Intelligent dialogue method and device, equipment, storage medium and program product

Abstract

The invention discloses an intelligent dialogue method and device, equipment, a storage medium and a program product, and relates to the field of human-computer interaction. The method comprises the following steps: displaying a dialog box interface; receiving a text input operation in the dialog box interface, wherein the text input operation is used for obtaining a first question text; based on the text input operation, displaying a first solution text for the first question text, and displaying at least one piece of text associated information, the text associated information comprising at least one of reply content of the first solution text and candidate question texts having semantic association relationship with the first question text; the text association information is used for providing shortcut operation elements for dialogue interaction with the system intelligent account for the first account. In this way, the text associated information serves as a shortcut operation element for dialogue interaction, and the efficiency of continuing intelligent dialogues subsequently is greatly improved. The method can be applied to various scenes such as cloud technology, artificial intelligence and intelligent transportation.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of human-computer interaction, and in particular to an intelligent dialogue method, device, equipment, storage medium and program product. Background Art

[0002] With the development of computer technology, intelligent dialogue technology has received widespread attention. Users can communicate with the system's intelligent account to know the answers to their questions in a more targeted manner, thereby improving the accuracy of information acquisition.

[0003] In related technologies, considering that the accuracy of question content has a great influence on the generation of answer content, intelligent dialogue technology usually focuses on improving the accuracy of question content. For example, based on the first half of the text content input by the user, multiple second half text contents are predicted for the user to choose from, thereby determining the question text in the intelligent dialogue process based on the user's selection operation.

[0004] Although the above process can improve the accuracy of the question content based on the partial text input by the user, the interaction of the intelligent dialogue is relatively simple, and usually only the corresponding answer text can be generated based on a question text once, which makes the manual interaction efficiency in the intelligent dialogue scenario low and affects the user experience. Summary of the invention

[0005] The embodiments of the present application provide an intelligent dialogue method, device, equipment, storage medium and program product, which can avoid the rigid problem of only replying to the current first question text with a single answer text, and use text-related information as a quick operation element for dialogue interaction, greatly improving the efficiency of subsequent intelligent dialogue. The technical solution is as follows.

[0006] In one aspect, an intelligent dialogue method is provided, the method comprising:

[0007] Displaying a dialog interface, wherein a first account is logged in to the dialog interface, and the first account is used to interact with the system smart account through the dialog interface;

[0008] Receiving a text input operation in the dialog interface, wherein the text input operation is used to obtain a first question text;

[0009] Based on the text input operation, display the first answer text for the first question text, and display at least one piece of text association information, where the text association information includes at least one of the reply content to the first answer text and the candidate question text that has a semantic association relationship with the first question text, and the text association information is used to provide the first account with a quick operation element for interacting with the system intelligent account in a dialogue.

[0010] On the other hand, a smart dialogue device is provided, and the device includes:

[0011] An interface display module, configured to display a dialogue box interface, where the first account is logged in to the dialogue box interface, and the first account is used to interact with the system intelligent account through the dialogue box interface;

[0012] An operation receiving module, configured to receive a text input operation in the dialogue box interface, and the text input operation is used to obtain a first question text;

[0013] A text display module, configured to display the first answer text for the first question text based on the text input operation, and display at least one piece of text association information, where the text association information includes at least one of the reply content to the first answer text and the candidate question text that has a semantic association relationship with the first question text, and the text association information is used to provide the first account with a quick operation element for interacting with the system intelligent account in a dialogue.

[0014] On the other hand, a computer device is provided, and the computer device includes a processor and a memory. At least one instruction, at least one program, a code set, or an instruction set is stored in the memory, and the at least one instruction, the at least one program, the code set, or the instruction set is loaded and executed by the processor to implement the smart dialogue method as described in any one of the above embodiments of the present application.

[0015] On the other hand, a computer-readable storage medium is provided. At least one instruction, at least one program, a code set, or an instruction set is stored in the storage medium, and the at least one instruction, the at least one program, the code set, or the instruction set is loaded and executed by a processor to implement the smart dialogue method as described in any one of the above embodiments of the present application.

[0016] On the other hand, a computer program product or a computer program is provided. The computer program product or the computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the smart dialogue method as described in any one of the above embodiments.

[0017] The beneficial effects brought by the technical solution provided by the embodiment of the present application at least include:

[0018] When the first account conducts a dialogue interaction with the system intelligent account through the dialog box interface, the first question text is obtained according to the text input in the dialog box interface, and then the first answer text for the first question text and at least one piece of text association information are displayed. The text association information is used to display at least one of the reply content to the first answer text and the candidate question text, so that the user can quickly conduct subsequent intelligent conversations with the system intelligent account through the text association information, avoiding the rigidity problem of a single reply to the current first question text only through the first answer text. Regarding the text association information as a quick operation element for dialogue interaction not only helps to greatly improve the efficiency of subsequent intelligent conversations but also helps the system intelligent account to provide more comprehensive and targeted answer content based on the first question text for the first account, thereby enhancing the user experience of the user. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0020] Figure 1 is a schematic diagram of the implementation environment provided by an exemplary embodiment of the present application;

[0021] Figure 2 is a flowchart of an intelligent dialogue method provided by an exemplary embodiment of the present application;

[0022] Figure 3 is a flowchart of an intelligent dialogue method provided by another exemplary embodiment of the present application;

[0023] Figure 4 is a schematic diagram of the interface of an intelligent dialogue provided by an exemplary embodiment of the present application;

[0024] Figure 5 is a schematic diagram of the interface of an intelligent dialogue provided by another exemplary embodiment of the present application;

[0025] Figure 6 is a flowchart of an intelligent dialogue method provided by still another exemplary embodiment of the present application;

[0026] Figure 7 is a flowchart of training and applying an intelligent dialogue model provided by an exemplary embodiment of the present application;

[0027] Figure 8 It is a schematic diagram of the model structure of a candidate dialogue model provided by an exemplary embodiment of the present application;

[0028] Figure 9 It is a block diagram of the structure of an intelligent dialogue device provided by an exemplary embodiment of the present application;

[0029] Figure 10 It is a block diagram of the structure of an intelligent dialogue device provided by another exemplary embodiment of the present application;

[0030] Figure 11 It is a block diagram of the structure of a terminal provided by an exemplary embodiment of the present application. Detailed implementation manners

[0031] To make the objectives, technical solutions, and advantages of the present application clearer, the following will further describe the embodiments of the present application in detail with reference to the accompanying drawings.

[0032] First, a brief introduction to the nouns involved in the embodiments of the present application is given.

[0033] Artificial Intelligence (AI): It 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. In other words, artificial intelligence is a comprehensive technology in computer science. It attempts to understand the essence of intelligence and produce a new intelligent machine that can react in a way similar to human intelligence. Artificial intelligence also studies the design principles and implementation methods of various intelligent machines to enable the machines to have the functions of perception, reasoning, and decision-making. Artificial intelligence technology is an interdisciplinary subject, involving a wide range of fields, including both hardware-level technologies and software-level technologies. The basic technologies of artificial intelligence generally include, for example, sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, pre-trained model technology, operation / interaction systems, mechatronics, etc. Among them, the pre-trained model is also called the large model or the basic model, and can be widely applied to downstream tasks in various major directions of artificial intelligence after fine-tuning. The software technologies of artificial intelligence mainly include several major directions such as computer vision technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0034] Machine Learning (ML): It is an interdisciplinary subject involving multiple fields such as probability theory, statistics, approximation theory, convex analysis, and algorithm complexity theory. It specializes in studying how computers simulate or implement human learning behaviors to acquire new knowledge or skills and reorganize the existing knowledge structure to continuously improve their own performance. Machine learning is the core of artificial intelligence and the fundamental way to make computers intelligent, and its applications cover all fields of artificial intelligence. Machine learning and deep learning usually include technologies such as artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and learning from demonstration.

[0035] In related technologies, considering that the accuracy of the question content has a great impact on the generation of the answer content, in intelligent dialogue technology, usually emphasis is placed on improving the accuracy of the question content. For example, based on the first half of the text content input by the user, multiple pieces of the second half of the text content for the user to choose from are predicted, and thus the question text in the intelligent dialogue process is determined based on the user's selection operation. Although the above process can improve the accuracy of the question content based on the partial text input by the user, the interaction of the intelligent dialogue is relatively single, usually only a corresponding answer text can be generated based on one question text, making the efficiency of the artificial interaction in the intelligent dialogue scenario relatively low and affecting the user experience of the user.

[0036] In the embodiments of the present application, an intelligent dialogue method is introduced. When the first account interacts with the system intelligent account through the dialog box interface, the first question text is obtained according to the text input in the dialog box interface, and then the first answer text for the first question text and at least one piece of text association information are displayed. It can avoid the problem of being restricted to a single reply to the current first question text only through the first answer text, and use the text association information as a quick operation element for dialogue interaction, which is not only beneficial to greatly improving the efficiency of subsequent intelligent dialogue, but also beneficial to the system intelligent account providing more comprehensive and targeted answer content for the first account, enhancing the user experience of the user. The intelligent dialogue method can be applied to various intelligent question-and-answer scenarios such as customer service scenarios, entertainment scenarios, education scenarios, personal assistant scenarios, etc., and the embodiments of the present application do not limit this.

[0037] It should be noted that before and during the process of collecting relevant data of the user, this application can display a prompt interface, a pop-up window or output a voice prompt message. The prompt interface, pop-up window or voice prompt message is used to prompt the user that their relevant data is being collected at present, so that this application only starts to execute the relevant steps of obtaining the user's relevant data after obtaining the confirmation operation of the user on the prompt interface or the pop-up window. Otherwise (that is, when the confirmation operation of the user on the prompt interface or the pop-up window is not obtained), the relevant steps of obtaining the user's relevant data are ended, that is, the relevant data of the user is not obtained. In other words, all user data collected by this application is collected with the consent and authorization of the user, and the collection, use and processing of relevant user data need to comply with relevant laws, regulations and standards in the relevant region.

[0038] Secondly, the implementation environment involved in the embodiments of this application is described. The intelligent dialogue method provided by the embodiments of this application can be implemented by the terminal alone, or by the server, or by the terminal and the server through data interaction. The embodiments of this application do not limit this. Optionally, the intelligent dialogue method implemented by the interaction between the terminal and the server is taken as an example for description.

[0039] Schematically, please refer to Figure 1 , in this implementation environment, the terminal 110 and the server 120 are involved, and the terminal 110 and the server 120 are connected through the communication network 130.

[0040] In some embodiments, an application program with an intelligent dialogue function is installed in the terminal 110.

[0041] Optionally, the terminal 110 displays a dialog box interface in the application program, where a first account is logged in to the dialog box interface, and the first account is used to interact with the system intelligent account through the dialog box interface. Schematically, the system intelligent account is an account that obtains and displays information by calling the server 120.

[0042] In some embodiments, the terminal 110 receives a text input operation in the dialog box interface.

[0043] Among them, the text input operation is used to obtain a first question text. Schematically, the first question text is the text content determined based on the text input operation of the first account in the text filling area. For example: the first question text is "How to make muffins".

[0044] Optionally, if the terminal 110 independently executes the intelligent dialogue method, the system intelligent account is an account controlled by the terminal 110 based on its own information acquisition function. For example, the terminal 110 pre-stores a knowledge base, queries the knowledge base based on the first question text (such as the first question text is an English word and the knowledge base is an offline dictionary, etc.), and displays the content queried through the system intelligent account.

[0045] Optionally, if the terminal 110 independently executes the intelligent dialogue method, the system intelligent account is an account that obtains information by calling the server 120 and displays the information. Schematically, after obtaining the first question text, the terminal 110 sends the first question text to the server 120 through the communication network 130. The server 120 performs text analysis processes such as semantic analysis on the first question text and generates the content to be displayed by the system intelligent account.

[0046] In some embodiments, based on the text input operation, the first answer text for the first question text is displayed, and at least one piece of text association information is displayed.

[0047] Schematically, the terminal 110 sends the first question text to the server 120 through the communication network 130. The server 120 generates at least one of the first answer text and at least one piece of text association information based on the first question text. For example: the server 120 calls a pre-trained intelligent dialogue system to analyze the first question text and generates the first answer text and / or text association information.

[0048] Optionally, there is a binding relationship between the system intelligent account and the server 120. Based on the server 120 generating the first answer text and text association information, the first answer text for the first question text answered by the system intelligent account and the text association information are displayed; or, the first answer text for the first question text answered by the system intelligent account is displayed, and the text association information is displayed in other areas of the dialogue box interface, etc.

[0049] Among them, the text association information includes at least one of the reply content to the first answer text and the candidate question text that has a semantic association relationship with the first question text.

[0050] Schematically, the text association information provides the reply content for replying to the first answer text and the candidate question text that facilitates the first account to continue the intelligent dialogue process with the intelligent dialogue account, so as to assist the first account and the system intelligent account to continue the dialogue interaction for the topic that has a semantic association relationship with the first question text. That is: the text association information is used to provide the first account with quick operation elements for dialogue interaction with the system intelligent account.

[0051] Schematically, based on the first account, a selection operation is performed on a certain text-related information among at least one text-related information, so that the candidate question text represented by the selected text-related information is used as the question for the next round of intelligent conversation, so that the first account can perform an efficient conversation interaction process with the system intelligent account.

[0052] It should be noted that the above terminals include but are not limited to mobile terminals such as mobile phones, tablet computers, portable laptop computers, intelligent voice interaction devices, intelligent household appliances, in-vehicle terminals, etc., and can also be implemented as desktop computers, etc.; the above servers can be independent physical servers, or a server cluster or distributed system composed of multiple physical servers, and can also be cloud servers that provide 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, Content Delivery Network (CDN), and big data and artificial intelligence platforms.

[0053] Among them, cloud technology refers to a hosting technology that unifies a series of resources such as hardware, application programs, and networks within a wide area network or local area network to realize data calculation, storage, processing, and sharing. Cloud technology is the general term for network technology, information technology, integration technology, management platform technology, application technology, etc. based on the cloud computing business model, and can form a resource pool, which can be used on demand and is flexible and convenient.

[0054] In some embodiments, the above server can also be implemented as a node in a blockchain system.

[0055] Combined with the above noun introduction and application scenarios, the intelligent conversation method provided by this application is described. Taking the application of this method to a terminal as an example, as Figure 2 shown, this method includes the following steps 210 to step 230.

[0056] Step 210, display a dialog box interface.

[0057] Among them, the first account is logged in to the dialog box interface; the first account is used to perform conversation interaction with the system intelligent account through the dialog box interface.

[0058] Schematically, the dialog box interface is the interface content corresponding to the intelligent conversation system. The intelligent conversation system is a system that provides conversation interaction for the user object, so that the user object can perform conversation interaction with the intelligent conversation account configured by the intelligent conversation system by logging in to the first account, so as to provide the information needed by the user object through the conversation interaction process.

[0059] In some embodiments, the intelligent conversation system is implemented as an application program configured on the terminal.

[0060] Schematically, application A is installed on the terminal. Application A has an intelligent conversation function, and a dialog box interface is displayed after application A is opened; or, after an interface display control is triggered after application A is opened, a dialog box interface is displayed, etc.

[0061] Optionally, based on the user object logging in to the first account in application A, the user object can have an intelligent conversation interaction with the system intelligent account on the dialog box interface corresponding to application A, so as to represent that the intelligent conversation interaction is between the first account and the system intelligent account.

[0062] In some embodiments, the intelligent conversation system is implemented as a conversation system that the terminal links to through a website URL.

[0063] Schematically, the user object inputs the website URL B corresponding to the intelligent conversation system through the terminal, so that the terminal links to the intelligent conversation system to display the corresponding dialog box interface; the user object can have an intelligent conversation interaction with the system intelligent account corresponding to the intelligent conversation system through the dialog box interface.

[0064] Optionally, based on the user object logging in to the first account in website URL B, the user object can have an intelligent conversation interaction with the system intelligent account on the dialog box interface corresponding to website URL B, so as to represent that the intelligent conversation interaction is between the first account and the system intelligent account.

[0065] In some embodiments, the intelligent conversation system is implemented as a mini-program that the terminal links to through a specific application.

[0066] Schematically, multiple applications are installed on the terminal, and at least one of them has a mini-program entry function. By entering mini-program C corresponding to the intelligent conversation system, the terminal links to the intelligent conversation system to display the corresponding dialog box interface; the user object can have an intelligent conversation interaction with the system intelligent account corresponding to the intelligent conversation system through the dialog box interface.

[0067] Optionally, based on the first account logged in by the user object in the specific application, the user object can have an intelligent conversation interaction with the system intelligent account on the dialog box interface corresponding to mini-program C in the specific application, so as to represent that the intelligent conversation interaction is between the first account and the system intelligent account.

[0068] In some embodiments, the intelligent conversation system is implemented as a system configured by the terminal itself.

[0069] Schematically, the terminal itself has an intelligent conversation function, which is implemented by an intelligent conversation system configured in itself. The intelligent conversation system is awakened by specific voice commands or specific gestures, etc., and a corresponding dialog box interface is displayed; the user can have an intelligent conversation interaction with the system intelligent account corresponding to the intelligent conversation system through the dialog box interface.

[0070] Optionally, based on the first account logged in by the user on the terminal, the user can have an intelligent conversation interaction with the system intelligent account through the dialog box interface, so as to represent that the intelligent conversation interaction is an interaction between the first account and the system intelligent account.

[0071] It should be noted that the relationship between the above intelligent conversation system and the terminal is only a schematic example. The intelligent conversation system can also be deployed on other computer devices that communicate with the terminal, such as other terminals (for example, if the above terminal is a smart watch, the other terminal is a smart phone wirelessly connected to the smart watch, etc.) or a server, which is not limited here.

[0072] Step 220, receive a text input operation in the dialog box interface.

[0073] Among them, the text input operation is used to obtain the first question text.

[0074] Schematically, various functions related to conversation interaction can be realized through the dialog box interface, such as: text input function, conversation like function, conversation evaluation function, etc.

[0075] Among them, the text input function is used to assist the first account in inputting the first question text for conversation interaction.

[0076] Optionally, the text input function is implemented as a text filling method. The dialog box interface includes a text filling area with a text input function. The user corresponding to the first account fills in the first question text in the text filling area, so that the terminal obtains the first question text.

[0077] Optionally, the text input function is implemented as a voice input method. During the display of the dialog box interface, the voice function can be awakened by voice, and the user corresponding to the first account expresses the first question text based on the voice function, so that the terminal obtains the first question text, etc.

[0078] In some embodiments, the first question text is used as the question content for asking the intelligent conversation system, so that the intelligent conversation system answers the current first question text.

[0079] Schematically, based on the text input operation corresponding to the first account, a first question text is determined. The intelligent dialogue system performs text analysis on the first question text, and uses the system intelligent account corresponding to the intelligent dialogue system to reply to the first account, thereby realizing the dialogue interaction between the first account and the system intelligent account.

[0080] For example, if the first question text input by the user of the first account based on the text input operation is "How to make a delicious muffin?", the terminal obtains the first question text, and then can call the intelligent dialogue system to perform text analysis on the first question text.

[0081] Step 230, based on the text input operation, display the first answer text for the first question text, and display at least one text association information.

[0082] Schematically, after the terminal receives the text input operation and determines the corresponding first question text, it calls the intelligent dialogue system to analyze the first question text to answer the first question text, and the text content generated after the answer is called the first answer text.

[0083] In some embodiments, after the intelligent dialogue system analyzes the first question text, it generates a first answer text for the first question text.

[0084] Among them, the first answer text is the answer content of the intelligent dialogue system for the first question text. The first answer text can be expressed through the system intelligent account, that is, display the interface where the system intelligent account publishes the first answer text.

[0085] For example, after the intelligent dialogue system performs text analysis on the first question text "How to make a delicious muffin?", it generates a first answer text for the first question text, such as: First, mix flour, sugar, etc. evenly... etc.

[0086] In some embodiments, in addition to generating the first answer text, the intelligent dialogue system also generates text association information.

[0087] Among them, the text association information includes at least one of the reply content to the first answer text and the candidate question text having a semantic association relationship with the first question text.

[0088] For example, the text association information includes the reply content to the first answer text; or, the text association information includes the candidate question text having a semantic association relationship with the first question text; or, the text association information includes both the reply content to the first answer text and the candidate question text having a semantic association relationship with the first question text.

[0089] In addition, when there are multiple text association information, it may be implemented such that some of the text association information is implemented as the reply content to the first answer text, some of the text association information is implemented as candidate question texts that have a semantic association relationship with the first question text, and some of the text association information includes reply content and candidate question texts, etc., which are not limited herein.

[0090] Schematically, the reply content is the content that replies to the first answer text and can be used as a template speech for the first account to have a friendly interaction with the system intelligent account.

[0091] Optionally, the reply content is the text content preset by the intelligent dialogue system.

[0092] For example: The reply content is at least one of multiple preset speeches such as "Thank you for your answer", "Thank you for your reply", "Your answer is very helpful to me", etc.

[0093] Optionally, the reply content is the text content generated by the intelligent dialogue system based on the first answer text or the first question text.

[0094] For example: The first question text is "May I ask how to make a delicious muffin", and the first answer text is "First, mix flour, sugar, etc. evenly... etc.", and the intelligent dialogue system associates the text meaning represented by the first answer text and generates reply content such as "Thank you for providing such a detailed production method", "Thank you for your suggestion. It seems that making muffins is a bit simple", etc., which is at least one of multiple generated speeches.

[0095] Optionally, the candidate question text is the question content used to assist the first account to continue the dialogue interaction with the system intelligent account; the semantic association relationship is the condition limit for determining the candidate question text based on the first question text.

[0096] Schematically, the semantic association relationship is used to represent that the first semantic information of the first question text and the second semantic information of the candidate question text meet a preset similarity.

[0097] Among them, the preset similarity is the similarity preset in advance. For example, the semantics of the first question text are extracted to obtain the first semantic information representing the first question text (such as refined into a certain word, etc.), and then the second semantic information that meets the preset similarity is searched from data sources such as the preset vocabulary library based on the first semantic information, and then the candidate question text is generated based on the second semantic information.

[0098] For example, the first semantic information represented by the first question text can be refined as "muffin making method". This first semantic information focuses on "muffin making". Considering that during the "muffin making" process, it is also highly relevant to information such as "selection of muffin materials" and "other muffin making methods", if information such as "selection of muffin materials" and "other muffin making methods" is used as the second semantic information, and the first semantic information and the second semantic information meet the preset similarity relationship, then candidate question texts semantically related to the first question text are generated, such as: "How to select fresh muffin materials", "Are there any other making methods", "Are there any other suggestions for making muffins", etc.

[0099] Schematically, the semantic association relationship is used to represent that the first keyword in the first question text appears in the candidate question text; or, the semantic association relationship is used to represent that a second keyword having a preset similarity to the first keyword in the first question text exists in the candidate question text, etc.

[0100] In some embodiments, the candidate question text can also be text content that is semantically related to the first answer text.

[0101] Schematically, the first question text is "Excuse me, how to make a delicious muffin", and the first answer text is "First, mix flour, sugar, etc. evenly... etc.". Since the muffin making materials - flour, sugar, etc. exist in the first answer text, questions such as "How to select fresh flour" and "How to select suitable sugar" can be generated based on the muffin making materials as candidate question texts, etc.

[0102] Among them, the text association information is used to provide a first account with quick operation elements for dialogue interaction with the system intelligent account.

[0103] Optionally, for at least one piece of text association information, some of the text association information contains reply content for replying to the first answer content, some of the text association information contains candidate question texts, and some of the text association information contains both reply content and candidate question texts; or, in each piece of text association information of at least one piece of text association information, both reply content and candidate question texts exist, etc.

[0104] In some embodiments, the first answer text and at least one piece of text association information can also be implemented as alternative display content.

[0105] Schematically, after determining the first question text, only the first answer text for the first question text is displayed; or, after determining the first question text, only at least one piece of text association information is displayed.

[0106] In some embodiments, at least one piece of text-related information is displayed in a designated area in the dialog interface as content for selection by the first account, and then the interaction status of the subsequent dialog interaction is determined according to the selection status of the first account.

[0107] Optionally, the reply content and candidate question texts are collectively referred to as text-related information, and the user can select only the reply content, only the candidate question texts, or all the contents including the reply content and the candidate question texts.

[0108] Illustratively, at least one text-related information includes the reply content "Thank you for your answer", candidate question text 1 "How to choose fresh flour", candidate question text 2 "Are there any other suggestions for making muffins", and "Thank you for your answer, so how to choose fresh flour" and other contents.

[0109] Optionally, when only the reply content is selected, the dialogue interaction between the first account and the system smart account is deemed to be terminated; or, when only the reply content is selected and no additional text input operation or other operation is received from the first account within a preset time period, the dialogue interaction between the first account and the system smart account is deemed to be terminated, etc.

[0110] Optionally, after the first account selects a candidate question text or text-related information including a candidate question text, it is deemed that the first account continues to ask questions to the system smart account through the candidate question text. The intelligent dialogue system needs to perform text analysis based on the candidate question text and display the corresponding analysis results through the terminal, such as displaying a second answer text corresponding to the candidate question text, and may also display text-related information determined based on the candidate question text.

[0111] It is worth noting that the above are merely illustrative examples and are not limited to the embodiments of the present application.

[0112] In summary, when the first account conducts a dialogue interaction with the system intelligent account through the dialogue box interface, the first question text is obtained based on the text input in the dialogue box interface, and then the first answer text for the first question text and at least one piece of text-related information are displayed. The reply content to the first answer text and at least one of the candidate question texts are presented through the text-related information, so that the user can quickly conduct subsequent intelligent conversations with the system intelligent account through the text-related information, avoiding the rigidity of a single reply to the current first question text only through the first answer text. Taking the text-related information as a quick operation element for dialogue interaction not only helps to greatly improve the efficiency of subsequent intelligent conversations, but also helps the system intelligent account to provide more comprehensive and targeted answer content based on the first question text, thereby enhancing the user experience of the user.

[0113] In an optional embodiment, each piece of text-related information displayed consists of a reply content and a candidate question text. According to the selection operation of the first account for at least one piece of text-related information, an intelligent conversation is continued with the system intelligent account through the selected text-related information. Schematically, as Figure 3 shown, the above Figure 2 shown embodiment can also be implemented as the following steps 310 to step 350; wherein, Figure 2 after the step 230 shown, the following steps 340 to step 350 are further included.

[0114] Step 310, display the dialogue box interface.

[0115] Among them, the first account is logged in to the dialogue box interface, and the first account is used to conduct a dialogue interaction with the system intelligent account through the dialogue box interface.

[0116] Schematically, the system intelligent account is the account corresponding to the intelligent dialogue system and can express the analysis result of the intelligent dialogue system; a one-to-one dialogue interaction process can be carried out between the first account and the system intelligent account in the dialogue box interface.

[0117] As Figure 4 shown, it is a schematic diagram of the dialogue box interface, where the first account 410 is represented by the account identifier A, and the system intelligent account 420 is represented by the account identifier B.

[0118] In an optional embodiment, the dialogue box interface is the interface corresponding to the dialogue group, and the dialogue group includes multiple accounts including the first account and the system intelligent account, such as: the second account, the third account, etc.

[0119] Schematically, a conversation group is a group composed of at least one object account and a system intelligent account. An object account is an account controlled by a user, such as the above-mentioned first account, second account, etc.; a system intelligent account is an account controlled by an intelligent conversation system, etc.

[0120] For example: The first account invites the second account to form a conversation group 1 with the system intelligent account together, and the first account can also invite other object accounts to join the conversation group 1; or, the first account is invited by the second account to join the conversation group 2, and the conversation group 2 includes the system intelligent account and may also include other object accounts, etc.

[0121] With the help of the conversation group, the first account can not only have a conversation interaction with the system intelligent account, but also enable the system intelligent account to perform text analysis on the conversation interaction between the first account and other accounts.

[0122] Step 320, receive a text input operation in the dialog box interface.

[0123] Among them, the text input operation is used to obtain the first question text. That is: through the text input operation, the first account can input the first question text for conversation interaction.

[0124] Optionally, the text filling operation for the text filling area in the dialog box interface is used as the text input operation.

[0125] Schematically, the user corresponding to the first account enables the terminal to obtain the first question text by filling the first question text in the text filling area.

[0126] In an optional embodiment, a trigger operation with a system wake-up instruction is received as the text input operation.

[0127] Among them, the system wake-up instruction is used to wake up the system intelligent account to provide intelligent conversation interaction services; the trigger operation can be implemented as an input operation for inputting text content, or can be implemented as a voice operation for inputting text content, etc.

[0128] Optionally, the system wake-up instruction is implemented as pre-set text content. For example: The system wake-up instruction is the system account name of the system intelligent account, such as: System L, Little l, etc.; or, the system wake-up instruction is content customized by the user, such as: Hi, wake up, etc.

[0129] In some embodiments, the dialog box interface is the interface corresponding to the conversation group, and the conversation group includes multiple accounts including the first account and the system intelligent account.

[0130] Considering that in the process of dialogue interaction among multiple accounts through a dialogue group, some dialogue content requires the participation of the system intelligent account, while some dialogue content does not want the participation of the system intelligent account. Therefore, the presence of the system wake-up instruction in the dialogue content can be used to measure whether the dialogue content requires the participation of the system intelligent account.

[0131] Optionally, receive the trigger operation of the first text content input as a text input operation, and the first text content includes a system wake-up instruction.

[0132] Schematically, when the first account performs a text input operation through the dialog box interface corresponding to the dialogue group, the first account can make the terminal obtain the first question text for dialogue interaction with the system intelligent account by adding the system account name during the process of inputting the text content, that is, the first text content is implemented as the first question text.

[0133] For example: The system wake-up instruction is the system account name "Xiaol" corresponding to the system intelligent account. The dialogue group includes the first account, the second account, and the system intelligent account; when the first account inputs "Xiaol, help me check how to get to scenic spot B more conveniently" in the dialog box interface, based on the inclusion of the system wake-up instruction "Xiaol" therein, the first text content is taken as the first question text asked by the first account to the system intelligent account. Therefore, the system intelligent account needs to participate in this dialogue, that is, call the intelligent dialogue system to perform text analysis on this content, so as to provide an answer to the first question text for both the first account and the second account through the analysis process.

[0134] Optionally, receive the trigger operation of the second text content input as an account interaction operation, and the second text content does not include a system wake-up instruction; the account interaction operation is used for dialogue interaction with other accounts except the system intelligent account.

[0135] Schematically, when the first account does not add the system account name during the process of inputting the text content, it is regarded that the first account conducts dialogue interaction with other accounts except the system intelligent account (such as the second account and the third account in the dialogue group), that is, the input text content is the second text content.

[0136] For example: The system wake-up instruction is the system account name "Xiaol" corresponding to the system intelligent account. The dialogue group includes the first account, the second account, and the system intelligent account; when the first account inputs "Let's go to scenic spot B for a visit" in the dialog box interface, based on the non-inclusion of the system wake-up instruction "Xiaol" therein, the content is taken as the second text content for dialogue interaction between the first account and the second account, that is, the system intelligent account does not participate in this dialogue, that is, does not call the intelligent dialogue system to perform text analysis on this content.

[0137] Similarly, other object accounts such as the second account can also perform the above process through the corresponding dialog box interface, which will not be elaborated here. With the system wake-up instruction, the intelligent dialogue system can be conditionally called for the text analysis process, avoiding the problem of low analysis efficiency caused by continuous text analysis and improving the expression efficiency of the system intelligent account.

[0138] It should be noted that the above is only a schematic example, and the embodiments of the present application are not limited thereto.

[0139] Step 330, based on the text input operation, display the first answer text for the first question text, and display at least one piece of text association information composed of at least one of the reply content and the candidate question text.

[0140] Among them, the text association information includes the reply content to the first answer text and the candidate question text that has a semantic association relationship with the first question text. The text association information is used to provide the first account with quick operation elements for dialogue interaction with the system intelligent account.

[0141] Schematically, as Figure 4 shown, it is a schematic diagram of the dialog box interface. Based on the text input operation of the first account 410, the terminal determines that the first question text 431 is "How to make a delicious muffin?", and the intelligent dialogue system performs text analysis based on the first question text 431 and generates the first answer text 432 for the first question text 431. The first answer text 432 is "Making a delicious muffin is not difficult. Here is a simple muffin-making method..."; the first answer text 432 is the text content generated after answering the first question text 431.

[0142] In addition, in addition to displaying the first answer text, at least one piece of text association information composed of at least one of the reply content and the candidate question text will also be displayed.

[0143] Optionally, for at least one piece of text association information, there is part of the text association information that has the reply content for answering the first answer content, part of the text association information that has the candidate question text, and part of the text association information that has both the reply content and the candidate question text; or, in each piece of text association information of at least one piece of text association information, there is both the reply content and the candidate question text, etc.

[0144] Optionally, taking the reply content and the candidate question text as examples when there is text association information, the text association information includes at least one reply content and at least one candidate question text. Illustratively, a piece of text association information is "Thank you for providing such a detailed production method! Can I make various muffins? Are there any other suggestions for these muffins?", and this piece of text association information includes one reply content "Thank you for providing such a detailed production method" and two candidate question texts, namely "Can I make various muffins?" and "Are there any other suggestions for these muffins?" etc.

[0145] As Figure 4 shown, in the dialog box interface, in addition to the first question text 431 posted by the first account 410 and the first answer text 432 replied by the system intelligent account, three pieces of text association information 433 are also shown, namely "Thank you for your suggestion. It seems that making muffins is a bit simple. But I have a question. How to select fresh flour", "Thank you for providing such a detailed production method! Can I make various muffins? Are there any other suggestions for these muffins?" and "When stirring the batter, what are the advantages of using a rubber spatula? Why".

[0146] In an optional embodiment, at least one piece of text association information is displayed based on the semantic association degree between multiple pieces of text association information and the first question text.

[0147] Among them, there is a semantic association relationship between multiple pieces of text association information and the first question text.

[0148] Illustratively, after determining the first question text, first determine the corpus content that has a semantic association relationship with the first question text from the preset corpus as the text association information, so as to obtain multiple pieces of text association information.

[0149] In addition, determine the semantic association degree between multiple pieces of text association information and the first question text, and the semantic association degree is used to characterize the semantic similarity between the text association information and the first question text.

[0150] Optionally, sort multiple pieces of text association information based on the semantic association degree to obtain an information sorting result; select the first k pieces of text association information from the information sorting result as the text association information to be displayed, that is, display the first k pieces of text association information, where k is a positive integer.

[0151] That is: at least one piece of text association information is the information obtained by screening multiple pieces of text association information based on the semantic association degree.

[0152] In an optional embodiment, an information comment area is displayed.

[0153] Among them, the information comment area is used to evaluate at least one piece of text association information.

[0154] Optionally, the information comment area is used to evaluate multiple text-associated information together, such as fixed-status evaluations like good reviews, medium reviews, bad reviews, or other text comments, etc.

[0155] Optionally, at least one text-associated information corresponds to an information comment area respectively, and the text-associated information can be evaluated specifically through the information comment area corresponding to it.

[0156] It should be noted that the above are only illustrative examples, and the embodiments of the present application are not limited thereto.

[0157] In an optional embodiment, in the dialog box interface, in addition to displaying the first question text, the first answer text, and at least one text-associated information, an evaluation control corresponding to the first answer text is also displayed; receiving a trigger operation on the evaluation control to evaluate the content of the first answer text.

[0158] Among them, the evaluation control is used to evaluate the content of the first answer text.

[0159] Illustratively, after the system intelligent account publishes the first answer text corresponding to the first question text, multiple function controls are displayed on the dialog box interface. Different function controls correspond to different evaluation functions. Based on the trigger operation of the first account on the function control, the first account can evaluate the first answer text specifically, which is beneficial to the update and improvement of the intelligent dialogue system.

[0160] Optionally, the evaluation control includes at least one of multiple controls such as a generation quality evaluation control, a like control, a share control, an evaluation control, etc. For example: the first account triggers an operation on the quality evaluation control to express whether the first answer text meets the expectations of the user when asking the first question; or, the first account triggers an operation on the like control to express that the first answer text is helpful to the user; or, the first account triggers an operation on the share control, and can send the first answer text and / or the first question text and / or the dialog box interface to other objects; or, the first account triggers an operation on the evaluation control, and can make a text evaluation of the answering situation of the first answer text, etc.

[0161] Such as Figure 4 As shown, there are also multiple evaluation controls in the dialog box interface. Among them, the quality evaluation control 441 is used to evaluate the quality of the first answer text, such as evaluations in multiple dimensions like getting better, about the same, getting worse, etc. Among them, "last time" can be either the answer text earlier in the current dialogue interaction round or the answer text in the previous dialogue interaction round, etc.

[0162] Among them, the like control 442 is used to give a positive evaluation to the first answer text, that is, to indicate agreement with the answer; the dislike control 443 is used to give a negative evaluation to the first answer text, that is, to indicate disagreement with the answer; the comment control 444 is used to deeply evaluate the first answer text in the form of text, images, etc.; the forward control 445 is used to share at least one of the first answer text, the first question text, at least one text association information, and the current dialog interface (such as screenshots, screen recordings, etc.), so that other objects can know the reply situation of the system intelligent account to the first question text.

[0163] In an optional embodiment, in the dialog interface, in addition to displaying the first question text, the first answer text, and at least one text association information, a text update control is also displayed.

[0164] Among them, the text update control is used to update the text content of the first answer text.

[0165] Optionally, upon receiving a trigger operation on the text update control, the updated first answer text is displayed.

[0166] Illustratively, if the first account believes that the first answer text replied by the system intelligent account to the first question text does not meet its expectations, the text update control can be triggered to update the first answer text. For example: if the user believes that the first answer text is too long, the text update control can be triggered to update the first answer text; or, if the user believes that the first answer text is inaccurate, the text update control can be triggered to update the first answer text, etc.

[0167] Optionally, the text update control has a first update condition configuration function, which is used to limit the generation of the updated first answer text.

[0168] Illustratively, upon receiving a trigger operation on the text update control, a first condition configuration window is displayed; upon receiving a first condition configuration operation on the first condition configuration window, the updated first answer text is displayed, and the updated first answer text meets the text configuration conditions corresponding to the first condition configuration operation.

[0169] For example: the text configuration condition determined after the first condition configuration operation is that the text length of the first answer text is less than 300. Then, based on the first condition configuration operation, it is necessary to limit the text length of the updated first answer text to be less than 300, etc.

[0170] Such as Figure 4As shown, there is also a text update control 450 in the dialog box interface. Based on the triggering operation of the text update control 450, the intelligent dialogue system is called to update the first answer text based on the first question text, so as to regenerate the first answer text and obtain the updated first answer text expressed by the system intelligent account.

[0171] Optionally, the user can also evaluate the updated first answer text through the above evaluation control, which will not be elaborated here.

[0172] In an optional embodiment, in the dialog box interface, in addition to displaying the first question text, the first answer text, and at least one piece of text association information, an information update control is also displayed.

[0173] Among them, the information update control is used to update the information of at least one piece of text association information.

[0174] Optionally, upon receiving the triggering operation on the information update control, at least one piece of updated text association information is displayed.

[0175] Illustratively, if the first account believes that at least one piece of text association information provided by the system intelligent account is not helpful for subsequent dialogue interactions, the information update control can be triggered to update at least one piece of text association information.

[0176] Optionally, the information update control has a second update condition configuration function for restricting the generation of at least one piece of updated text association information.

[0177] Illustratively, upon receiving the triggering operation on the information update control, a second condition configuration window is displayed; upon receiving the second condition configuration operation on the second condition configuration window, at least one piece of updated text association information is displayed, and the updated text association information conforms to the information configuration conditions corresponding to the second condition configuration operation.

[0178] For example: The information configuration conditions determined after the second condition configuration operation are: the presence of the keywords "muffin" and "production materials". Then, based on the second condition configuration operation, it is necessary to restrict the presence of the two keywords "muffin" and "production materials" in the updated text association information.

[0179] It should be noted that the above is only an illustrative example, and the embodiments of the present application are not limited thereto.

[0180] Step 340, receive the selection operation for the first text association information among at least one piece of text association information.

[0181] Schematically, after displaying at least one text-associated information, the first account can select at least one text-associated information so as to perform a subsequent intelligent conversation process through the selected text-associated information.

[0182] Wherein, the first text-associated information is any one of at least one text-associated information, and the first text-associated information is a text-associated information determined based on the selection operation of the first account.

[0183] Schematically, the first text-associated information includes a first candidate question text, and the first candidate question text is a question selected by the first account and used as a question to assist the first account in subsequent conversation interaction with the system intelligent account.

[0184] Optionally, the selection operation can be implemented as a triggering operation such as a click operation or a long-press operation, or can be implemented in forms such as a "voice operation" or a "gesture operation", which is not limited here.

[0185] As Figure 4 shown, based on the click operation of the first account on "What are the advantages of using a rubber spatula when stirring batter? Why" in at least one text-associated information 433 as the selection operation, this text-associated information is used as the first text-associated information.

[0186] Step 350, based on the selection operation, display a dialog box interface for continuing the conversation interaction with the first candidate question text as the second question text.

[0187] Schematically, after the first account selects the first text-associated information, determine the first candidate question text in the first text-associated information, and then use the first candidate question text as the second question text to perform a subsequent conversation interaction process.

[0188] As Figure 4 shown, based on the click operation of the first account on "What are the advantages of using a rubber spatula when stirring batter? Why" in at least one text-associated information 433 as the selection operation, determine that the first candidate question text in the first text-associated information is "What are the advantages of using a rubber spatula when stirring batter? Why", and use this first candidate question text as the second question text to continue the subsequent conversation interaction.

[0189] Wherein, the acquisition method of the second question text is different from that of the first question text, but the functions achieved are the same. Schematically, both the first question text and the second question text are used as questions expressed by the first account, aiming to obtain the content replied by the system intelligent account.

[0190] As Figure 5As shown, the first account 510 is represented by account identifier A, and the system intelligent account 520 is represented by account identifier B; based on performing the above selection operation on the dialog box interface as shown in Figure 4 shown, "What are the advantages of using a rubber spatula when stirring batter? Why" is used as the second question text 530 displayed in the dialog box interface as shown in Figure 5 shown. This second question text 530 is regarded as a question raised again by the first account after the first question text, enabling the intelligent dialogue system to perform text analysis based on the second question text 530.

[0191] In some embodiments, a dialog box interface for continuing the dialogue interaction with the first text association information is displayed.

[0192] In some embodiments, a second answer text for answering the second question text is displayed.

[0193] As Figure 5 shown, a second answer text 540 for answering the second question text 530 is displayed to express "the advantages of using a rubber spatula" and "the reasons for achieving such advantages", etc.

[0194] In some embodiments, in addition to displaying the second answer text, at least one text association information having a semantic association relationship with the second question text is also displayed.

[0195] Among them, for the content of determining at least one text association information corresponding to the second question text, reference can be made to the content of determining at least one text association information corresponding to the first question text above, and details will not be elaborated here.

[0196] As Figure 5 shown, in addition to displaying the second answer text 540, at least one text association information of types such as "Thank you for your answer. I also want to ask how to quickly make a vegetarian dish at home", "Thank you for your explanation! Then, when making muffins, why do we need to preheat the baking pan and apply oil", "Thank you for your explanation! Then, when making muffins, why do we need to preheat the baking pan and apply oil" is also displayed.

[0197] In an alternative embodiment, an information merging operation for the second text association information and the third text association information is received.

[0198] Among them, the second text association information consists of the first reply content and the second candidate question text, and the third text association information consists of the second reply content and the third candidate question text.

[0199] Schematically, the second text association information and the third text association information are different text association information among at least one piece of text association information. The information merging operation is used to merge the second text association information and the third text association information.

[0200] Optionally, the information merging operation is implemented as a multi-selection operation on the second text association information and the third text association information; or, the information merging operation is implemented as a sliding operation of dragging the second text association information and the third text association information to a specified area, etc.

[0201] In an optional embodiment, in response to the information merging operation, a dialog box interface for continuing the dialogue interaction with the combined reply content and the combined question text is displayed.

[0202] Among them, the combined reply content includes at least one of the first reply content and the second reply content, and the combined question text is the text content obtained by merging the second candidate question text and the third candidate question text.

[0203] Schematically, the second text association information is "Thank you for your answer. I also want to know how to obtain flour", and the third text association information is "Your answer is great. Do I need any additional tools?"; Based on this, the first reply content corresponding to the second text association information is "Thank you for your answer", and the second candidate question text corresponding to the second text association information is "I also want to know how to obtain flour"; The second reply content corresponding to the third text association information is "Your answer is great", and the third candidate question text corresponding to the third text association information is "Do I need any additional tools?".

[0204] The information merging operation is implemented as a multi-selection operation on the second text association information and the third text association information.

[0205] Based on the first reply content "Thank you for your answer" and the second reply content "Your answer is great", the combined reply content is obtained. For example, the combined reply content is: "Thank you for your answer. Your answer is great", that is, comprehensive citation, or it can also be implemented as "Thank you for your answer" or "Your answer is great", that is, alternative citation, etc.

[0206] Based on the second candidate question text and the third candidate question text, the combined question content is obtained, and the combined question text is the summary form of the two question texts. For example, the combined question text is: "What additional tools do I need and how to obtain flour?", so that through the information merging operation, the answers to multiple questions can be known at one time, improving the efficiency of question answering.

[0207] In an optional embodiment, a text input operation for obtaining the third question text is received.

[0208] Among them, the third question text is the question content obtained after another text input operation on the dialog box interface.

[0209] For example: The user first performs a text input operation on the dialog box interface to input the first question text as "How to make bread"; based on the text input operation, the first answer text - "Steps of making bread" is displayed, and at least one text association information is displayed; then, the user does not select at least one text association information for the time being, but performs a text input operation again to input the third question text "What will happen if the egg white is not whipped until foamy".

[0210] In an optional embodiment, in response to a dialog association relationship existing between the first question text and the third question text, based on at least one text association information and the third question text, the text association information for the third question text is displayed.

[0211] Schematically, after receiving the first question text and the subsequently input third question text, the intelligent dialogue system automatically determines the dialog association degree between the first question text and the third question text, and when the dialog association degree meets the preset association degree, it is considered that there is a dialog association relationship between the first question text and the third question text.

[0212] Optionally, the dialog association degree is implemented as the semantic intersection degree between the first question text and the third question text. For example: The first question text "How to make bread" makes a semantic description around the bread making method, and the third question text "What will happen if the egg white is not whipped until foamy" makes a semantic description around the "egg white" material in the bread making process. The semantic intersection degree between the two question texts is 30%. This semantic intersection degree of 30% is regarded as the dialog association degree and compared with the preset association degree. If the preset association degree is 20%, it is considered that there is a dialog association relationship between the first question text and the third question text.

[0213] The above implementation of the dialog association degree as the semantic intersection degree is only a schematic example. The dialog association degree can also be implemented as at least one of various forms such as the keyword coincidence ratio, semantic similarity, etc., which is not limited here.

[0214] When it is determined that there is a dialog association relationship between the first question text and the third question text, based on the second question text that the user has not selected from at least one text association information as the question content, therefore, in order to include the semantic content represented by at least one text association information corresponding to the first question text, when displaying the text association information corresponding to the third question text, the intelligent dialogue system will comprehensively analyze the information with at least one text association information corresponding to the first question text and the third question text.

[0215] Schematically, a piece of text association information corresponding to the first question text is "Thank you for your reply. I would like to know what kind of flour is best for making bread." The third question text is "What will happen if the egg whites are not whipped until foamy?" Based on the dialogue association relationship between the first question text and the third question text, the intelligent dialogue system synthesizes "Thank you for your reply. I would like to know what kind of flour is best for making bread." and "What will happen if the egg whites are not whipped until foamy?" to generate text association information corresponding to the third question text, such as: "Thank you for your reply. Are there any additional points to note regarding the reasonable use of ingredients for making bread?" etc.

[0216] This schematically shows the content of generating a piece of text association information corresponding to the third question text based on a piece of text association information and the third question text. It is also possible to generate multiple pieces of text association information corresponding to the third question text based on a piece of text association information and the third question text, or generate a piece of text association information corresponding to the third question text based on multiple pieces of text association information and the third question text, or generate multiple pieces of text association information corresponding to the third question text based on multiple pieces of text association information and the third question text, etc. This is not limited here.

[0217] In an optional embodiment, in response to the existence of a dialogue association relationship between the first question text and the third question text, based on the first answer text and the third question text, the third answer text is displayed.

[0218] Optionally, during the process of not yet finishing displaying the first answer text for the first question text or not yet displaying the first answer text, if the third question text input by the user is received, the intelligent dialogue system can merge the first answer text that has not yet been fully displayed and the content for answering the third question text, so as to display the third answer text. This third answer text can not only answer the first question text but also answer the third question text, etc.

[0219] It should be noted that the above is only a schematic example, and the embodiments of the present application are not limited thereto.

[0220] In summary, by presenting at least one of the reply content to the first answer text and the candidate question text through the text association information, it enables the user to quickly conduct subsequent intelligent conversations with the system intelligent account through the text association information, avoiding the rigidity problem of a single reply to the current first question text only through the first answer text. Regarding the text association information as a quick operation element for dialogue interaction not only helps to greatly improve the efficiency of subsequent intelligent conversations but also enables the system intelligent account to provide more comprehensive and targeted answer content for the first account based on the first question text, thereby enhancing the user experience of the user.

[0221] In an embodiment of the present application, the content of selecting the first text association information and using the first candidate question information in the first text association information as the second question text to continue the intelligent dialogue interaction is introduced. By means of the text association information, a form of fast and continuous intelligent dialogue is provided for the first account, thereby improving the human-computer interaction efficiency of the intelligent dialogue. In addition, the first account can also efficiently discuss similar topics with other target accounts through the dialogue group at the same time, which also helps the system intelligent account to provide more perfect interaction services for multiple target accounts and further improves the human-computer interaction efficiency.

[0222] In an optional embodiment, in addition to being implemented in the form of the above reply content and the second question text, at least one displayed text association information can also flexibly display content such as location information and redemption association information as the text association information according to the content of the first question text. Schematically, as Figure 6 shown, the above Figure 2 shown embodiment can also be implemented as the following steps 610 to step 642; the content of displaying at least one text association information in step 230 shown above Figure 2 can also be implemented as the following steps 641 to step 642.

[0223] Step 610, display a dialog box interface.

[0224] Among them, the first account is logged in to the dialog box interface, and the first account is used to interact with the system intelligent account through the dialog box interface.

[0225] Schematically, the dialog box interface is the interface content corresponding to the intelligent dialogue system, and the intelligent dialogue system is a system that provides dialogue interaction for the user object.

[0226] Step 610 has been introduced in the above steps 210 and 310, and will not be elaborated here.

[0227] Step 620, receive the text input operation in the dialog box interface.

[0228] Among them, the text input operation is used to obtain the first question text.

[0229] Schematically, various functions related to dialogue interaction can be realized through the dialog box interface, such as: text input function, dialogue like function, dialogue evaluation function, etc. Optionally, the text input function is implemented in a text filling manner, and the dialog box interface includes a text filling area with a text input function. The user object corresponding to the first account fills the first question text in the text filling area, so that the terminal obtains the first question text.

[0230] In some embodiments, the first query text serves as the question content for querying the intelligent dialogue system, and is used to enable the intelligent dialogue system to answer the current first query text.

[0231] Step 620 has been introduced in the above steps 220 and 320, and will not be elaborated here.

[0232] Step 630: Based on the text input operation, display the first answer text for the first query text.

[0233] Illustratively, the terminal determines the first query text based on the text input operation, and calls the intelligent dialogue system according to the text input operation to perform text analysis on the first query text, so as to generate the first answer text for answering the first query text.

[0234] Optionally, the intelligent dialogue system is a system deployed on the terminal, and the terminal can directly call the intelligent dialogue system; or, the intelligent dialogue system is a system deployed on other devices, and the terminal can indirectly call the intelligent dialogue system through the communication process with other devices, etc.

[0235] Illustratively, the intelligent dialogue system receives the first query text for text analysis to generate the first answer text for answering the first query text; then, the intelligent dialogue system feeds back the first answer text to the terminal so that the terminal displays the first answer text. For example: The terminal displays the first answer text on the dialog box interface.

[0236] In an alternative embodiment, at least one piece of multimedia answer data is displayed as the first answer text.

[0237] Among them, the multimedia answer data is used to represent the multimedia data for answering the first query text in multimedia form.

[0238] Illustratively, in addition to expressing the content for answering the first query text in text form, the content for answering the first query text can also be expressed in multimedia form, that is, using the multimedia answer data as the first answer text.

[0239] Optionally, the multimedia answer data is implemented in the form of video data, that is, the multimedia form is video; or the multimedia answer data is implemented in the form of audio data, that is, the multimedia form is audio, etc.

[0240] In some embodiments, in response to receiving a trigger operation for the first multimedia answer data, play the first multimedia content corresponding to the first multimedia answer data.

[0241] Among them, the first multimedia content includes at least one of video content and audio content, and the first multimedia content is used to answer the first query text in multimedia form.

[0242] Schematically, the first multimedia answer data is the selected multimedia data among at least one multimedia answer data. The first multimedia answer data is selected according to a trigger operation, and the first multimedia content corresponding to the first multimedia answer data is determined. Then, the first multimedia content is played, so as to answer the first question text in a more vivid manner.

[0243] In some embodiments, in response to receiving a trigger operation for the first multimedia answer data, a multimedia web page corresponding to the first multimedia answer data is jumped to and displayed; the first multimedia content corresponding to the first multimedia answer data is played in the multimedia web page.

[0244] Schematically, if the first multimedia content corresponding to the first multimedia answer data needs to be presented through a multimedia web page, after receiving the trigger operation for the first multimedia answer data, the multimedia web page is jumped to and displayed, so as to play the first multimedia content in the multimedia web page.

[0245] In an alternative embodiment, a solution website is displayed as the first answer text.

[0246] Wherein, the solution website is used to comprehensively express the content for answering the first question text.

[0247] Schematically, for the convenience of more comprehensive analysis and understanding of the first question text, the intelligent dialogue system provides the solution website as the first answer text for the first account through the system intelligent account. The first account can obtain the content for more comprehensive answer to the first question text by clicking on the solution website, etc.

[0248] It should be noted that the above is only a schematic example, and the embodiments of the present application are not limited thereto.

[0249] In an alternative embodiment, in addition to displaying the first answer text on the dialog box interface, at least one piece of text-related information is also displayed. The at least one piece of text-related information can be implemented as the information composed of at least one of the above-mentioned reply content and candidate question text, and can also be implemented as other information for replying or asking again around the first question text.

[0250] Schematically, according to the difference in the implementation form of the text-related information, the content of the text-related information can also be implemented as the following step 641 or step 642.

[0251] Step 641, in response to the text keyword corresponding to the geographical name included in the first question text, the location information corresponding to the geographical name is displayed.

[0252] Schematically, when the intelligent dialogue system analyzes the first question text, the intelligent dialogue system can also obtain the text keywords in the first question text.

[0253] Optionally, the first question text consists of multiple text words, and the text keywords are at least one text word with the largest word weight in the first question text. Schematically, the text keywords are text words determined based on the word weights of the text words in the first question text.

[0254] Schematically, keyword extraction algorithms are deployed in the intelligent dialogue system, such as: at least one of multiple keyword extraction algorithms such as Term Frequency–Inverse Document Frequency (TF-IDF), TextRank, and Latent Dirichlet Allocation (LDA).

[0255] For example: The intelligent dialogue system analyzes each of the multiple text words in the first question text by means of the deployed text ranking algorithm, and selects the text word with the largest word weight in the first question text as the text keyword.

[0256] In some embodiments, a map including a location identifier is displayed as location information.

[0257] Among them, the location identifier is used to represent the location coordinates corresponding to the geographical name.

[0258] Schematically, if the first question text is "How to get to Scenic Spot D?", and the geographical name "Scenic Spot D" is determined based on the keyword extraction algorithm, then based on the text keyword corresponding to the geographical name in the first question text, when displaying the location information, a map including a location identifier is displayed, such as displaying the map content centered on Scenic Spot D as the location information, where Scenic Spot D is represented by the location identifier.

[0259] Optionally, the location information can also be expressed by longitude and latitude. Schematically, the longitude and latitude corresponding to Scenic Spot D are displayed as the location information.

[0260] In some embodiments, a map link for jumping to the map viewing interface is displayed as location information.

[0261] Among them, the map link is used to display the map area including the geographical name in the map viewing interface. Schematically, after clicking the map link, the map viewing interface is jumped to and map content is displayed in the map viewing interface, where a certain map area is displayed centered on the location of the geographical name to show the situations of other locations near the geographical name.

[0262] In some embodiments, the terminal corresponds to the first position, and the geographical name corresponds to the second position.

[0263] Schematically, determine the current position of the terminal, that is, determine the first position; based on the text keyword corresponding to the geographical name included in the first question text, determine the position corresponding to the geographical name, that is, determine the second position.

[0264] Optionally, display the planned path from the first position to the second position as location information.

[0265] Schematically, express the location information through a route traffic map. Schematically, display the content such as the route / public transportation / driving route section from the current position to scenic spot D as location information, etc.

[0266] It should be noted that the above is only a schematic example, and the embodiments of the present application are not limited thereto.

[0267] Step 642, in response to the text keyword corresponding to the item resource included in the first question text, display at least one exchange association information as text association information.

[0268] Among them, the text keyword is the text vocabulary determined based on the vocabulary weight of the text vocabulary in the first question text; the exchange association information is used to assist in exchanging item resources.

[0269] Schematically, the first question text is "how to make muffins". Based on the keyword extraction algorithm, the text keyword is determined to be "muffins". Then, based on the extracted text keyword corresponding to the item resource "muffins", display the exchange association information used to assist in exchanging this item resource - muffins.

[0270] Optionally, express the exchange association information through an exchange website. Schematically, display the websites for purchasing "muffins" and for purchasing "muffin materials" as exchange association information.

[0271] Optionally, express the exchange association information by displaying a jump inquiry prompt message. Schematically, display the jump inquiry prompt for jumping to the application program for exchanging item resources as exchange association information. For example: display "Do you want to jump to application program T" to purchase materials related to "muffins", etc.

[0272] In an optional embodiment, the dialog box interface includes a setting control, and the setting control is used to configure the display situation of the text association information.

[0273] In some embodiments, when the setting control is in the enabled state, based on the text input operation, display the first answer text for the first question text, and display at least one text association information.

[0274] Schematically, based on the opening operation of the setting control by the user, the setting control is presented in an open state. Therefore, in addition to being able to display the first answer text, at least one piece of text-related information can also be displayed.

[0275] In some embodiments, when the setting control is in a closed state, based on a text input operation, the first answer text for the first question text is displayed; wherein, at least one piece of text-related information is in a hidden state.

[0276] Schematically, based on the closing operation of the setting control by the user, the setting control is presented in a closed state. Therefore, at least one piece of text-related information is hidden, that is, only the first answer text is displayed, avoiding interference to the user from the text-related information.

[0277] In some embodiments, based on the opening operation of the setting control to use the expansion content as text-related information, the expansion content is displayed as text-related information, and the expansion content includes at least one of the above position information and redemption-related information; based on the closing operation of the setting control to use the expansion content as text-related information, the expansion content is hidden.

[0278] By means of the flexible configuration of the open state and closed state of the setting control, the above content can achieve the purpose of flexibly adjusting the text-related information, not only effectively avoiding the problem of a large amount of cached data caused by always displaying the text-related information, but also avoiding the problem of affecting the user experience.

[0279] It should be noted that the above realization of the text-related information as redemption-related information or position coordinate information is only a schematic example. The text-related information can also be realized as a combination of the above redemption-related information and position information (such as there are multiple text keywords, or one text keyword indicates multiple meanings), or can be realized as a combination of the above reply content, candidate question text, redemption-related information, and position information or at least two of them. Of course, the text-related information corresponding to the first question text can also be realized as content determined based on the first question text and / or the first answer text. Similarly, the text-related information corresponding to the second question text can also be realized as content determined based on at least one of the first question text, the first answer text, the second question text, and the second answer text. Other similar situations will not be elaborated, and the embodiments of the present application do not limit this either.

[0280] In summary, by presenting the reply content to the first answer text and at least one piece of information in the candidate question text through text association information, it enables the user to quickly conduct subsequent intelligent conversations with the system intelligent account through the text association information, avoiding the rigidity of a single reply to the current first question text only through the first answer text. Taking the text association information as a quick operation element for dialogue interaction not only helps to greatly improve the efficiency of subsequent intelligent conversations but also enables the system intelligent account to provide more comprehensive and targeted answer content for the first account based on the first question text, thereby enhancing the user experience of the user.

[0281] In the embodiments of the present application, content such as location information and redemption association information is introduced as text association information, enriching the implementation forms of text association information and facilitating the first account to perform diverse and efficient intelligent dialogue operations based on the text association information, further enhancing the human-computer interaction efficiency.

[0282] In an optional embodiment, when the above intelligent dialogue method is executed through a terminal, the terminal can analyze the first question text based on the intelligent dialogue system configured by itself and obtain the first answer text and text association information; the terminal can also communicate with the server. The terminal sends the first question text to the server, and the server analyzes the first question text based on the intelligent dialogue system configured by itself. After obtaining the first answer text and text association information, it sends them to the terminal to implement the dialogue interaction process.

[0283] Schematically, the intelligent dialogue system is a pre-configured system in which a pre-trained intelligent dialogue model is deployed. Thus, when the intelligent dialogue system is called to analyze questions such as the first question text and the second question text, the intelligent dialogue system performs text analysis on the questions through the pre-deployed intelligent dialogue model, and then generates and / or invokes content such as the first answer text and at least one piece of text association information.

[0284] Optionally, as Figure 7 shown, the candidate dialogue model is trained through the following steps 710 to 740 to obtain the intelligent dialogue model; and then through the following step 750, in the way of the intelligent dialogue system calling the intelligent dialogue model, at least one piece of text association information is generated based on the text input operation, as well as content such as the first answer text and at least one piece of text association information for the first question text.

[0285] Step 710, obtain a pair of sample data.

[0286] Among them, the pair of sample data is used to train the candidate dialogue model and obtain the intelligent dialogue model.

[0287] Schematically, the candidate dialogue model is the model to be trained. Optionally, the candidate dialogue model is a model that has been pre-trained to a certain extent and has certain text analysis capabilities and intelligent dialogue capabilities.

[0288] Schematically, the candidate dialogue model is any one of a variety of basic models such as the BigScience Large Open-science Open-access Multilingual Language Model (BLOOM), the Finetuned Language Net (FLN), and the Large Language Model MetaArtificial Intelligence (LLaMa).

[0289] Optionally, taking the BLOOM model as an example of the candidate dialogue model, BLOOM(7B) is used as the basic model, where 7B is used to represent the parameter scale, and there are also several other BLOOM models with different parameter scales such as 560M, 1.1B, 1.7B, 3B, 7.1B, and 176B.

[0290] Among them, BLOOM(7B) is based on the Transformer structure and makes many changes to the original Transformer architecture. For example, an optional position embedding scheme or a novel activation function.

[0291] Among them, the sample data pair consists of at least one sample question data and sample answer data for answering at least one sample question data.

[0292] Schematically, there is a corresponding relationship between the sample question data and the sample answer data. For example: determining a sample answer statement for answering a sample question data based on one sample question data; or determining a sample answer statement for answering a sample question data based on multiple sample question data; or determining multiple sample answer statements for answering a sample question data based on one sample question data; or determining multiple sample answer statements for answering a sample question data based on multiple sample question data, etc.

[0293] In an optional embodiment, at least one sample question data and sample answer data for answering at least one sample question data are obtained; based on the answer correlation relationship between the sample question data and the sample answer data, the at least one question data and the at least one sample answer data are spliced to generate a sample data pair.

[0294] Among them, the answer association relationship is used to represent the relationship that the sample answer data is used to answer the corresponding sample question data.

[0295] Schematically, obtain a sample question data and a sample answer data for answering the sample question data, and form a sample data pair 1 based on the sample question data and the sample answer data. The sample data pair 1 is represented as: sample question data Q - sample answer data A.

[0296] Schematically, obtain multiple sample question data and multiple sample answer data, and determine the corresponding situation between the sample question data and the sample answer data based on the answer association relationship, so as to form at least one sample data pair from the multiple sample question data and the multiple sample answer data.

[0297] For example: Obtain 2 sample question data and 2 sample answer data. The sample question data Q1 is answered by the sample answer data A1, and the sample question data Q2 is answered by the sample answer data A2. And the sample question data Q1 and the sample question data Q2 are the content obtained in a round of dialogue interaction process. Then form a sample data pair 2 based on the 2 sample question data and the 2 sample answer data. The sample data pair 2 is represented as: sample question data Q1 - sample answer data A1 - sample question data Q2 - sample answer data A2, etc.

[0298] Or, obtain 2 sample question data and 3 sample answer data. The sample question data Q1 is answered by the sample answer data A1, and the sample question data Q2 is answered by the sample answer data A2 and the sample answer data A3. And the sample question data Q1 and the sample question data Q2 are the content obtained in a round of dialogue interaction process. Then form a sample data pair 3 based on the 2 sample question data and the 3 sample answer data. The sample data pair 3 is represented as: sample question data Q1 - sample answer data A1 - sample question data Q2 - sample answer data A2 - sample answer data A3, etc.

[0299] Or, obtain 2 sample question data and 2 sample answer data. The sample question data Q1 is answered by the sample answer data A1, and the sample question data Q2 is answered by the sample answer data A2 and the sample answer data A3. And the sample question data Q1 and the sample question data Q2 are the content obtained in two different rounds of dialogue interaction processes respectively. Then form two sample data pairs based on the 2 sample question data and the 2 sample answer data, which are the sample data pair 4, represented as: sample question data Q1 - sample answer data A1, and the sample data pair 5, represented as: sample question data Q2 - sample answer data A2, etc.

[0300] The number of dialogue interaction turns involved in the above content is used to represent the process from the start to the end of the dialogue interaction. It may end the dialogue interaction after one question, or after one question and one answer, or after one question and multiple answers, or after multiple questions and one answer, or after multiple questions and multiple answers, etc. The number of dialogue interaction turns will not be elaborated here.

[0301] In some embodiments, the following two dialogue datasets are used as the datasets for obtaining sample data pairs.

[0302] (1) Knowledge-driven Conversation (KdConv): A Chinese multi-turn dialogue dataset driven by multi-domain knowledge, which contains 4.5K dialogue interactions and 86K utterances from three domains (movies, music, and travel), with an average number of turns of 19.

[0303] (2) Natural Conversation (Natural Conv) is an open dialogue dataset, which is topic-driven Chinese dialogue generation. This dataset is closer to human conversations and has natural attributes, including complete natural environments such as scenario assumptions, free topic expansion, greetings, etc. It contains approximately 400K sentences and 19.9K conversations, covering multiple domains (including but not limited to sports, entertainment, and technology). The average number of turns is 20, which is significantly longer than other dialogue datasets.

[0304] The above two dialogue datasets are only illustrative examples, and the embodiments of the present application are not limited thereto.

[0305] Among them, a sample data pair corresponds to a data label, and the data label is used to represent the question content after the last sample question data in the sample data pair.

[0306] In some embodiments, considering the input length limit of the model, the number of dialogue turns can be appropriately reduced, such as limiting the input number of dialogue turns within 5 turns. That is, the dialogue discourse represented by the sample data pair is within 5 turns.

[0307] Optionally, dialogue data with different numbers of turns are uniformly sampled in different topics to construct training data - sample data pairs for supervised fine-tuning (SFT), so that the sampled data covers various numbers of turns within 5 turns and covers various topics, thereby maintaining the richness and diversity of the training data.

[0308] Schematically, a constructed sample data pair is shown as follows.

[0309] A: Do you know about the advantages and disadvantages of solar power generation?

[0310] B: Of course. The advantages of solar power generation include being clean, renewable, low-carbon and environmentally friendly; the disadvantages are the relatively high cost, large initial investment and the dependence on the abundance of sunlight resources in most areas.

[0311] H: Then, what is the current development level of solar power generation technology?

[0312] A: Currently, significant progress has been made in solar power generation technology, with the efficiency gradually increasing and the power generation cost continuously decreasing, which makes solar power generation increasingly popular globally.

[0313] H: Can you give an example of the solar policies in some regions?

[0314] A: Of course. Taking Country D as an example, in 2000, it provided strong policy support for solar power generation by implementing Regulation K, making Country D a leader in the solar power generation industry. Other countries have also successively introduced policies to support solar power generation and are committed to the development of the solar power generation industry.

[0315] Next question: (blank)

[0316] Label for the next question (answer): "Where is the largest solar power station in our country?"

[0317] Among them, H and A above represent different objects. H represents the object asking the question, and A represents the object answering the question. The above dialogue interaction represents 3 rounds of dialogue (one question and one answer). Taking the above 3 rounds of dialogue as a sample data pair and the label of the above next question as the label corresponding to the sample data pair, the purpose of the candidate dialogue model is to predict the next question based on the sample data pair and compare the predicted next question with the label of the next question, so as to carry out the model training process.

[0318] Schematically, by constructing a dataset containing multi-round data with different numbers of rounds and the richness of topics, the richness of the dataset can make the model more robust and generalizable, so that in different scenarios, it can learn the above information of the dialogue based on different rounds and predict the questions that the user may ask next.

[0319] That is: the sample data pair corresponds to the data label, and the data label is used to represent the content of the question after the last sample question data.

[0320] Step 720, analyze the sample data pair through the candidate dialogue model to obtain the prediction result corresponding to the sample data pair.

[0321] Among them, the prediction result is the result after predicting the question content.

[0322] Schematically, taking the above BLOOM (7B) model as an example of a candidate dialogue model, as Figure 8 shown, it is a schematic diagram of the BLOOM model structure. Based on the BLOOM (7B) model structure that retains the causal decoder-only, the following two changes have been made to the BLOOM model.

[0323] (1) Attention with Linear Biases Enables Input Length Extrapolation, that is: ALiBi positional embedding

[0324] Compared with adding positional information in the embedding layer, ALiBi directly decays the attention scores based on the distances between keys and queries. Although the initial motivation of ALiBi was that it could extrapolate to longer sequences, it was found that it could also bring more balanced training and better downstream performance on the original sequence length, surpassing learnable vectors and rotary vectors.

[0325] (2) Normalization layer 810, that is: Embedding LayerNorm

[0326] In the initial experiment of training the BLOOM (104B) parameter model, normalization operation (layer normalization) was first attempted immediately after the embedding layer, and it was found that this could significantly improve training stability. Although this process has a penalty on zero-shot generalization, an additional normalization layer was added after the first embedding layer of the BLOOM model to avoid training instability. In addition, float16 was used in the initial BLOOM (104B) experiment, while bfloat16 was used in the final training. Because float16 has always been considered the cause of many instabilities observed during the training of LLMs. Bfloat16 has the potential to alleviate the need for Embedding LayerNorm.

[0327] Schematically, the sample data pair is sample question data Q1 - sample answer data A1 - sample question data Q2 - sample answer data A2. At this time, the candidate dialogue model performs text analysis on this sample data pair, aiming to predict the question content Q3 after the sample answer data A2, and taking the predicted content as the prediction result.

[0328] Step 730, determine the loss value based on the difference between the prediction result and the data label.

[0329] In an optional embodiment, the task based on the candidate dialogue model is to predict the content of the question. Therefore, when the candidate dialogue model performs text analysis on the sample data pair, in addition to predicting the content after the last sample question data based on the sample data pair, it will also predict the next sample question data based on the previous sample question data.

[0330] Schematically, the sample data pair is sample question data Q1 - sample answer data A1 - sample question data Q2 - sample answer data A2. At this time, when the candidate dialogue model performs text analysis on this sample data pair, in addition to outputting the prediction result obtained by predicting the question content Q3 after the sample answer data A2, the candidate dialogue model may also output another prediction result obtained by predicting the sample question data Q2 after the sample answer data A1, resulting in multiple prediction results. Furthermore, when determining the loss value based on the difference between the prediction result and the data label, there is a determination bias.

[0331] Based on this, in some embodiments, a masked feature representation is generated, and the masked feature representation is used to mask other prediction results except the prediction result; other prediction results are masked with the masked feature representation, and the loss value is obtained based on the difference between the prediction result and the data label.

[0332] Among them, other prediction results are the results obtained by predicting at least one sample question data.

[0333] Schematically, with the help of the masked feature representation, it is beneficial to determine the loss value for accurately training the candidate dialogue model based on the difference between the unique prediction result and the data label corresponding to the sample data pair.

[0334] For example, the prediction result obtained by predicting the sample question data Q2 is masked with the masked feature representation, so that the loss value is determined based on the prediction result obtained by predicting the question content Q3 after the sample answer data A2 and the difference between the data label.

[0335] That is: when performing SFT on the candidate dialogue model, each sample data pair participating in the training generally consists of two parts, namely the instruction (prompt / instruction) and the answer (answer). That is, some answers in the form of questions and answers need to be provided to the model for learning. After having the experience of the previous pre-training, the essence of SFT is to perform the next token prediction process. Only because it is more desirable for the model to focus on the prediction of the data label part, a masked feature representation (mask vector) can be generated to mask the part where the loss value (loss) is not desired to be calculated.

[0336] Schematically, what is done is to splice the instruction (question data) and the answer (solution data), and add a start and end symbol on both sides of the answer, so as to splice the question and answer into a complete training data through '[PAD]', including the sample data pair and the corresponding data label. The purpose of doing this is to enable the model to capture the correlation between the question and the answer and understand the context of the entire conversation.

[0337] Among them, SFT is a transfer learning technique that performs additional training on the basis of a large pre-trained model to adapt to specific tasks. These tasks usually include classification, entity recognition, question answering, etc. In supervised fine-tuning, a labeled dataset is used, and in this dataset, each input sample has a corresponding label or target output. Using the supervised learning method, the Bloom model is subjected to SFT based on the spliced training text.

[0338] Optionally, during the training process, hyperparameters such as the learning rate and batch size need to be adjusted to optimize the model performance. By adopting the Bloom (7B) model and performing SFT training, it helps to better achieve the instruction association function, thereby improving the naturalness and coherence of the human-computer dialogue.

[0339] Step 740, training the candidate dialogue model with the loss value until an intelligent dialogue model is obtained.

[0340] Schematically, after obtaining the loss value, the candidate dialogue model is trained with the loss value. In response to the loss value no longer decreasing, an intelligent dialogue model is obtained; or, in response to the decrease value of the loss value being less than the preset value, an intelligent dialogue model is obtained, etc.

[0341] Step 750, in response to a text input operation, calling the intelligent dialogue model configured on the computer device to generate at least one text association information and a first solution text for the first question text.

[0342] Among them, the intelligent dialogue model is a pre-trained model. For example: the intelligent dialogue model is the model obtained through the above fine-tuning training.

[0343] Schematically, in response to a text input operation, the terminal calls the intelligent dialogue system to analyze the first question text. The intelligent dialogue system analyzes the first question text based on the intelligent dialogue model deployed on it to predict the first solution text corresponding to the first question text, and can also predict the questions that the first account may ask later based on the first question text and / or the first solution text, so as to generate at least one candidate question text, and then generate at least one text association information by integrating at least one of the reply content and the at least one candidate question text.

[0344] The interface content related to the above step 750 may refer to the above Figure 2 , Figure 3 and Figure 5 illustrated embodiments, which will not be elaborated here.

[0345] It should be noted that the above is only a schematic example, and the embodiments of the present application are not limited thereto.

[0346] In summary, by presenting at least one of the reply content to the first answer text and the candidate question text through text association information, so that the user can quickly conduct subsequent intelligent conversations with the system intelligent account through the text association information, avoiding the rigidity problem of simply replying to the current first question text with the first answer text, and using the text association information as a quick operation element for dialogue interaction, which not only helps to greatly improve the efficiency of subsequent intelligent conversations, but also helps the system intelligent account to provide more comprehensive and targeted answer content based on the first question text, thereby enhancing the user experience of the user.

[0347] In the embodiments of the present application, the content of obtaining the intelligent dialogue model by training the candidate dialogue model and applying the intelligent dialogue model is introduced. The candidate dialogue model is trained by sample data composed of sample question data and sample answer data. When predicting the prediction result corresponding to the sample data pair, other sample answer data are masked by mask features, so that the candidate dialogue model focuses on the question content after the last sample answer data. Then, based on the prediction result and the data label representing the question data after the last sample answer data, the loss value for more targeted model training of the candidate dialogue model is determined, and thus the candidate dialogue model is trained by the loss value until the intelligent dialogue model is obtained; the intelligent dialogue model is deployed in the intelligent dialogue system, so that the terminal can generate the answer text and text association information corresponding to the question text by flexibly invoking the intelligent dialogue system, improving the human-computer interaction efficiency in the intelligent dialogue process.

[0348] Figure 9 is a structural block diagram of an intelligent dialogue device provided by an exemplary embodiment of the present application. As Figure 9 shown, the device includes the following parts:

[0349] The interface display module 910 is used to display a dialog box interface, and the first account is logged in to the dialog box interface, and the first account is used to conduct dialogue interaction with the system intelligent account through the dialog box interface;

[0350] The operation receiving module 920 is used to receive a text input operation in the dialog box interface, and the text input operation is used to obtain a first question text;

[0351] A text display module 930, configured to display a first answer text for the first question text based on the text input operation, and display at least one piece of text association information, where the text association information includes at least one of a reply content for the first answer text and a candidate question text having a semantic association relationship with the first question text, and the text association information is used to provide a quick operation element for the first account to perform a dialogue interaction with the system intelligent account.

[0352] In an optional embodiment, the text display module 930 is further configured to receive a selection operation for a first text association information in the at least one piece of text association information, where the first text association information is composed of the reply content and a first candidate question text; and based on the selection operation, display the dialog box interface for continuing the dialogue interaction with the first candidate question text as the second question text.

[0353] In an optional embodiment, the text display module 930 is further configured to display a text update control, where the text update control is used to update the text content of the first answer text; and receive a trigger operation for the text update control, and display the updated first answer text.

[0354] In an optional embodiment, the text display module 930 is further configured to, in response to a text keyword corresponding to a geographical name included in the first question text, display location information corresponding to the geographical name; where the text keyword is at least one text word determined based on the word weights of multiple text words in the first question text.

[0355] In an optional embodiment, the terminal corresponds to a first location, and the geographical name corresponds to a second location;

[0356] The text display module 930 is further configured to display a map including a location identifier as the location information, where the location identifier is used to represent the location coordinates corresponding to the geographical name; or display a map link for jumping to a map viewing interface as the location information, where the map link is used to display a map area including the geographical name in the map viewing interface; or display a planned path from the first location to the second location as the location information.

[0357] In an optional embodiment, the text display module 930 is further configured to, in response to a text keyword corresponding to an item resource included in the first question text, display at least one piece of exchange association information as the text association information; where the text keyword is at least one text word determined based on the word weights of multiple text words in the first question text, and the exchange association information is used to assist in exchanging the item resource.

[0358] In an optional embodiment, the dialog box interface includes a setting control for configuring the display of text association information;

[0359] The text display module 930 is further configured to, when the setting control is in an enabled state, display the first answer text for the first question text and the at least one piece of text association information based on the text input operation; or, when the setting control is in a disabled state, display the first answer text for the first question text based on the text input operation; wherein the at least one piece of text association information is in a hidden state.

[0360] In an optional embodiment, the text display module 930 is further configured to receive an information merging operation for second text association information and third text association information, where the second text association information is composed of a first reply content and a second candidate question text, and the third text association information is composed of a second reply content and a third candidate question text; in response to the information merging operation, display the dialog box interface for continuing the conversation interaction with the combined reply content and the combined question text, where the combined reply content includes at least one of the first reply content and the second reply content, and the combined question text is the text content obtained by merging the second candidate question text and the third candidate question text.

[0361] In an optional embodiment, the text display module 930 is further configured to display the at least one piece of text association information based on the semantic association degree between multiple pieces of text association information and the first question text; wherein there is a semantic association relationship between the multiple pieces of text association information and the first question text, and the at least one piece of text association information is the information obtained by screening the multiple pieces of text association information based on the semantic association degree.

[0362] In an optional embodiment, the operation receiving module 920 is further configured to receive, in the dialog box interface, a trigger operation with a system wake-up instruction as the text input operation, where the system wake-up instruction is used to wake up the system intelligent account to provide intelligent conversation interaction services.

[0363] In an optional embodiment, the dialog box interface is an interface corresponding to a conversation group, and the conversation group includes multiple accounts including the first account and the system intelligent account;

[0364] The operation receiving module 920 is further configured to receive a triggering operation for inputting the first text content as the text input operation, where the first text content includes the system wake-up instruction; receive a triggering operation for inputting the second text content as an account interaction operation, where the second text content does not include the system wake-up instruction, and the account interaction operation is used to perform a dialogue interaction with an account other than the system intelligent account.

[0365] In an alternative embodiment, the text display module 930 is further configured to display an information update control for updating the at least one text-associated information; receive a triggering operation on the information update control, and display the updated at least one text-associated information.

[0366] In an alternative embodiment, the text display module 930 is further configured to receive the text input operation for obtaining the third question text, where the third question text is the question content obtained after the text input operation is performed again on the dialog box interface; in response to a dialogue association relationship existing between the first question text and the third question text, based on the at least one text-associated information and the third question text, display the text-associated information for the third question text.

[0367] In an alternative embodiment, the device is deployed in a computer device;

[0368] As Figure 10 shown, the device further includes:

[0369] A model calling module 940, configured to, in response to the text input operation, call an intelligent dialogue model configured on the computer device to generate the at least one text-associated information and generate the first answer text for the first question text, where the intelligent dialogue model is a pre-trained model.

[0370] In an alternative embodiment, the model calling module 940 is further configured to obtain a sample data pair for training a candidate dialogue model and obtaining the intelligent dialogue model, where the sample data pair is composed of at least one sample question data and sample answer data for answering the at least one sample question data, the sample data pair corresponds to a data label, and the data label is used to represent the question content after the last sample question data in the sample data pair; analyze the sample data pair through the candidate dialogue model to obtain a prediction result corresponding to the sample data pair, where the prediction result is the result after predicting the question content; determine a loss value based on the difference between the prediction result and the data label; and train the candidate dialogue model with the loss value until the intelligent dialogue model is obtained.

[0371] In an alternative embodiment, the model calling module 940 is further configured to obtain the at least one sample query data and sample answer data for answering the at least one sample query data; based on the answer association relationship between the sample query data and the sample answer data, perform data combination on the at least one query data and at least one sample answer data to generate the sample data pair.

[0372] In an alternative embodiment, the model calling module 940 is further configured to generate a mask feature representation for masking other prediction results except the prediction result, where the other prediction results are the results obtained after predicting the at least one sample query data; mask the other prediction results with the mask feature representation, and obtain the loss value based on the difference between the prediction result and the data label.

[0373] In summary, by presenting at least one of the reply content to the first answer text and the candidate query text through text association information, it enables the user to quickly conduct subsequent intelligent conversations with the system intelligent account through the text association information, avoiding the limitation of a single reply to the current first query text only through the first answer text. Taking the text association information as a quick operation element for dialogue interaction not only helps to greatly improve the efficiency of subsequent intelligent conversations but also enables the system intelligent account to provide more comprehensive and targeted answer content for the first account based on the first query text, thereby enhancing the user experience of the user.

[0374] It should be noted that: for the intelligent dialogue device provided in the above embodiment, only the division of the above functional modules is used for illustration. In actual applications, the above functions can be allocated to different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the intelligent dialogue device provided in the above embodiment and the intelligent dialogue method embodiment belong to the same concept, and the specific implementation process is detailed in the method embodiment, which will not be elaborated here.

[0375] Figure 11The block diagram of the electronic device 1100 provided by an exemplary embodiment of the present application is shown. The electronic device 1100 may be a portable mobile terminal, such as: a smart phone, a vehicle-mounted terminal, a tablet computer, an MP3 player (Moving Picture Experts Group Audio Layer III), an MP4 (Moving Picture Experts Group Audio Layer IV) player, a notebook computer or a desktop computer. The electronic device 1100 may also be referred to by other names such as a user equipment, a portable terminal, a laptop terminal, a desktop terminal, etc.

[0376] Generally, the electronic device 1100 includes: a processor 1101 and a memory 1102.

[0377] The processor 1101 may include one or more processing cores, such as a quad-core processor, an octa-core processor, etc. The processor 1101 may be implemented in at least one hardware form of DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), PLA (Programmable Logic Array). The processor 1101 may also include a main processor and a coprocessor. The main processor is a processor for processing data in the wake state, also known as the CPU (Central Processing Unit); the coprocessor is a low-power processor for processing data in the standby state. In some embodiments, the processor 1101 may be integrated with a GPU (Graphics Processing Unit), and the GPU is responsible for rendering and drawing the content to be displayed on the display screen. In some embodiments, the processor 1101 may also include an AI (Artificial Intelligence) processor, and the AI processor is used to process the computational operations related to machine learning.

[0378] The memory 1102 may include one or more computer-readable storage media, and the computer-readable storage media may be non-transitory. The memory 1102 may also include high-speed random access memory and non-volatile memory, such as one or more disk storage devices, flash storage devices. In some embodiments, the non-transitory computer-readable storage medium in the memory 1102 is used to store at least one instruction, and the at least one instruction is used to be executed by the processor 1101 to implement the intelligent dialogue method provided by the method embodiment of the present application.

[0379] In some embodiments, the electronic device 1100 may further optionally include: a peripheral device interface 1103 and at least one peripheral device. The processor 1101, the memory 1102, and the peripheral device interface 1103 may be connected through a bus or signal lines. Each peripheral device may be connected to the peripheral device interface 1103 through a bus, signal lines, or a circuit board. Specifically, the peripheral device includes at least one of a radio frequency circuit 1104, a display screen 1105, an audio circuit 1107, and a power supply 1109.

[0380] The peripheral device interface 1103 can be used to connect at least one peripheral device related to I / O (Input / Output) to the processor 1101 and the memory 1102. In some embodiments, the processor 1101, the memory 1102, and the peripheral device interface 1103 are integrated on the same chip or circuit board; in some other embodiments, any one or two of the processor 1101, the memory 1102, and the peripheral device interface 1103 can be implemented on a separate chip or circuit board, and this embodiment does not limit this.

[0381] The radio frequency circuit 1104 is used to receive and transmit RF (Radio Frequency) signals, also known as electromagnetic signals. The radio frequency circuit 1104 communicates with a communication network and other communication devices through electromagnetic signals. The radio frequency circuit 1104 converts an electrical signal into an electromagnetic signal for transmission, or converts the received electromagnetic signal into an electrical signal. Optionally, the radio frequency circuit 1104 includes: an antenna system, an RF transceiver, one or more amplifiers, a tuner, an oscillator, a digital signal processor, a codec chipset, a subscriber identity module card, and so on. The radio frequency circuit 1104 can communicate with other terminals through at least one wireless communication protocol. The wireless communication protocol includes but is not limited to: the World Wide Web, a metropolitan area network, an intranet, each generation of mobile communication networks (2G, 3G, 4G, and 5G), a wireless local area network, and / or a WiFi (Wireless Fidelity) network. In some embodiments, the radio frequency circuit 1104 may further include a circuit related to NFC (Near Field Communication), and this application does not limit this.

[0382] The display screen 1105 is used to display the UI (User Interface). The UI may include graphics, text, icons, videos, and any combination thereof. When the display screen 1105 is a touch display screen, the display screen 1105 also has the ability to collect touch signals on or above the surface of the display screen 1105. The touch signal can be input to the processor 1101 as a control signal for processing. At this time, the display screen 1105 can also be used to provide virtual buttons and / or a virtual keyboard, also known as soft buttons and / or a soft keyboard. In some embodiments, there can be one display screen 1105, which is provided on the front panel of the electronic device 1100; in other embodiments, there can be at least two display screens 1105, which are respectively provided on different surfaces of the electronic device 1100 or are in a folding design; in other embodiments, the display screen 1105 can be a flexible display screen, which is provided on the curved surface or the folding surface of the electronic device 1100. Even, the display screen 1105 can be set to an irregular non-rectangular shape, that is, a special-shaped screen. The display screen 1105 can be prepared using materials such as LCD (Liquid Crystal Display) and OLED (Organic Light-Emitting Diode).

[0383] The audio circuit 1107 may include a microphone and a speaker. The microphone is used to collect sound waves of the user and the environment, and convert the sound waves into electrical signals and input them to the processor 1101 for processing, or input them to the radio frequency circuit 1104 to achieve voice communication. For the purpose of stereo collection or noise reduction, there can be multiple microphones, which are respectively provided at different parts of the electronic device 1100. The microphone can also be an array microphone or an omnidirectional collection type microphone. The speaker is used to convert the electrical signal from the processor 1101 or the radio frequency circuit 1104 into sound waves. The speaker can be a traditional thin film speaker or a piezoelectric ceramic speaker. When the speaker is a piezoelectric ceramic speaker, it can not only convert the electrical signal into sound waves audible to humans, but also convert the electrical signal into sound waves inaudible to humans for uses such as ranging. In some embodiments, the audio circuit 1107 may also include a headphone jack.

[0384] The power supply 1109 is used to supply power to each component in the electronic device 1100. The power supply 1109 can be alternating current, direct current, a disposable battery, or a rechargeable battery. When the power supply 1109 includes a rechargeable battery, the rechargeable battery can be a wired rechargeable battery or a wireless rechargeable battery. A wired rechargeable battery is a battery charged through a wired line, and a wireless rechargeable battery is a battery charged through a wireless coil. The rechargeable battery can also be used to support fast charging technology.

[0385] In some embodiments, the electronic device 1100 further includes one or more sensors 1110. The one or more sensors 1110 include, but are not limited to: a pressure sensor 1113, an optical sensor 1115, and a proximity sensor 1116.

[0386] The pressure sensor 1113 may be disposed on the side frame of the electronic device 1100 and / or the lower layer of the display screen 1105. When the pressure sensor 1113 is disposed on the side frame of the electronic device 1100, it can detect the holding signal of the user on the electronic device 1100, and the processor 1101 performs left / right hand recognition or shortcut operations according to the holding signal collected by the pressure sensor 1113. When the pressure sensor 1113 is disposed on the lower layer of the display screen 1105, the processor 1101 controls the operable controls on the UI interface according to the pressure operation of the user on the display screen 1105. The operable controls include at least one of button controls, scroll bar controls, icon controls, and menu controls.

[0387] The optical sensor 1115 is used to collect the ambient light intensity. In one embodiment, the processor 1101 can control the display brightness of the display screen 1105 according to the ambient light intensity collected by the optical sensor 1115. Specifically, when the ambient light intensity is high, the display brightness of the display screen 1105 is increased; when the ambient light intensity is low, the display brightness of the display screen 1105 is decreased. In another embodiment, the processor 1101 can also dynamically adjust the shooting parameters of the camera module according to the ambient light intensity collected by the optical sensor 1115.

[0388] The proximity sensor 1116, also known as a distance sensor, is usually disposed on the front panel of the electronic device 1100. The proximity sensor 1116 is used to collect the distance between the user and the front of the electronic device 1100. In one embodiment, when the proximity sensor 1116 detects that the distance between the user and the front of the electronic device 1100 is gradually decreasing, the processor 1101 controls the display screen 1105 to switch from the lit state to the off state; when the proximity sensor 1116 detects that the distance between the user and the front of the electronic device 1100 is gradually increasing, the processor 1101 controls the display screen 1105 to switch from the off state to the lit state.

[0389] Those skilled in the art can understand that Figure 11 the structure shown in

[0390] does not constitute a limitation on the electronic device 1100, and may include more or fewer components than shown in the figure, or combine certain components, or adopt a different component layout. Figure 2The terminal or server shown. The computer device includes a processor and a memory. At least one instruction, at least one program, a code set, or an instruction set is stored in the memory. The at least one instruction, at least one program, the code set, or the instruction set is loaded and executed by the processor to implement the intelligent dialogue method provided by each of the above method embodiments.

[0391] An embodiment of the present application further provides a computer-readable storage medium. At least one instruction, at least one program, a code set, or an instruction set is stored on the computer-readable storage medium. The at least one instruction, at least one program, the code set, or the instruction set is loaded and executed by the processor to implement the intelligent dialogue method provided by each of the above method embodiments.

[0392] An embodiment of the present application further provides a computer program product or a computer program. The computer program product or the computer program includes computer instructions. The computer instructions are stored in a computer-readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the intelligent dialogue method described in any one of the above embodiments.

[0393] Optionally, the computer-readable storage medium may include: read-only memory (ROM, Read Only Memory), random access memory (RAM, Random Access Memory), solid state drives (SSD, Solid State Drives), or optical discs, etc. Among them, the random access memory may include resistive random access memory (ReRAM, Resistance RandomAccess Memory) and dynamic random access memory (DRAM, Dynamic Random Access Memory). The serial numbers of the above embodiments of the present application are only for description and do not represent the advantages and disadvantages of the embodiments.

[0394] Those of ordinary skill in the art can understand that all or part of the steps to implement the above embodiments can be completed by hardware, or can be completed by a program instructing related hardware. The program can be stored in a computer-readable storage medium. The storage medium mentioned above can be read-only memory, a magnetic disk, or an optical disc, etc.

[0395] The above are only optional embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. An intelligent dialogue method, characterized in that, the method includes: displaying a dialogue box interface logged in with a first account, where the first account is used to conduct dialogue interactions with the system intelligent account through the dialogue box interface; receiving a text input operation in the dialogue box interface, where the text input operation is used to obtain a first question text; based on the text input operation, displaying a first answer text for the first question text, and displaying at least one piece of text association information, where the text association information includes at least one of a reply content to the first answer text and a candidate question text that has a semantic association relationship with the first question text, and the text association information is used to provide a first account with a quick operation element for conducting dialogue interactions with the system intelligent account.

2. The method according to claim 1, characterized in that, after displaying at least one piece of text association information, it further includes: receiving a selection operation for a first text association information in the at least one piece of text association information, where the first text association information is composed of the reply content and a first candidate question text; based on the selection operation, displaying the dialogue box interface for continuing the dialogue interaction with the first candidate question text as the second question text.

3. The method according to claim 1 or 2, characterized in that, the method further includes: displaying a text update control for updating the text content of the first answer text; receiving a trigger operation for the text update control and displaying the updated first answer text.

4. The method according to claim 1 or 2, characterized in that, the method includes: in response to the first question text including a text keyword corresponding to a geographical name, displaying location information corresponding to the geographical name; wherein, the text keyword is at least one text word determined based on the word weights of multiple text words in the first question text.

5. The method according to claim 4, characterized in that, the method is executed by a terminal, the terminal corresponds to a first location, and the geographical name corresponds to a second location; the displaying location information corresponding to the geographical name includes: displaying a map including a location identifier as the location information, where the location identifier is used to represent the location coordinates corresponding to the geographical name; or, displaying a map link for jumping to a map viewing interface as the location information, where the map link is used to display a map area including the geographical name in the map viewing interface; or, displaying a planned path from the first location to the second location as the location information.

6. The method according to claim 1 or 2, characterized in that, the displaying at least one piece of text association information includes: in response to the first question text including a text keyword corresponding to an item resource, displaying at least one piece of exchange association information as the text association information; Among them, the text keyword is at least one text word determined based on the word weights of multiple text words in the first question text, and the redemption association information is used to assist in redeeming the item resources.

7. The method according to claim 1 or 2, wherein, the dialog box interface includes a setting control for configuring the display situation of the text association information; based on the text input operation, display a first answer text for the first question text, and display at least one piece of text association information, including: when the setting control is in an on state, based on the text input operation, display the first answer text for the first question text, and display the at least one piece of text association information; or, when the setting control is in an off state, based on the text input operation, display the first answer text for the first question text; wherein, the at least one piece of text association information is in a hidden state.

8. The method according to claim 1 or 2, wherein, after displaying the at least one piece of text association information, further includes: receiving an information merging operation for the second text association information and the third text association information, the second text association information is composed of a first reply content and a second candidate question text, and the third text association information is composed of a second reply content and a third candidate question text; in response to the information merging operation, display the dialog box interface for continuing the conversation interaction with the combined reply content and the combined question text, the combined reply content includes at least one of the first reply content and the second reply content, and the combined question text is the text content obtained by merging the second candidate question text and the third candidate question text.

9. The method according to claim 1 or 2, wherein, displaying the at least one piece of text association information includes: display the at least one piece of text association information based on the semantic association degree between multiple pieces of text association information and the first question text; wherein, there is a semantic association relationship between the multiple pieces of text association information and the first question text, and the at least one piece of text association information is the information obtained by screening the multiple pieces of text association information based on the semantic association degree.

10. The method according to claim 1 or 2, wherein, receiving the text input operation in the dialog box interface includes: in the dialog box interface, receive a trigger operation with a system wake-up instruction as the text input operation, and the system wake-up instruction is used to wake up the system intelligent account to provide intelligent dialogue interaction services.

11. The method according to claim 10, wherein, the dialog box interface is the interface corresponding to the conversation group, and the conversation group includes multiple accounts including the first account and the system intelligent account; receiving a trigger operation with a system wake-up instruction as the text input operation includes: receive a trigger operation for inputting the first text content as the text input operation, and the first text content includes the system wake-up instruction; Receiving a trigger operation for the second text content as an account interaction operation, where the second text content does not include the system wake-up instruction, and the account interaction operation is used for dialogue interaction with other accounts outside the system intelligent account.

12. The method according to claim 1 or 2, wherein, the method further includes: Receiving the text input operation for obtaining the third question text, where the third question text is the question content obtained after the text input operation is performed again on the dialog box interface; In response to a dialogue association relationship existing between the first question text and the third question text, based on the at least one text association information and the third question text, displaying the text association information for the third question text.

13. The method according to claim 1 or 2, wherein, the method is executed by a computer device, and the method further includes: In response to the text input operation, calling an intelligent dialogue model configured on the computer device to generate the at least one text association information, and generating the first answer text for the first question text, where the intelligent dialogue model is a pre-trained model.

14. The method according to claim 13, wherein, the method further includes: Obtaining a sample data pair, where the sample data pair is used to train a candidate dialogue model to obtain the intelligent dialogue model, the sample data pair is composed of at least one sample question data and sample answer data for answering the at least one sample question data, the sample data pair corresponds to a data label, and the data label is used to represent the question content after the last sample question data in the sample data pair; Analyzing the sample data pair through the candidate dialogue model to obtain a prediction result corresponding to the sample data pair, where the prediction result is the result after predicting the question content; Based on the difference between the prediction result and the data label, determining a loss value; Training the candidate dialogue model with the loss value until the intelligent dialogue model is obtained.

15. The method according to claim 14, wherein, the obtaining of the sample data pair includes: Obtaining the at least one sample question data and sample answer data for answering the at least one sample question data; Based on the answer association relationship between the sample question data and the sample answer data, performing data combination on the at least one question data and at least one sample answer data to generate the sample data pair.

16. The method according to claim 14, wherein, the determining of the loss value based on the difference between the prediction result and the data label includes: Generating a mask feature representation, where the mask feature representation is used to cover other prediction results except the prediction result, and the other prediction results are the results obtained after predicting the at least one sample question data; Covering the other prediction results with the mask feature representation, and based on the difference between the prediction result and the data label, obtaining the loss value.

17. An intelligent dialogue device, It is characterized in that the device includes: an interface display module, configured to display a dialog box interface, on which a first account is logged in, and the first account is used to conduct dialogue interaction with the system intelligent account through the dialog box interface; an operation receiving module, configured to receive a text input operation in the dialog box interface, and the text input operation is used to obtain a first question text; a text display module, configured to display a first answer text for the first question text based on the text input operation, and display at least one piece of text association information, where the text association information includes at least one of a reply content to the first answer text and a candidate question text that has a semantic association relationship with the first question text, and the text association information is used to provide the first account with a quick operation element for conducting dialogue interaction with the system intelligent account.

18. A computer device It is characterized in that the computer device includes a processor and a memory, and at least one segment of program is stored in the memory, and the at least one segment of program is loaded and executed by the processor to implement the intelligent dialogue method according to any one of claims 1 to 16.

19. A computer-readable storage medium It is characterized in that at least one segment of program is stored in the storage medium, and the at least one segment of program is loaded and executed by a processor to implement the intelligent dialogue method according to any one of claims 1 to 16.

20. A computer program product It is characterized in that it includes computer instructions, and when the computer instructions are executed by a processor, the intelligent dialogue method according to any one of claims 1 to 16 is implemented.

Images