Intelligent reply method, intelligent reply device, electronic equipment and storage medium

Abstract

The embodiment of the invention provides an intelligent reply method, an intelligent reply device, electronic equipment and a storage medium, and belongs to the technical field of artificial intelligence and financial science and technology. The method comprises the steps of obtaining original dialogue data of a target object in each round of dialogue; the target object comprises a first participating object and a second participating object; performing semantic understanding on the original dialogue data to obtain preliminary semantic information; performing intention recognition on the preliminary semantic information to obtain preliminary intention data of the target object; performing emotion recognition on the preliminary semantic information to obtain preliminary emotion data of the target object; acquiring preliminary reply content based on the preliminary intention data; obtaining a target virtual human image and a target action based on the preliminary emotion data; and controlling the target virtual human image to execute the target action and reply based on the preliminary reply content. Through the embodiment of the invention, the reply efficiency and accuracy can be improved.

Description

Technical Field

[0001] The present application relates to the fields of artificial intelligence and financial technology, and in particular to an intelligent reply method, an intelligent reply device, an electronic device, and a storage medium. Background Art

[0002] Smart replies are usually used in the field of smart customer service. Smart reply technology can automatically communicate with customers through preset rules and knowledge bases. For example, in insurance scenarios in the financial field, the main way of pre-sales consultation and after-sales maintenance is to answer manual customer service. In insurance sales scenarios in the financial field, it is mainly insurance agents who meet with customers through manual communication. In actual scenarios, manual customer service can usually only query relevant knowledge of the insurance industry for questions raised by users, and the service scope of manual customer service is limited and cannot provide users with comprehensive and accurate insurance services in a timely manner. Insurance agents are unable to provide comprehensive answer services in a timely manner when meeting with customers due to knowledge limitations. Summary of the invention

[0003] The main purpose of the embodiments of the present application is to propose an intelligent reply method, an intelligent reply device, an electronic device, and a storage medium, aiming to improve the efficiency and accuracy of the reply.

[0004] To achieve the above-mentioned purpose, a first aspect of an embodiment of the present application proposes an intelligent reply method, the method comprising:

[0005] Acquire original conversation data of the target object in each round of conversation; wherein the target object includes a first participant and a second participant, and the original conversation data includes at least one of the following: text data, voice data;

[0006] Performing semantic understanding on the original conversation data to obtain preliminary semantic information;

[0007] Performing intent recognition on the preliminary semantic information to obtain preliminary intent data of the target object;

[0008] Performing emotion recognition on the preliminary semantic information to obtain preliminary emotion data of the target object;

[0009] Obtaining preliminary response content based on the preliminary intent data;

[0010] Acquire a target virtual human image and a target action based on the preliminary emotion data;

[0011] The target virtual human image is controlled to execute the target action and to make a reply based on the preliminary reply content.

[0012] In some embodiments, the preliminary intention data includes customer intention data of the first participating object, the customer intention data includes a customer intention category, and performing intention recognition on the preliminary semantic information to obtain the preliminary intention data of the target object includes:

[0013] Classifying the preliminary semantic information to obtain first semantic information of the first participating object;

[0014] Performing vector conversion on the first semantic information to obtain a first semantic vector;

[0015] Determining an intent category probability value of the first semantic information according to the first semantic vector;

[0016] Obtaining the maximum value of the intention category probability values ​​to obtain a target intention probability value;

[0017] The customer intention category is filtered out from the preset intention data according to the target intention probability value.

[0018] In some embodiments, the original conversation data includes original speech data, and performing semantic understanding on the original conversation data to obtain preliminary semantic information includes:

[0019] Performing speech recognition on the original speech data to obtain preliminary text information;

[0020] Extracting keywords from the preliminary text information to obtain semantic keywords and keyword slots of the semantic keywords;

[0021] The preliminary semantic information is obtained by performing context understanding based on the semantic keywords and the keyword slots.

[0022] In some embodiments, the original conversation data further includes original text data, and before performing keyword extraction on the preliminary text information to obtain semantic keywords and keyword slots of the semantic keywords, performing semantic understanding on the original conversation data to obtain preliminary semantic information further includes:

[0023] Performing semantic recognition on the original text data to obtain semantic recognition information;

[0024] Acquire the timestamps of the semantic recognition information and the preliminary text information to obtain a semantic timestamp;

[0025] Merging the semantic recognition information and the preliminary text information based on the semantic timestamp to obtain semantic merged information;

[0026] The semantic merging information is updated to the preliminary text information.

[0027] In some embodiments, the customer intention data also includes customer objection information, and obtaining preliminary reply content based on the preliminary intention data includes:

[0028] Acquire customer information of the first participant to obtain customer portrait information of the first participant;

[0029] Classify the objection based on the customer objection information to obtain the current objection category;

[0030] Obtaining text encoding information based on the customer portrait information, the customer intention category and the current objection category;

[0031] Normalizing the text encoding information to obtain text normalization information;

[0032] Perform residual connection on the text normalization information to obtain the preliminary reply content.

[0033] In some embodiments, performing emotion recognition on the preliminary semantic information to obtain preliminary emotion data of the target object includes:

[0034] Performing voiceprint recognition on the original conversation data to obtain preliminary voiceprint features;

[0035] Emotion recognition is performed according to the preliminary semantic information and the preliminary voiceprint features to obtain the preliminary emotion data.

[0036] In some embodiments, controlling the target virtual human image to perform the target action and responding based on the preliminary response content includes:

[0037] Performing a reply response detection on the second participant to obtain a reply response result; wherein the reply response result includes a first response result and a second response result, the first response result is used to indicate that the second participant replies to the first participant within a first preset time period or that the second participant does not reply to the first participant within the first preset time period, and the second response result is used to indicate that a response indication from the second participant is received within a second preset time period or that the response indication is not received within the second preset time period;

[0038] If the reply result indicates that the second participant has not replied to the first participant within the first preset time period, the target virtual human image is controlled to perform the target action and reply based on the preliminary reply content; and / or, if the second response result indicates that the response indication has not been received within the second preset time period, the target virtual human image is controlled to perform the target action and reply based on the preliminary reply content.

[0039] To achieve the above-mentioned purpose, a second aspect of an embodiment of the present application provides an intelligent reply device, the device comprising:

[0040] A dialogue acquisition module, used to acquire original dialogue data of the target object in each round of dialogue; wherein the target object includes a first participant and a second participant, and the original dialogue data includes at least one of the following: text data, voice data;

[0041] A semantic processing module, used for performing semantic understanding on the original conversation data to obtain preliminary semantic information;

[0042] An intention recognition module, used to perform intention recognition on the preliminary semantic information to obtain preliminary intention data of the target object;

[0043] An emotion recognition module, used for performing emotion recognition on the preliminary semantic information to obtain preliminary emotion data of the target object;

[0044] A reply content generation module, used for obtaining preliminary reply content based on the preliminary intention data;

[0045] A virtual object generation module, used to obtain a target virtual human image and a target action based on the preliminary emotion data;

[0046] The reply module is used to control the target virtual human image to execute the target action and reply based on the preliminary reply content.

[0047] To achieve the above objectives, a third aspect of an embodiment of the present application proposes an electronic device, which includes a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the method described in the first aspect is implemented.

[0048] To achieve the above-mentioned purpose, the fourth aspect of an embodiment of the present application proposes a storage medium, which is a computer-readable storage medium, and the storage medium stores a computer program. When the computer program is executed by a processor, the method described in the first aspect is implemented.

[0049] The intelligent reply method, intelligent reply device, electronic device, and storage medium proposed in the embodiments of the present application obtain the original dialogue data of the target object in each round of dialogue, and perform semantic understanding on the original dialogue data to obtain preliminary semantic information, perform intent recognition on the preliminary semantic information to obtain preliminary intent data of the target object, perform emotion recognition on the preliminary semantic information to obtain preliminary emotion data of the target object, thereby obtaining preliminary reply content based on the preliminary intent data, and obtaining the target virtual human image and target action based on the preliminary emotion data, so as to control the target virtual human image to perform the target action and reply based on the preliminary reply content. The embodiments of the present application can improve the efficiency and accuracy of replies. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 is a flow chart of the intelligent reply method provided by an embodiment of the present application;

[0051] Figure 2 yes Figure 1 Step 102 is a flow chart of an embodiment;

[0052] Figure 3 yes Figure 1 Step 103 is a flowchart of another embodiment;

[0053] Figure 4 yes Figure 1 Flow chart of step 103 in;

[0054] Figure 5 yes Figure 1 Flowchart of step 105 in;

[0055] Figure 6 is a schematic diagram of the structure of the intelligent reply device provided in an embodiment of the present application;

[0056] Figure 7 It is a schematic diagram of the hardware structure of the electronic device provided in the embodiment of the present application. DETAILED DESCRIPTION

[0057] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0058] It should be noted that, although the functional modules are divided in the device schematic diagram and the logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first", "second", etc. in the specification, claims and the above drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0059] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0060] First, some nouns involved in this application are analyzed:

[0061] Artificial intelligence (AI) is a new technical science that studies and develops theories, methods, technologies and application systems for simulating, extending and expanding human intelligence. AI is a branch of computer science. AI attempts to understand the essence of intelligence and produce a new intelligent machine that can respond in a similar way to human intelligence. Research in this field includes robots, language recognition, image recognition, natural language processing and expert systems. AI can simulate the information process of human consciousness and thinking. AI is also a theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0062] Natural language processing (NLP): NLP uses computers to process, understand and apply human languages ​​(such as Chinese, English, etc.). NLP is a branch of artificial intelligence and an interdisciplinary subject between computer science and linguistics. It is often referred to as computational linguistics. Natural language processing includes grammatical analysis, semantic analysis, and text understanding. Natural language processing is often used in technical fields such as machine translation, handwritten and printed character recognition, speech recognition and text-to-speech conversion, information intent recognition, information extraction and filtering, text classification and clustering, public opinion analysis and opinion mining. It involves data mining related to language processing, machine learning, knowledge acquisition, knowledge engineering, artificial intelligence research, and linguistic research related to language computing.

[0063] ChatGPT (Chat Generative Pre-trained Transformer) is a chatbot program and a generative pre-trained transformation model. As a natural language processing tool driven by artificial intelligence technology, ChatGPT is based on the Transformer neural network architecture and can be used to process sequence data. ChatGPT has language understanding and text generation capabilities. ChatGPT can train models by connecting a large number of corpora that contain real-world conversations, allowing ChatGPT to interact with users based on the context of the chat. In actual scenarios, ChatGPT can handle tasks such as writing emails, video scripts, copywriting, translation, and code.

[0064] Smart replies are usually used in the field of smart customer service. Smart reply technology can automatically communicate with customers through preset rules and knowledge bases. For example, in insurance scenarios in the financial field, the main way of pre-sales consultation and after-sales maintenance is to answer manual customer service. In insurance sales scenarios in the financial field, it is mainly insurance agents who meet with customers through manual communication. In actual scenarios, manual customer service can usually only query relevant knowledge of the insurance industry for questions raised by users, and the service scope of manual customer service is limited and cannot provide users with comprehensive and accurate insurance services in a timely manner. Insurance agents are unable to provide comprehensive answer services in a timely manner when meeting with customers due to knowledge limitations.

[0065] Based on this, the embodiments of the present application provide an intelligent reply method, an intelligent reply device, an electronic device, and a storage medium, aiming to improve the efficiency and accuracy of the reply.

[0066] The intelligent reply method, intelligent reply device, electronic device, and storage medium provided in the embodiments of the present application are specifically illustrated through the following embodiments. First, the intelligent reply method in the embodiments of the present application is described.

[0067] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Among them, artificial intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0068] AI basic technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. AI software technologies mainly include computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0069] The intelligent reply method provided in the embodiment of the present application relates to the field of artificial intelligence technology. The intelligent reply method provided in the embodiment of the present application can be applied to the terminal, can also be applied to the server side, and can also be software running in the terminal or the server side. In some embodiments, the terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, etc.; the server side can be configured as an independent physical server, or it can be configured as a server cluster or distributed system composed of multiple physical servers, and can also be configured as a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms; the software can be an application that implements the intelligent reply method, etc., but is not limited to the above forms.

[0070] The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application can also be practiced in distributed computing environments, in which tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0071] It should be noted that in each specific implementation of the present application, when it comes to the need to perform relevant processing based on data such as user information, user voice, user behavior, user historical data, and user attribute information and other data related to user identity or characteristics, the user's permission or consent will be obtained first, and the collection, use and processing of these data will comply with relevant laws, regulations and standards. In addition, when the embodiment of the present application needs to obtain the user's sensitive personal information, the user's separate permission or consent will be obtained through a pop-up window or jump to a confirmation page. After clearly obtaining the user's separate permission or consent, the necessary user-related data for the normal operation of the embodiment of the present application will be obtained.

[0072] Figure 1 is an optional flowchart of the intelligent reply method provided in the embodiment of the present application. Figure 1The method may include but is not limited to steps 101 to 107.

[0073] Step 101, obtaining original conversation data of the target object in each round of conversation; wherein the target object includes a first participant and a second participant, and the original conversation data includes at least one of the following: text data, voice data;

[0074] Step 102, semantic understanding is performed on the original conversation data to obtain preliminary semantic information;

[0075] Step 103, performing intent recognition on the preliminary semantic information to obtain preliminary intent data of the target object;

[0076] Step 104, performing emotion recognition on the preliminary semantic information to obtain preliminary emotion data of the target object;

[0077] Step 105, obtaining preliminary reply content based on the preliminary intention data;

[0078] Step 106, obtaining a target virtual human image and a target action based on the preliminary emotion data;

[0079] Step 107 , controlling the target virtual human image to perform a target action and respond based on the preliminary response content.

[0080] Steps 101 to 107 shown in the embodiment of the present application are performed by obtaining the original conversation data of the target object in each round of conversation, performing semantic understanding on the original conversation data to obtain preliminary semantic information, performing intent recognition on the preliminary semantic information to obtain preliminary intent data of the target object, performing emotion recognition on the preliminary semantic information to obtain preliminary emotion data of the target object, and obtaining preliminary reply content based on the preliminary intention data, and obtaining a target virtual human image and a target action based on the preliminary emotion data, thereby controlling the target virtual human image to perform the target action and reply based on the preliminary reply content. The efficiency and accuracy of the reply can be improved through the embodiment of the present application.

[0081] In step 101 of some embodiments, the target object may include a first participant and at least a second participant. In one application scenario, the first participant may be a policyholder, a customer consultant, and the second participant may be an insurance agent, an insurance agent manager, etc. The original conversation data includes text data, and the original conversation data may also include voice data. The original conversation data may also include text data and voice data at the same time, that is, the target object may conduct a conversation by text input, or by text and voice input. In a financial scenario, the text data may be, for example, a comparison of key insurance information and product terms. In other application scenarios, the original conversation data may also include image data, that is, the target object may input images in addition to text and voice. The input method for the target object to input the original conversation data includes at least one of the following: written input, touch screen operation, remote control operation, etc.

[0082] See also Figure 2 In step 102 of some embodiments, the original conversation data includes original speech data, and the original conversation data is semantically understood to obtain preliminary semantic information, which may include but is not limited to:

[0083] Step 201, performing speech recognition on the original speech data to obtain preliminary text information;

[0084] Step 202, extracting keywords from the preliminary text information to obtain semantic keywords and keyword slots of the semantic keywords;

[0085] Step 203 , context understanding is performed based on semantic keywords and keyword slots to obtain preliminary semantic information.

[0086] In step 201 of some embodiments, the original voice data may be subjected to voice recognition based on the audio signal recognition (ASR) technology to obtain preliminary text information. The technical principle of obtaining preliminary text information by performing voice recognition on the original voice data is not limited in the embodiments of the present application. In an insurance sales scenario, the original voice data may be the voice content from the insurance agent, and the preliminary text information obtained after performing voice recognition on the voice content may be: "Hello, Mr. Wang, regarding the insurance product A you are consulting, I would like our intelligent robot to answer your questions." The original voice data may also be the voice content from the insurance consultant, and the preliminary text information obtained after performing voice recognition on the voice content may be: "Regarding the insurance product A you just mentioned, you mentioned that there are corresponding purchase restrictions. Please let me know the specific conditions. Thank you."

[0087] In step 202 of some embodiments, for example, the preliminary text information includes: "I want to purchase insurance product A". After keyword extraction is performed on the preliminary text information, it is determined that the insurance category that the insured wants to purchase is "insurance product A", wherein "insurance product A" is a keyword, and the attribute of the keyword "insurance product A" is a keyword slot, and the keyword slot of "insurance product A" can be expressed as insurance.

[0088] In step 203 of some embodiments, a slot-value pair can be obtained based on the semantic keyword and the keyword slot. For example, in combination with the embodiment of step 202, if the text information includes: "I want to buy insurance product A", the slot-value pair is obtained based on the keyword "insurance product A" and the keyword slot "insurance": [insurance product A] insurance. Multiple groups of slot-value pairs can be obtained through step 203, and context understanding can be performed based on the multiple groups of slot-value pairs, thereby obtaining preliminary semantic information.

[0089] In the embodiment of the present application, the voice data is converted into text information through steps 201 to 203, and keywords are extracted from the text information to perform context understanding to obtain preliminary semantic information, facilitate subsequent intent recognition, and improve the accuracy of intent recognition.

[0090] In some embodiments, see Figure 3 The original conversation data also includes original text data. Before step 202 extracts keywords from the preliminary text information to obtain semantic keywords and keyword slots of semantic keywords, step 102 performs semantic understanding on the original conversation data to obtain preliminary semantic information, which may also include but is not limited to:

[0091] Step 301, performing semantic recognition on the original text data to obtain semantic recognition information;

[0092] Step 302, obtaining the timestamps of the semantic recognition information and the preliminary text information to obtain a semantic timestamp;

[0093] Step 303, merging the semantic recognition information and the preliminary text information based on the semantic timestamp to obtain semantic merged information;

[0094] Step 304: Update the semantic merge information to preliminary text information.

[0095] In step 301 of some embodiments, the original text data may be input from the target object, such as insurance knowledge query information, insurance purchase consultation information, or insurance underwriting request information. Semantic recognition information is obtained by performing semantic recognition on the original text data.

[0096] In step 302 of some embodiments, the semantic timestamp includes a first timestamp and a second timestamp. The first timestamp is obtained by acquiring the timestamp of the semantic recognition information, and the second timestamp is obtained by acquiring the timestamp of the preliminary text information.

[0097] In step 303 of some embodiments, the semantic recognition information and the preliminary text information are merged based on the semantic timestamp, and the obtained semantic merged information is information based on time sorting.

[0098] In step 304 of some embodiments, the semantic merge information replaces the preliminary text information to update the preliminary text information. In the insurance sales scenario, the intelligent reply method of the embodiment of the present application is applied to an intelligent reply system (the system can be an intelligent reply robot), which can receive the user's voice input, text input, and image input. The intelligent reply system merges the input data based on the timestamp input by the user to achieve data merging of multimodal inputs, so that the customer's intention and emotion recognition can be performed based on the context, and the accuracy of intention recognition and emotion recognition can be improved, so as to output reply content that meets the user's needs to the user.

[0099] In the embodiment of the present application, through steps 301 to 304, multimodal data processing of text data and voice data can be implemented, which facilitates subsequent intention recognition and emotion recognition and improves the diversity and experience of interaction with users.

[0100] See also Figure 4 In some embodiments, the preliminary intention data includes customer intention data of the first participant, the customer intention data includes customer intention categories, and step 103 may include but is not limited to:

[0101] Step 401, classifying the preliminary semantic information to obtain first semantic information of the first participant;

[0102] Step 402, performing vector conversion on the first semantic information to obtain a first semantic vector;

[0103] Step 403, determining the intent category probability value of the first semantic information according to the first semantic vector;

[0104] Step 404, obtaining the maximum value of the intention category probability values ​​to obtain the target intention probability value;

[0105] Step 405 , filtering out customer intent categories from the preset intent data according to the target intent probability value.

[0106] In actual application scenarios, the original conversation data may be the conversation content of at least two interlocutors, for example, it may be a conversation between two interlocutors or a conversation between three interlocutors. The preliminary semantic information obtained by semantically understanding the original conversation data through steps 201 to 203 may be the semantic information of at least one speaker. In order to identify the intention of each speaker, the preliminary semantic information is first classified through step 401 to obtain the semantic information of each speaker, for example, the first semantic information of the first participant may be obtained.

[0107] In step 402 of some embodiments, the first semantic information may be vectorized based on a pre-trained language model to obtain a first semantic vector. In one application scenario, the pre-trained language model may be a natural language processing (NLP) model, and the first semantic information may be semantically recognized and vectorized by the NLP model to obtain a first semantic vector corresponding to the first semantic information. In the embodiment of the present application, the NLP model may be semantically recognized and characterized with a small number of sample sentences and annotations, thereby improving the accuracy of semantic recognition and further improving the accuracy of subsequent intent recognition.

[0108] In step 403 of some embodiments, the process of intent recognition can be regarded as a text-based multi-classification task, and the probability value of the first semantic information belonging to each intent category is determined in multiple pre-set intent categories (preset intent data) through the fully connected layer hidden layer, so as to determine the customer intent data of the first semantic information in multiple intent categories, that is, the final intent category. Specifically, in an application scenario, the multi-classification output of the fully connected layer hidden layer is normalized using the softmax function to obtain the probability value of the first semantic information belonging to each intent category.

[0109] In step 404 and step 405 of some embodiments, the maximum value of the intention category probability value is used as the target intention probability value, and the intention category corresponding to the target intention probability value is used as the customer intention data. The preset intention data can be preset according to the actual situation. In the insurance scenario, the preset intention data can be used to indicate the willingness to purchase insurance. For example, it can include five intention categories such as very willing, willing, general, not very willing, and completely unwilling. Taking the purchase of insurance product A by a user as an example, it is determined that the intention category probability value of the intention category of purchasing insurance product A is [very willing] is 0.5, the intention category probability value of the intention category of purchasing insurance product A is [willing] is 0.2, the intention category probability value of purchasing insurance product A is [general] is 0.1, the intention category probability value of purchasing insurance product A is [not very willing] is 0.1, and the intention category probability value of purchasing insurance product A is [completely unwilling] is 0.1. The intention category probability value of the intention category of purchasing insurance product A is [very willing] is 0.5, which is the maximum value, thereby determining that the intention category of the user to purchase insurance product A is [very willing].

[0110] In the embodiment of the present application, through steps 401 to 405, the first semantic information of the first participant (e.g., a customer) can be screened out from a multi-person conversation, and the first semantic information can be vector processed to determine the intention of the first participant based on the first semantic vector. Through the embodiment of the present application, the accuracy of intent recognition can be improved.

[0111] In step 104 of some embodiments, emotion recognition is performed on the preliminary semantic information to obtain preliminary emotion data of the target object, which may include but is not limited to:

[0112] Perform voiceprint recognition on the original conversation data to obtain preliminary voiceprint features;

[0113] Emotion recognition is performed based on preliminary semantic information and preliminary voiceprint features to obtain preliminary emotion data.

[0114] In actual application scenarios, the target object can communicate through voice and text input at the same time, and the original conversation data can be subjected to voiceprint recognition to obtain preliminary voiceprint features. The voiceprint features such as the tone and intonation of the target object can be obtained through the preliminary voiceprint features, and then combined with the specific conversation content of the target object (i.e., preliminary semantic information), the target object can be emotionally recognized to obtain preliminary emotional data of the target object. In the embodiment of the present application, the voiceprint features of the target object and the preliminary semantic information of the specific conversation are combined to improve the accuracy of emotion recognition. For example, in the insurance scenario, the voiceprint features in the original conversation data of the target object characterize that the target object has a relatively loud volume (higher than the preset reference volume) and a relatively fast speech speed (higher than the preset reference speech speed), and the preliminary semantic information contains words with a heavier tone, such as "Why can't a 60-year-old man buy insurance A?" contains the rhetorical question "why", thereby identifying that the current emotion of the target object includes anger and impatience.

[0115] In some embodiments, the customer intent data also includes customer objection information, see Figure 5 Step 105 obtains preliminary response content based on the preliminary intention data, which may include but is not limited to:

[0116] Step 501, obtaining customer information of a first participant, and obtaining customer portrait information of the first participant;

[0117] Step 502, classifying the objection based on the customer objection information to obtain the current objection category;

[0118] Step 503, obtaining text encoding information based on the customer portrait information, the customer intention category and the current objection category;

[0119] Step 504, normalizing the text encoding information to obtain text normalization information;

[0120] Step 505, perform residual connection on the text normalization information to obtain preliminary response content.

[0121] In step 501 of some embodiments, customer portrait information of the first participant can be obtained based on relevant information input by the first participant. For example, in a scenario where a user consults about insurance underwriting, corresponding customer portrait information can be obtained based on the policy number input by the user, or customer portrait information related to the identity information input by the user can be queried from a database. The customer portrait information includes user identity portrait, user preference portrait, user purchasing power portrait, etc.

[0122] In step 502 of some embodiments, the customer objection information may be: "The insurance package you just mentioned does not seem to cover all important medical services." After objection classification based on the customer objection information, the current objection category is obtained as: insurance product coverage.

[0123] In step 503 of some embodiments, the customer profile information, customer intent category, and current objection category may be uniformly encoded based on a preset encoder to obtain text encoding information, thereby facilitating normalization processing in step 504 to obtain text normalization information. The preset encoder may be a pre-trained Transformer encoder, which has the ability to understand context and capture the dependencies between elements in the text, thereby achieving deep semantic understanding and contextual relationship analysis of the overall input information, and outputting a matrix represented by numerical values ​​to obtain text encoding information.

[0124] In step 504 of some embodiments, the text encoding information is normalized so that the text normalization information has stronger generalization ability.

[0125] In step 505 of some embodiments, residual connections may be performed on the text normalization information based on the residual block to obtain preliminary reply content. In this embodiment, the residual block may include multiple linear connection layers, each of which has a jump connection, so that multiple linear connection layers can learn the potential information in the text normalization information, further improving the generalization ability, so as to generate higher quality and more accurate preliminary reply content when processing customer objection information.

[0126] In steps 501 to 505 of the embodiment of the present application, customer objection information is processed in combination with customer portrait information and customer intent category, so that preliminary response content that is more in line with the user's current psychological and actual needs can be generated based on the user's personal preferences and current intent category, thereby improving the user experience.

[0127] In actual application scenarios, massive data can be obtained to obtain customer portraits and train insurance knowledge models, where the insurance knowledge model can be a ChatGPT model. The massive data includes a large amount of conversation data between insurance agents and customers, relevant course data sold by insurance agents, a large amount of policy data and customer information, insurance-related professional books, etc., so as to construct a massive amount of insurance-related knowledge, and train the insurance knowledge model based on a large number of cases and research reports closed by insurance agents, so that the customer portrait can be matched with the insurance knowledge model, and applied to the intelligent AI auxiliary system to assist insurance agents in conducting business through the personal information provided by the customer, and search and analyze the insurance preferences, possible risks and matching products and services that meet the customer in the insurance knowledge model.

[0128] Taking the actual application scenario of insurance agents meeting with customers as an example, the embodiments of the present application can be applied to intelligent AI-assisted systems, which can be displayed in the form of digital people and communicated through a voice interaction system. After the first participant provides personal information, key insurance information and product terms comparison are displayed by analyzing customer needs. When the first participant raises questions and objections, the embodiments of the present application can generate preliminary response content to provide answers and suggestions for handling derived needs. As the second participant, the insurance agent can quickly answer customer questions, provide comprehensive insurance solutions, and demonstrate professional sales capabilities through the intelligent AI-assisted system.

[0129] Through the embodiment of the present application, the second participant can provide comprehensive insurance sales support when meeting the first participant with the help of the intelligent AI auxiliary system. The embodiment of the present application is presented in a multimodal interactive form, and combined with the ChatGPT model to analyze customer needs and provide insurance knowledge. Compared with traditional sales scenarios, the embodiment of the present application can improve sales efficiency, personalized services, support professional sales, and improve customer satisfaction.

[0130] In step 106 of some embodiments, a corresponding target virtual human image and target action are determined based on the preliminary emotional data. For example, if the preliminary emotional data indicates that the customer is angry, a target action of making a heart shape with hands is obtained, and a cute and funny image is obtained as the target virtual human image, such as a funny cartoon image or an avatar that the customer likes.

[0131] In some embodiments, step 107 may include, but is not limited to:

[0132] Performing a reply response detection on the second participant to obtain a reply response result; wherein the reply response result includes a first response result and a second response result, the first response result is used to indicate that the second participant has replied to the first participant within a first preset time period or that the second participant has not replied to the first participant within the first preset time period, and the second response result is used to indicate that a response indication from the second participant has been received within a second preset time period or that a response indication has not been received within the second preset time period;

[0133] If the reply response result indicates that the second participant has not replied to the first participant within the first preset time period, the target virtual human image is controlled to perform the target action and reply based on the preliminary reply content; and / or, if the second response result indicates that the response indication has not been received within the second preset time period, the target virtual human image is controlled to perform the target action and reply based on the preliminary reply content.

[0134] In an actual application scenario, the first preset time length and the second preset time length may be set to be the same. In other application scenarios, the first preset time length may be greater than the second preset time length or the first preset time length may be less than the second preset time length.

[0135] In an application scenario in the financial field, the first participant is a customer and the second participant is an insurance agent. During the conversation in which the insurance agent explains insurance to the customer, if it is detected that the insurance agent has not responded to the customer within 3 seconds (the first preset time length), and / or the response instruction issued by the insurance agent within 3 seconds (the second preset time length) is not received, the intelligent AI assisted system will respond to the customer on its own. Taking the scenario of responding to the customer's bidding consultation as an example, the cartoon image is controlled to speak the preliminary reply content in a voice with a heart gesture: Beautiful lady, you can purchase insurance A in the following ways.

[0136] In an application scenario where an insurance agent meets with a customer, the embodiment of the present application can provide detailed insurance information to the customer, answer the customer's questions and handle the customer's needs. Due to time and knowledge constraints, agents are unable to reply and respond to customers when faced with a large amount of customer information and questions. Compared with traditional sales assistance tools that usually only provide static information and limited interactive capabilities, the embodiment of the present application can provide multi-modal interaction methods such as digital humans, robots and voice interactions.

[0137] The intelligent reply method provided in the embodiment of the present application is based on the ChatGPT model and can be used in an intelligent AI-assisted system, thereby providing comprehensive auxiliary support for agents (such as insurance agents), providing comprehensive customer information, key insurance information, objection questions and answers, derivative needs and product terms comparison, etc. when meeting with customers. In the form of multimodal interaction, such as digital humans, robots and voice interaction systems, intelligent all-round sales assistance is achieved, the professional capabilities of agents are demonstrated, and sales efficiency and customer satisfaction are improved.

[0138] The intelligent reply method of the embodiment of the present application can be applied to an intelligent AI-assisted system, which can be displayed in a multimodal form, such as a digital human, a robot, and other multimodal forms. For example, in a face-to-face meeting between an insurance agent and a customer, the insurance agent can communicate with the customer through a digital human or a robot. The intelligent AI-assisted system pre-stores comprehensive customer information. The intelligent AI-assisted system can provide comprehensive customer information to the insurance agent by analyzing the customer's personal information, family situation, occupation, etc., wherein the customer information may include but is not limited to the customer's preferences, needs, risk tolerance, etc., to help the agent better understand the customer and provide corresponding insurance solutions. The intelligent AI-assisted system can also provide key insurance information and product terms comparison, risk gaps and protection gaps, etc. according to the customer's needs and circumstances. The insurance agent can accurately explain to the customer the characteristics, insured amount, premium, claims process, etc. of different insurance products, as well as the comparison and advantages and disadvantages between different products. The intelligent AI-assisted system can understand the objections, questions and derivative needs raised by the customer, and provide corresponding answers and suggestions. Insurance agents can use the intelligent AI assistance of the system to better answer customers' questions and handle customer needs, and provide professional consulting services.

[0139] The digital human in the embodiment of the present application refers to a virtual character display in a humanoid image, presented through images and animations. The digital human image is more likely to resonate with the younger generation of customers and attract the attention of the younger generation of customers. In addition, the digital human can convey emotions and provide a more vivid interactive experience through facial expressions, gestures and animations. Insurance agents can obtain more detailed insurance information, product comparisons and other related content through a touch screen or by clicking on the body parts of the digital human. The robot in the embodiment of the present application refers to a physical robot with autonomous movement and perception capabilities that can interact with customers in a physical space. The robot can display insurance product samples, provide physical interactions and demonstrations on site, and increase customer participation. In addition, the robot communicates directly with customers through voice and body language to create a more intimate and human experience. Insurance agents can control the robot's movements and demonstrations through remote control or voice commands to demonstrate the characteristics and advantages of insurance products.

[0140] In addition, for customers who want convenient and fast interaction, the embodiments of the present application can achieve fast interaction through voice recognition and speech synthesis technology without manual operation or text input. And, for customers who want both convenience and personalized service, the embodiments of the present application can understand the customer's verbal input through voice recognition technology and provide personalized insurance solutions according to customer needs.

[0141] The embodiments of the present application can provide insurance agents with optional interaction forms, including digital humans, robots, and voice interaction systems, based on the customer's age, preferences, emotional needs, and characteristics of the interaction scenario.

[0142] See also Figure 6 The embodiment of the present application also provides an intelligent reply device, which can implement the above-mentioned intelligent reply method, and the device includes:

[0143] A dialogue acquisition module, used to acquire original dialogue data of the target object in each round of dialogue; wherein the target object includes a first participant and a second participant, and the original dialogue data includes at least one of the following: text data, voice data;

[0144] The semantic processing module is used to understand the original conversation data and obtain preliminary semantic information;

[0145] An intention recognition module is used to perform intention recognition on the preliminary semantic information to obtain preliminary intention data of the target object;

[0146] An emotion recognition module is used to perform emotion recognition on the preliminary semantic information to obtain preliminary emotion data of the target object;

[0147] A reply content generation module, used for obtaining preliminary reply content based on preliminary intention data;

[0148] A virtual object generation module, used to obtain a target virtual human image and target action based on preliminary emotion data;

[0149] The reply module is used to control the target virtual human image to perform the target action and reply based on the preliminary reply content.

[0150] In some embodiments, the original conversation data includes original voice data, and the semantic processing module can be specifically used to implement:

[0151] Perform speech recognition on the original speech data to obtain preliminary text information;

[0152] Perform keyword extraction on the preliminary text information to obtain semantic keywords and keyword slots of semantic keywords;

[0153] Based on semantic keywords and keyword slots, context understanding is performed to obtain preliminary semantic information.

[0154] Specifically, the semantic processing module can be used to implement the above steps 201 to 203, which will not be described in detail here.

[0155] In some embodiments, the original conversation data also includes original text data, and the semantic processing module can be specifically used to implement:

[0156] Perform semantic recognition on the original text data to obtain semantic recognition information;

[0157] Obtaining timestamps of semantic recognition information and preliminary text information to obtain a semantic timestamp;

[0158] Merging the semantic recognition information and the preliminary text information based on the semantic timestamp to obtain semantic merged information;

[0159] Update the semantic merge information to the preliminary text information.

[0160] Specifically, the semantic processing module can also be used to implement the above steps 301 to 304, which will not be described in detail here.

[0161] See also Figure 4 In some embodiments, the preliminary intention data includes customer intention data of the first participant, the customer intention data includes customer intention categories, and the intention identification module can also be used to implement:

[0162] Classifying the preliminary semantic information to obtain first semantic information of the first participating object;

[0163] Performing vector conversion on the first semantic information to obtain a first semantic vector;

[0164] Determine an intent category probability value of the first semantic information according to the first semantic vector;

[0165] Get the maximum value of the intention category probability value to get the target intention probability value;

[0166] Filter customer intent categories from preset intent data based on target intent probability values.

[0167] Specifically, the intention recognition module can be used to implement the above steps 401 to 405, which will not be repeated here.

[0168] In some embodiments, the customer intention data also includes customer objection information, and the reply content generation module can also be used to implement:

[0169] Acquire customer information of the first participant and obtain customer portrait information of the first participant;

[0170] Classify objections based on customer objection information to obtain the current objection category;

[0171] Get text coding information based on customer profile information, customer intention category and current objection category;

[0172] Normalizing the text encoding information to obtain text normalization information;

[0173] Perform residual connection on the normalized information of the text to obtain the preliminary response content.

[0174] Specifically, the reply content generation module can be used to implement the above steps 501 to 505, which will not be described in detail here.

[0175] The specific implementation of the intelligent reply device is basically the same as the specific implementation of the above-mentioned intelligent reply method, and will not be repeated here.

[0176] The embodiment of the present application also provides an electronic device, the electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the above-mentioned intelligent reply method when executing the computer program. The electronic device can be any intelligent terminal including a tablet computer, a car computer, etc.

[0177] See also Figure 7 , Figure 7 The hardware structure of an electronic device of another embodiment is illustrated, and the electronic device includes:

[0178] The processor 701 may be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (Application Specific Integrated Circuit, ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present application;

[0179] The memory 702 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 702 can store an operating system and other applications. When the technical solution provided in the embodiment of this specification is implemented by software or firmware, the relevant program code is stored in the memory 702, and the processor 701 calls and executes the intelligent reply method of the embodiment of this application;

[0180] Input / output interface 703, used to implement information input and output;

[0181] Communication interface 704, used to realize communication interaction between the device and other devices, which can be realized through wired mode (such as USB, network cable, etc.) or wireless mode (such as mobile network, WIFI, Bluetooth, etc.);

[0182] A bus 705 that transmits information between the various components of the device (e.g., the processor 701, the memory 702, the input / output interface 703, and the communication interface 704);

[0183] The processor 701 , the memory 702 , the input / output interface 703 and the communication interface 704 are connected to each other in communication within the device via a bus 705 .

[0184] An embodiment of the present application also provides a storage medium, which is a computer-readable storage medium and stores a computer program. When the computer program is executed by a processor, the above-mentioned intelligent reply method is implemented.

[0185] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely disposed relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0186] The intelligent reply method, intelligent reply device, electronic device, and storage medium provided in the embodiments of the present application obtain the original dialogue data of the target object in each round of dialogue, perform semantic understanding on the original dialogue data to obtain preliminary semantic information, perform intent recognition on the preliminary semantic information to obtain preliminary intent data of the target object, perform emotion recognition on the preliminary semantic information to obtain preliminary emotion data of the target object, and obtain preliminary reply content based on the preliminary intent data, obtain the target virtual human image and target action based on the preliminary emotion data, thereby controlling the target virtual human image to perform the target action and reply based on the preliminary reply content to improve the efficiency and accuracy of the reply. The embodiments of the present application can generate preliminary reply content that is more in line with the user's current psychological and actual needs based on the user's personal preferences and current intent categories, thereby improving the user experience. In addition, the embodiments of the present application are presented in a multimodal interactive form, and are combined with the ChatGPT model to analyze customer needs and provide insurance knowledge. Compared with traditional sales scenarios, the embodiments of the present application can improve sales efficiency, personalized services, support professional sales, and improve customer satisfaction.

[0187] The embodiments described in the embodiments of the present application are intended to more clearly illustrate the technical solutions of the embodiments of the present application and do not constitute a limitation on the technical solutions provided in the embodiments of the present application. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0188] Those skilled in the art will appreciate that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than shown in the figures, or a combination of certain steps, or different steps.

[0189] The device embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0190] Those skilled in the art will appreciate that all or some of the steps in the methods disclosed above, and the functional modules / units in the systems and devices may be implemented as software, firmware, hardware, or a suitable combination thereof.

[0191] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0192] It should be understood that in the present application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the objects associated before and after are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0193] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the above units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0194] The units described above 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 may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0195] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0196] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it 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 all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including multiple instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, referred to as ROM), random access memory (Random Access Memory, referred to as RAM), disk or optical disk and other media that can store programs.

[0197] The preferred embodiments of the present invention are described above with reference to the accompanying drawings, but the scope of the rights of the present invention is not limited thereto. Any modification, equivalent substitution and improvement made by a person skilled in the art without departing from the scope and essence of the present invention should be within the scope of the rights of the present invention.

Claims

1. An intelligent reply method, characterized in that: The method comprises: Acquire original conversation data of the target object in each round of conversation; wherein the target object includes a first participant and a second participant, and the original conversation data includes at least one of the following: text data, voice data; Performing semantic understanding on the original conversation data to obtain preliminary semantic information; Performing intent recognition on the preliminary semantic information to obtain preliminary intent data of the target object; Performing emotion recognition on the preliminary semantic information to obtain preliminary emotion data of the target object; Obtaining preliminary response content based on the preliminary intent data; Acquire a target virtual human image and a target action based on the preliminary emotion data; The target virtual human image is controlled to execute the target action and to make a reply based on the preliminary reply content.

2. The method according to claim 1, characterized in that The preliminary intention data includes customer intention data of the first participant object, the customer intention data includes customer intention categories, and the preliminary intention data of the target object obtained by performing intention recognition on the preliminary semantic information includes: Classifying the preliminary semantic information to obtain first semantic information of the first participating object; Performing vector conversion on the first semantic information to obtain a first semantic vector; Determining an intent category probability value of the first semantic information according to the first semantic vector; Obtaining the maximum value of the intention category probability values ​​to obtain a target intention probability value; The customer intention category is filtered out from the preset intention data according to the target intention probability value.

3. The method according to claim 2, characterized in that The original conversation data includes original speech data, and the performing semantic understanding on the original conversation data to obtain preliminary semantic information includes: Performing speech recognition on the original speech data to obtain preliminary text information; Extracting keywords from the preliminary text information to obtain semantic keywords and keyword slots of the semantic keywords; The preliminary semantic information is obtained by performing context understanding based on the semantic keywords and the keyword slots.

4. The method according to claim 3, characterized in that: The original conversation data also includes original text data. Before extracting keywords from the preliminary text information to obtain semantic keywords and keyword slots of the semantic keywords, performing semantic understanding on the original conversation data to obtain preliminary semantic information further includes: Performing semantic recognition on the original text data to obtain semantic recognition information; Acquire the timestamps of the semantic recognition information and the preliminary text information to obtain a semantic timestamp; Merging the semantic recognition information and the preliminary text information based on the semantic timestamp to obtain semantic merged information; The semantic merging information is updated to the preliminary text information.

5. The method according to claim 4, characterized in that The customer intention data also includes customer objection information, and obtaining preliminary reply content based on the preliminary intention data includes: Acquire customer information of the first participant to obtain customer portrait information of the first participant; Classify the objection based on the customer objection information to obtain the current objection category; Obtaining text encoding information based on the customer portrait information, the customer intention category and the current objection category; Normalizing the text encoding information to obtain text normalization information; Perform residual connection on the text normalization information to obtain the preliminary reply content.

6. The method according to claim 3, characterized in that: The performing emotion recognition on the preliminary semantic information to obtain preliminary emotion data of the target object includes: Performing voiceprint recognition on the original conversation data to obtain preliminary voiceprint features; Emotion recognition is performed according to the preliminary semantic information and the preliminary voiceprint features to obtain the preliminary emotion data.

7. The method according to any one of claims 1 to 6, characterized in that: The controlling the target virtual human image to perform the target action and respond based on the preliminary response content includes: Performing a reply response detection on the second participant to obtain a reply response result; wherein the reply response result includes a first response result and a second response result, the first response result is used to indicate that the second participant replies to the first participant within a first preset time period or that the second participant does not reply to the first participant within the first preset time period, and the second response result is used to indicate that a response indication from the second participant is received within a second preset time period or that the response indication is not received within the second preset time period; If the reply result indicates that the second participant has not replied to the first participant within the first preset time period, the target virtual human image is controlled to perform the target action and reply based on the preliminary reply content; and / or, if the second response result indicates that the response indication has not been received within the second preset time period, the target virtual human image is controlled to perform the target action and reply based on the preliminary reply content.

8. An intelligent answering device, characterized in that: The device comprises: A dialogue acquisition module, used to acquire original dialogue data of the target object in each round of dialogue; wherein the target object includes a first participant and a second participant, and the original dialogue data includes at least one of the following: text data, voice data; A semantic processing module, used for performing semantic understanding on the original conversation data to obtain preliminary semantic information; An intention recognition module, used to perform intention recognition on the preliminary semantic information to obtain preliminary intention data of the target object; An emotion recognition module, used for performing emotion recognition on the preliminary semantic information to obtain preliminary emotion data of the target object; A reply content generation module, used for obtaining preliminary reply content based on the preliminary intention data; A virtual object generation module, used to obtain a target virtual human image and a target action based on the preliminary emotion data; The reply module is used to control the target virtual human image to execute the target action and reply based on the preliminary reply content.

9. An electronic device, characterized in that: The electronic device comprises a memory and a processor, the memory stores a computer program, and the processor implements the method according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images