication channel and a second historical dialogue record based on the private domain communication channel; inputting the current dialogue record into a pre-established message generation model, and returning the obtained current round of local-end messages to the client. The embodiment of the present application establishes an outbound communication channel with the client, sends a channel establishment request message for the private domain communication channel to the client through the outbound communication channel, and when the private domain communication channel is successfully established, receives the current round client message sent by the client through the outbound communication channel or the private domain communication channel, generates a current conversation record based on the current round client message, the first historical conversation record based on the outbound communication channel, and the second historical conversation record based on the private domain communication channel, inputs the current conversation record into a pre-established message generation model, and returns the obtained current round end message to the client. On the one hand, during an outbound call with the client, the message generation model can be used to combine the current conversation records on multiple communication channels of the outbound communication channel and the private domain communication channel to generate the current round end message, thereby ensuring the accurate generation of the current round end message and improving the customer conversion rate. On the other hand, after an outbound call with the client ends, the private domain communication channel can continue to interact with the client in conversation based on the private domain communication channel to carry out subsequent business operations, effectively saving business operation costs and improving business revenue.
[0102] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device embodiments described above are merely schematic. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architectures, functions and operations of the devices, methods and computer program products according to multiple embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of the code, and the module, program segment or a part of the code contains one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or action, or can be implemented using a combination of dedicated hardware and computer instructions.
[0103] In addition, the functional modules in each embodiment of the present application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.
[0104] If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a number of instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard drives, read-only memories (ROM), random access memories (RAM), magnetic disks or optical disks.
[0105] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any modifications or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.
Claims
1. A large-scale model dialogue method based on multimedia, characterized in that: include: Establishing an outbound communication channel with the client, and sending a channel establishment request message for a private domain communication channel to the client through the outbound communication channel; When the private domain communication channel is successfully established, receiving a current round of client messages sent by the client through the outbound communication channel or the private domain communication channel; Generate a current conversation record based on the current round of client messages, a first historical conversation record based on the outbound communication channel, and a second historical conversation record based on the private domain communication channel; The current conversation record is input into a pre-established message generation model, and the obtained current turn-side message is returned to the client.
2. The method according to claim 1, characterized in that Before sending a channel establishment request message for a private domain communication channel to the client through the outbound communication channel, the method further includes: Determining that the private domain communication channel is not established with the client; A channel establishment inquiry message for the private domain communication channel is sent to the client through the outbound communication channel, and it is determined that a channel establishment consent message returned by the client is received.
3. The method according to claim 1, characterized in that The current round of client messages is a multimodal message sent by the client through the private domain communication channel; Before generating the current conversation record according to the current round of client messages, the first historical conversation record based on the outbound communication channel, and the second historical conversation record based on the private domain communication channel, the method further includes: Determining a target modality message recognition model corresponding to the message modality according to the message modality of the current round of client messages; Inputting the current round of client messages into the target modal message recognition model to obtain a recognition result; Concatenate the current round client message and the recognition result to obtain a target current round client message; The current round client message is updated to the target current round client message.
4. The method according to claim 1, wherein Generating a current conversation record based on the current round of client messages, a first historical conversation record based on the outbound communication channel, and a second historical conversation record based on the private domain communication channel includes: The current round of client messages, all messages in the first historical conversation record, and all messages in the second historical conversation record are concatenated in order of message sending and receiving times from earliest to latest to obtain the current conversation record.
5. The method according to claim 1, wherein The message generation model includes a question extraction agent, an intention recognition agent and a message generation agent; The step of inputting the current conversation record into a pre-established message generation model and returning the obtained current turn-side message to the client includes: Extracting at least one customer question from the current conversation record by the question extraction agent; identifying the customer's intention based on the at least one customer question by the intention identification agent; An intelligent agent is generated through the message to determine an answer that meets the customer's intention, and the current turn-end message is generated based on the answer.
6. The method according to claim 1, characterized in that The method further comprises: In the case that the current round of the client's message fails to pass the user's review, the target agent terminal is triggered to interact with the client through the outbound communication channel.
7. The method according to any one of claims 1 to 6, characterized in that The number of the outbound communication channel is one, and the number of the private domain communication channel is at least one.
8. A large-scale model dialogue device based on multimedia, characterized in that: include: A channel establishment module is configured to establish an outbound communication channel with a client, and send a channel establishment request message for a private domain communication channel to the client via the outbound communication channel; a message receiving module, configured to receive a current round of client messages sent by the client through the outbound communication channel or the private communication channel when the private communication channel is successfully established; a message processing module, configured to generate a current conversation record based on the current round of client messages, a first historical conversation record based on the outbound communication channel, and a second historical conversation record based on the private domain communication channel; The message sending module is used to input the current conversation record into a pre-established message generation model and return the obtained current turn-side message to the client.
9. An electronic device, characterized in that: The method comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor; when the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.
10. A computer-readable storage medium, characterized in that The computer-readable storage medium includes a stored computer program; wherein, when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the method according to any one of claims 1 to 7.