Conversation robot construction method, system and equipment and storage medium

Abstract

The invention provides a construction method, system and device of a dialogue robot and a storage medium, and the construction method builds a digital employee platform as a visual dialogue robot development low-code platform, so that common business personnel can also log in and access the digital employee platform. The dialogue robot meeting the self service requirements is simply, conveniently and quickly constructed in the digital employee platform, and enterprises are assisted to realize high-quality digital operation.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and more specifically, to a construction method, system, device and storage medium of a conversational robot. Background Art

[0002] Existing conversational robot products are usually developed for general fields. They can respond to some basic questions raised by users in general fields, but they are difficult to meet the needs of some business fields that require strong professionalism and broad knowledge.

[0003] At present, if users want to develop a professional conversational robot for their own business field, they need to hand it over to experienced IT (Information Technology) developers to carry out traditional code development based on the functions that the conversational robot needs to have in the target business field. However, IT developers are much less familiar with actual business scenarios than ordinary business personnel, and because traditional code development has high requirements for professional knowledge and technology, business personnel who are more familiar with actual business scenarios cannot participate in the development process of conversational robots, making it difficult to customize conversational robots for actual business scenarios. Summary of the invention

[0004] In view of this, the present application provides a method, system, device and storage medium for constructing a conversational robot. By creating a digital employee platform as a visual low-code platform for developing conversational robots, ordinary business personnel can also log in to access the digital employee platform, and easily and quickly construct a conversational robot that meets their own business needs in the digital employee platform, thereby helping enterprises achieve high-quality digital operations.

[0005] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, preferred embodiments are specifically cited below and described in detail with reference to the attached drawings.

[0006] In a first aspect, an embodiment of the present application provides a method for constructing a conversational robot, which is applied to a target user terminal in a construction system, wherein the construction system further includes: a digital employee platform, a cloud assistant management platform, and a cloud assistant application; wherein the construction method includes:

[0007] In response to a target user's login access request to the digital employee platform, displaying a target conversation robot pre-created by the target user in the cloud assistant management platform in the user interface of the target user;

[0008] In response to the target user adding an instruction to a knowledge base of the target conversational robot in the digital employee platform, configuring the knowledge base included in the knowledge base adding instruction as a question-and-answer knowledge base of the target conversational robot;

[0009] In response to a function configuration instruction of the target user for the target conversational robot in the digital employee platform, automatically generating a question-and-answer logic matching the function configuration instruction according to function description information and / or function module identifier contained in the function configuration instruction;

[0010] In response to the target user importing the question-and-answer knowledge base and the question-and-answer logic into the natural language processing engine in the digital employee platform, the original dialogue model deployed in the natural language processing engine is trained according to the imported question-and-answer knowledge base and the question-and-answer logic, and the trained original dialogue model is configured as the dialogue model of the target dialogue robot to obtain a configured target dialogue robot.

[0011] In a second aspect, an embodiment of the present application provides a system for constructing a conversational robot, the system comprising: a digital employee platform, a cloud assistant management platform, and a cloud assistant application; wherein the digital employee platform is used to:

[0012] In response to a target user's login access request to the digital employee platform, displaying a target conversation robot pre-created by the target user in the cloud assistant management platform in the user interface of the target user;

[0013] In response to the target user adding an instruction to a knowledge base of the target conversational robot in the digital employee platform, configuring the knowledge base included in the knowledge base adding instruction as a question-and-answer knowledge base of the target conversational robot;

[0014] In response to a function configuration instruction of the target user for the target conversational robot in the digital employee platform, automatically generating a question-and-answer logic matching the function configuration instruction according to function description information and / or function module identifier contained in the function configuration instruction;

[0015] In response to the target user importing the question-and-answer knowledge base and the question-and-answer logic into the natural language processing engine in the digital employee platform, the original dialogue model deployed in the natural language processing engine is trained according to the imported question-and-answer knowledge base and the question-and-answer logic, and the trained original dialogue model is configured as the dialogue model of the target dialogue robot to obtain a configured target dialogue robot.

[0016] In a third aspect, an embodiment of the present application provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above-mentioned method for constructing a conversational robot when executing the computer program.

[0017] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the above-mentioned method for constructing a conversational robot are executed.

[0018] The technical solution provided by the embodiments of the present application may have the following beneficial effects:

[0019] The embodiments of the present application provide a method, device, equipment and storage medium for constructing a conversational robot. By creating a digital employee platform as a visual low-code platform for developing conversational robots, ordinary business personnel can also log in to access the digital employee platform, and easily and quickly construct a conversational robot that meets their own business needs on the digital employee platform, thereby helping enterprises achieve high-quality digital operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0021] Figure 1 A schematic diagram of the structure of a conversational robot construction system provided in an embodiment of the present application is shown;

[0022] Figure 2 A schematic diagram of a flow chart of a method for constructing a conversational robot provided in an embodiment of the present application is shown;

[0023] Figure 3 A schematic diagram of an interface interaction of a newly added knowledge base provided in an embodiment of the present application is shown;

[0024] Figure 4 An interface schematic diagram of a zero-code development interface provided in an embodiment of the present application is shown;

[0025] Figure 5 A schematic diagram of a robot creation page provided in an embodiment of the present application is shown;

[0026] Figure 6 A schematic diagram of the structure of an electronic device 600 provided in an embodiment of the present application. DETAILED DESCRIPTION

[0027] To make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. It should be understood that the drawings in the present application only serve the purpose of explanation and description and are not used to limit the scope of protection of the present application. In addition, it should be understood that the schematic drawings are not drawn in real proportion. The flowchart used in this application shows the operations implemented according to some embodiments of the present application. It should be understood that the operations of the flowchart can be implemented out of sequence, and the steps without logical context can be reversed in order or implemented simultaneously. In addition, those skilled in the art can add one or more other operations to the flowchart under the guidance of the content of the present application, or remove one or more operations from the flowchart.

[0028] In addition, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. The components of the embodiments of the present application described and shown in the drawings here can be arranged and designed in various configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application claimed for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present application.

[0029] It should be noted that the term "comprising" will be used in the embodiments of the present application to indicate the existence of the features declared thereafter, but does not exclude the addition of other features.

[0030] here, Figure 1 A schematic diagram of a construction system of a conversational robot provided in an embodiment of the present application is shown. A construction method of a conversational robot in one embodiment of the present application can be applied to Figure 1 The construction system of the conversational robot shown.

[0031] Specifically, refer to Figure 1 As shown, the constructed system includes: a target user terminal 110 corresponding to the target user, a digital employee platform 101, a cloud assistant management platform 102, and a cloud assistant application 103; wherein, business personnel within the enterprise (i.e., target users) can use their own account and password to log in to the cloud assistant management platform 102 through the target user terminal 110, and create a conversation robot suitable for their own specific business scenarios in the cloud assistant management platform 102;

[0032] The business personnel can also use their own account and password to log in and access the digital employee platform 101 through the target user terminal 110. After logging in and accessing the digital employee platform 101, they can view the conversational robot previously created in the cloud assistant management platform 102 in the user interface, and add one or more knowledge bases that meet the above-mentioned specific business scenarios to the created conversational robot in the digital employee platform 101 as the knowledge reserve of the conversational robot;

[0033] In addition, in the zero-code development environment provided by the digital employee platform 101, the business personnel can simply specify the specific functions that the conversational robot needs to have, and control the digital employee platform 101 to automatically generate question-answering logic that matches the specified specific functions. In this way, the business personnel only need to import the added knowledge base and the automatically generated question-answering logic into the NLP (Natural Language Processing) engine (i.e., natural language processing engine) in the digital employee platform 101, and then train in the NLP engine to obtain a conversational model that can answer questions according to the question-answering logic within the knowledge reserve corresponding to the added knowledge base, and then configure the conversational model to the previously created conversational robot to obtain the configured target conversational robot.

[0034] Subsequently, business personnel can add the configured target dialogue robot to the cloud assistant application 103, so that ordinary users (including other users within the enterprise except the above-mentioned business personnel and users outside the enterprise who have registered for the cloud assistant application 103) can also use their own account and password to log in to access the cloud assistant application 103 through their own user terminal 120, and have a conversation interaction experience with the above-mentioned target dialogue robot in the cloud assistant application 103.

[0035] Combine the following Figure 1 The construction system shown in the figure exemplifies each step in the construction method of the conversational robot provided in the embodiment of the present application:

[0036] refer to Figure 2 As shown, Figure 2 A schematic flow chart of a method for constructing a conversational robot provided in an embodiment of the present application is shown, wherein the method for constructing the conversational robot includes steps S201-S204; specifically:

[0037] S201, in response to a target user's login access request to a digital employee platform, displaying a target conversational robot pre-created by the target user in a cloud assistant management platform in a user interface of the target user.

[0038] Here, reference Figure 1The target users can be employees within the enterprise. Figure 1 The target user terminal 110 shown can represent the terminal device held by the above-mentioned target user; wherein, the digital employee platform 101 supports multi-channel access methods, and the target user can log in and access the digital employee platform 101 on the target user terminal 110 through various access methods such as Web interface, mobile App (Application), and mini-programs, so as to meet the usage habits and needs of different users and improve user experience.

[0039] Specifically, in response to receiving a login access request sent by the target user terminal 110 (i.e., a login access request of the target user to the digital employee platform), the digital employee platform 101 can perform data transmission with the target user terminal 110, so that the target user terminal 110 can display the user interface of the target user (i.e., the user interface of the target user in the digital employee platform 101) on the target user terminal 110 after receiving the platform data transmitted by the digital employee platform 101; wherein, based on the uniqueness of the user identity of the target user (i.e., the login account bound by the target user in the digital employee platform 101 and the cloud assistant management platform 102 is the same, wherein the login account can be a mobile phone number or a corporate email address) and the digital employee platform 101 and the cloud assistant management platform 102 belong to the same construction system (i.e., a communication connection is established between the digital employee platform 101 and the cloud assistant management platform 102), therefore, after entering the digital employee platform 101, the target user can see the target dialogue robot previously created in the cloud assistant management platform 102 in his own user interface (i.e., the target dialogue robot previously created by the target user in the cloud assistant management platform 102 is displayed in the user interface).

[0040] S202: In response to the target user issuing an instruction for adding a knowledge base to the target conversational robot in the digital employee platform, configuring the knowledge base included in the instruction for adding the knowledge base as a question-and-answer knowledge base of the target conversational robot.

[0041] Here, for the target user, the cloud assistant management platform 102 and the digital employee platform 101 are two different management platforms in the construction system; wherein, the target user can create a dialogue robot by himself in the cloud assistant management platform 102, and the created dialogue robot (i.e., the above-mentioned target dialogue robot) can be synchronously displayed in the user interface of the target user in the digital employee platform 101, and the target user can manage and configure the open capabilities of the created target dialogue robot in the digital employee platform 101 (including but not limited to: adding knowledge base, constructing question and answer logic, configuring dialogue model, etc.).

[0042] Specifically, a knowledge base management system is also deployed in the digital employee platform 101. The target user can click on the interface entry corresponding to the above knowledge base management system in the user interface. The target user terminal 110 can jump to the knowledge base management page corresponding to the above knowledge base management system in response to the target user's click operation on the above interface entry. The target user can select one or more existing knowledge bases matching the actual business needs from the knowledge base management page as the question and answer knowledge base of the above target dialogue robot according to actual business needs (that is, the above knowledge base addition instruction can be a selection operation for the existing knowledge base displayed on the knowledge base management page); the target user can also import other knowledge bases from an external platform (that is, other data platforms other than the digital employee platform 101) as a newly added knowledge base, and configure the imported newly added knowledge base as the question and answer knowledge base of the above target dialogue robot (that is, the above knowledge base addition instruction can also be a new operation for the external imported knowledge base).

[0043] Exemplary description, Figure 3 A schematic diagram of an interface interaction of a newly added knowledge base provided in an embodiment of the present application is shown. Figure 3 As shown, the target user clicks the Add Knowledge Base button displayed in the knowledge base management page, and a Add Knowledge Base page 300 will pop up in the current knowledge base management page; the Add Knowledge Base page 300 displays: knowledge base code (the knowledge base code corresponding to the existing knowledge base is automatically sorted and generated, and the target user does not need to fill it in), knowledge base name (mandatory, the target user can name the newly added knowledge base by himself), knowledge base description (used to fill in the description information about the purpose of the newly added knowledge base), knowledge base version (the version number of the newly added knowledge base, if it is necessary to do Version distinction can be filled in, otherwise it is optional), sorting (the arrangement order of the newly added knowledge base in the knowledge base management list), status (the status indicator of whether the newly added knowledge base is enabled), creating a user number (that is, the user number of the target user of the newly added knowledge base in the digital employee platform 101), creating a user name (that is, the user name of the target user of the newly added knowledge base in the digital employee platform 101); after completing the filling and input of various information in the above-mentioned new knowledge base page 300, the target user can click Submit to complete the addition of the knowledge base, and the target user can click Cancel to cancel the addition of the knowledge base.

[0044] S203, in response to the function configuration instruction of the target user for the target dialogue robot in the digital employee platform, automatically generate a question-and-answer logic matching the function configuration instruction according to the function description information and / or function module identifier contained in the function configuration instruction.

[0045] Here, a zero-code development environment is also deployed in the digital employee platform 101. The target user can click on the access entry corresponding to the above zero-code development environment in the user interface. The target user terminal 110 can jump into the zero-code development interface corresponding to the above zero-code development environment in response to the target user's click operation on the above access entry; wherein, unlike the traditional code development method, the target user does not need to have professional programming knowledge, and only needs to instruct the digital employee platform 101 to clearly understand which specific functions the target user needs for the target dialogue robot (i.e., the functions corresponding to the above function module identification and / or function description information) through simple operations in the zero-code development interface. The digital employee platform 101 can then be controlled to automatically generate question-and-answer logic that matches the above function module identification and / or function description information (i.e., the function configuration instructions).

[0046] Specifically, as an optional embodiment, the target user can determine the above function configuration instruction by selecting a function module identifier through the method shown in the following steps a1-a2:

[0047] Step a1: In response to the target user's access operation on the zero-code development interface in the digital employee platform, display the zero-code development interface in the user interface of the target user.

[0048] Here, the target user terminal 110 can display the zero-code development interface in the target user's user interface in response to the target user's access operation to the zero-code development interface in the digital employee platform 101 (e.g., clicking on the access entry corresponding to the above-mentioned zero-code development environment in the user interface).

[0049] Step a2: in response to the target user's selection operation on the target function module identifier in the zero-code development interface, determining the selected target function module identifier as the function module identifier included in the function configuration instruction.

[0050] here, Figure 4 A schematic diagram of a zero-code development interface provided by an embodiment of the present application is shown. Figure 4 As shown, the zero-code development interface 400 displays multiple functional module identifiers (i.e. Figure 4 The graphs shown, weather forecast, check-in statistics, visit plans, commissions, commission rates, OCR recognition, smart recommendations, etc.), where each functional module identifier corresponds to a robot function (e.g., smart recommendation is used to indicate the function of automatically pushing information to users when they log in to visit daily, and weather forecast is used to indicate the function of supporting users to query the weather).

[0051] Specifically, the target functional module identifier represents the functional module identifier selected by the target user, such as Figure 4 As shown, the target user can select the function module identifier (i.e., the target function module identifier) ​​contained in the function configuration instruction from the above-mentioned multiple function module identifiers displayed in the zero-code development interface 400, wherein the selected function module identifier can carry a "subscribed" mark, and the unselected function module identifier can carry an "unsubscribed" mark for distinction; in addition, the target user can also search for function module identifiers that match the input keywords in the zero-code development interface 400 by entering keywords according to actual business needs.

[0052] It should be noted that the above-mentioned selection operation may be a click operation, a drag operation, etc. on the function module logo. The embodiment of the present application does not impose any limitation on the specific operation type of the above-mentioned selection operation.

[0053] Specifically, as another optional embodiment, the target user may also determine the above function configuration instruction by inputting function description information through the method shown in the following steps b1-b2:

[0054] Step b1: In response to the target user's access operation on the zero-code development interface in the digital employee platform, display the zero-code development interface in the user interface of the target user.

[0055] Here, the specific implementation of step b1 is the same as that of the above step a1, and the repeated parts are not repeated here.

[0056] Step b2: In response to the target user's input operation on the target input box in the zero-code development interface, determine the text information input by the target user in the target input box as the function description information included in the function configuration instruction.

[0057] Specifically, Figure 4 As shown, in addition to selecting a function module identifier in the zero-code development interface 400, the target user can also choose to input text information used to describe a specific robot function in the target input box displayed in the zero-code development interface 400 through text input, so that the digital employee platform 101 can obtain the above text information input by the target user as the function description information included in the function configuration instruction.

[0058] S204, in response to the target user importing the question-and-answer knowledge base and the question-and-answer logic into the natural language processing engine in the digital employee platform, the original dialogue model deployed in the natural language processing engine is trained according to the imported question-and-answer knowledge base and the question-and-answer logic, and the trained original dialogue model is configured as the dialogue model of the target dialogue robot, thereby obtaining a configured target dialogue robot.

[0059] Here, the above-mentioned original dialogue model can be a large language model, or it can be an AI (Artificial Intelligence) model or other model with natural language processing capabilities. The embodiment of the present application does not impose any limitation on the specific model structure of the above-mentioned original dialogue model.

[0060] In the NLP engine (i.e., the above-mentioned natural language processing engine), by importing the above-mentioned question-and-answer logic and the above-mentioned question-and-answer knowledge base, based on deep learning technology, the above-mentioned original dialogue model can be trained to learn to output reply information that matches the input question information and conforms to the above-mentioned question-and-answer logic within the knowledge reserve corresponding to the above-mentioned question-and-answer knowledge base; wherein, the question information input in the model training stage can come from the above-mentioned imported question-and-answer knowledge base.

[0061] Based on the construction method shown in the above steps S201-S204, before executing the above step S201, the target user can also create the above target conversation robot in the cloud assistant management platform 102 according to the method shown in the following steps c1-c2, specifically:

[0062] Step c1: In response to the target user's robot creation request in the cloud assistant management platform, a robot creation page is displayed in the target user's user interface.

[0063] Here, the target user can use his own account and password to log in to the cloud assistant management platform 102 through the target user terminal 110, and the corresponding user interface of the target user in the cloud assistant management platform 102 will be displayed on the target user terminal 110; wherein, the user interface displays a creation button for the dialogue robot.

[0064] Specifically, the target user terminal 110 can display a robot creation page in the user interface of the target user in the cloud assistant management platform 102 in response to the target user's click operation on the above-mentioned creation button (i.e., the target user's robot creation request in the cloud assistant management platform 102).

[0065] Step c2: According to the multiple attribute parameters selected by the target user in the robot creation page, automatically create a target conversation robot that matches the multiple attribute parameters.

[0066] Here, the above-mentioned multiple attribute parameters include but are not limited to: robot type (used to determine the specific business scenario of the subsequent application of the conversational robot created this time), question and answer type (used to determine the specific model training method corresponding to the conversational model configured corresponding to the robot created this time in the natural language processing engine in the future), and open channels (used to determine the specific channels in which the conversational robot created this time can be put into use in the future).

[0067] Specifically, Figure 5 A schematic diagram of a robot creation page provided in an embodiment of the present application is shown. Figure 5 As shown, the robot creation page 500 displays the following multiple attribute parameters for the target user to select and configure: robot icon (the target user can upload a robot avatar corresponding to the conversation robot created this time), robot name (the target user can name the conversation robot created this time), robot age (the target user can customize the age information of the conversation robot created this time), robot type (individual chat means that the conversation robot created this time can provide one-to-one conversation services for a specific user in the future, and group chat means that the conversation robot created this time can provide one-to-many conversation services for all users in a user group), question and answer type (knowledge base robot means that when the original conversation model is trained in the natural language processing engine in the future, multiple question and answer pairs extracted from the imported question and answer knowledge base can be used for model training, and document robot means that when the original conversation model is trained in the natural language processing engine in the future, the entire document contained in the imported question and answer knowledge base can be used for model training), open channel (cloud assistant application means that the conversation robot created this time supports subsequent addition to cloud assistant application, and other application means that the conversation robot created this time also supports subsequent addition to other applications except cloud assistant application).

[0068] Here, referring to the above description of “Question and Answer Type”, it can be known that the target dialogue robots of different question and answer types have different model training methods in the natural language processing engine. The specific model training methods corresponding to the target dialogue robots of different question and answer types in the above step S204 are described in detail below:

[0069] Specifically, when the question and answer type of the target dialogue robot belongs to the knowledge base type, the above step S204 can be performed according to the method shown in the following steps d1-d3:

[0070] Step d1: When the question and answer type of the target dialogue robot belongs to the knowledge base type, multiple question and answer pairs are extracted from the document data contained in the question and answer knowledge base as candidate dialogue training data.

[0071] Here, if the target user creates the target conversational robot Figure 5 By selecting "knowledge base robot" as the question and answer type in the robot creation page 500 shown, it can be determined that the question and answer type of the target dialogue robot belongs to the knowledge base type. At this time, before training the original dialogue model in the natural language processing engine, the natural language processing engine can extract multiple question and answer pairs as candidate dialogue training data from the document data (i.e., multiple entire documents) contained in the imported question and answer knowledge base.

[0072] It should be noted that each of the above question-answer pairs consists of a set of matching question information and reply information; the embodiment of the present application does not impose any limitation on the specific number of question-answer pairs extracted.

[0073] Step d2: for each question-answer pair, input the question information in the question-answer pair into the original dialogue model to obtain predicted response information for the question information output by the original dialogue model.

[0074] Specifically, for each question-answer pair extracted, the question information in the question-answer pair is input into the original dialogue model, and the reply information corresponding to the input question information is predicted by the original dialogue model, so that the predicted reply information for the above question information output by the original dialogue model can be obtained.

[0075] Step d3: train the original dialogue model according to the semantic loss between the predicted reply information and the reply information in the question-answer pair and the logical loss between the question information in the question-answer pair and the predicted reply information and the question-answer logic until the original dialogue model reaches convergence.

[0076] Here, the reply information in the question-answer pair is equivalent to the real reply information corresponding to the above-mentioned question information. Therefore, according to the semantic loss between the above-mentioned predicted reply information and the reply information in the question-answer pair, the original dialogue model can be trained to learn how to output semantically accurate reply information (that is, semantically close to the above-mentioned real reply information); and the question information in the question-answer pair and the above-mentioned predicted reply information output by the original dialogue model constitute the question-answer pair prediction result corresponding to this model prediction. According to the logical loss between the question-answer pair prediction result and the imported question-answer logic (that is, whether the above-mentioned question-answer pair prediction result conforms to the imported question-answer logic), the original dialogue model can be trained to learn how to output reply information with accurate question-answer logic (that is, conforming to the imported question-answer logic); thus, based on the above-mentioned two aspects of model training methods, the above-mentioned original dialogue model is trained to learn to output reply information that matches the input question information and conforms to the above-mentioned question-answer logic within the knowledge reserve corresponding to the above-mentioned question-answer knowledge base.

[0077] Specifically, when the question and answer type of the target dialogue robot is a document type, the step S204 may be performed according to the method shown in the following steps e1-e3:

[0078] Step e1: When the question and answer type of the target dialogue robot belongs to the document type, the document data contained in the question and answer knowledge base is used as training sample data.

[0079] Here, if the target user creates the target conversational robot Figure 5 By selecting "document robot" as the question and answer type in the robot creation page 500 shown, it can be determined that the question and answer type of the target dialogue robot belongs to the document type. At this time, when training the original dialogue model in the natural language processing engine, the natural language processing engine can directly use the document data (i.e., multiple entire documents) contained in the imported question and answer knowledge base as training sample data.

[0080] Step e2: for each of the training sample data, input the training sample data into the original dialogue model, randomly extract question information from the training sample data as question sample information through the original dialogue model, and obtain predicted response sample information for the question sample information output by the original dialogue model.

[0081] Specifically, for each entire document data (i.e., each of the training sample data), when the entire document data is input into the original dialogue model, the original dialogue model can randomly extract a question information from the input entire document data as question sample information, and predict the reply information corresponding to the question sample information, so as to obtain the predicted reply information for the above question sample information output by the original dialogue model.

[0082] Step e3: Train the original dialogue model according to the semantic loss between the predicted response sample information and the real response information corresponding to the question sample information in the training sample data, and the logical loss between the question sample information and the predicted response sample information and the question-answering logic, until the original dialogue model converges.

[0083] Here, in the input training sample data, the reply information located behind the question sample information can be determined as the real reply information corresponding to the question sample information. Therefore, according to the semantic loss between the predicted reply information and the real reply information, the original dialogue model can be trained to learn how to output semantically accurate reply information (i.e., semantically close to the real reply information); and the question sample information and the predicted reply information output by the original dialogue model constitute the question and answer pair prediction result corresponding to this model prediction. According to the logical loss between the question and answer pair prediction result and the imported question and answer logic (i.e., whether the question and answer pair prediction result conforms to the imported question and answer logic), the original dialogue model can be trained to learn how to output reply information with accurate question and answer logic (i.e., conforming to the imported question and answer logic); thus, based on the above two aspects of model training methods, the original dialogue model is trained to learn to output reply information that matches the input question information and conforms to the above question and answer logic within the knowledge reserve corresponding to the above question and answer knowledge base.

[0084] After executing the above steps S201-S204 to obtain the configured target conversation robot, taking the open channel of the target conversation robot as the cloud assistant application 103 as an example, the target user can also open the configured target conversation robot to the target user group in the cloud assistant application 103 according to the method shown in the following steps f1-f3, specifically:

[0085] Step f1: In response to the target user configuring the target conversation robot in the digital employee platform to be open to the target user group in the cloud assistant application, user question information in the target user group is received from the cloud assistant application.

[0086] Here, reference Figure 5 As shown, in the cloud assistant management platform 102, the target user can select the cloud assistant application as the open channel of the target dialogue robot in the open channel of the target dialogue robot, and in the digital employee platform 101, the target user can configure a callback address for the target dialogue robot, wherein the callback address is used to determine the specific open object and open area corresponding to the target dialogue robot (equivalent to specifying the specific service object and service area of ​​the target dialogue robot according to the callback address).

[0087] Specifically, when executing the above step f1, the target user can configure the callback address of the target conversational robot in the digital employee platform 101 according to the specific group position of the target user group in the cloud assistant application 103, and copy the configured callback address, and synchronously configure the callback address of the target conversational robot in the cloud assistant management platform 102, so that the target conversational robot is configured to be open to the target user group in the cloud assistant application 103.

[0088] Step f2: input the user question information into the dialogue model configured by the target dialogue robot, and obtain the robot reply information of the dialogue model to the user question information.

[0089] Here, when user question information is detected in the target user group (i.e., user question information in the target user group is received), the user question information can be input into the dialogue model configured for the target dialogue robot (i.e., the trained original dialogue model) to obtain the robot response information of the above dialogue model for the above user question information.

[0090] Step f3: Send the robot reply information to the target user group through the target dialogue robot.

[0091] Here, the above robot reply information can be sent to the target user group through the target dialogue robot, so that users in the target user group can receive the robot reply information of the target dialogue robot and experience the process of dialogue interaction with the target dialogue robot.

[0092] Based on the construction method of the above-mentioned conversational robot provided in the embodiment of the present application, by creating a digital employee platform as a visual conversational robot development low-code platform, ordinary business personnel can also log in to access the digital employee platform, and easily and quickly construct a conversational robot that meets their own business needs in the digital employee platform, thereby helping enterprises to achieve high-quality digital operations.

[0093] Based on the same inventive concept, Figure 1 As shown, the present application also provides a construction system of a conversational robot corresponding to the construction method of the above-mentioned conversational robot. Since the principle of solving the problem by the construction system in the embodiment of the present application is similar to the above-mentioned construction method in the embodiment of the present application, the implementation of the construction system can refer to the implementation of the above-mentioned construction method, and the repeated parts will not be repeated.

[0094] like Figure 1 As shown, the construction system includes: a target user terminal 110 corresponding to the target user, a digital employee platform 101, a cloud assistant management platform 102 and a cloud assistant application 103; wherein the target user terminal 110 is used to:

[0095] In response to the target user (corresponding to Figure 1 The target user terminal 110 shown in the figure) responds to the login access request of the digital employee platform 101, and displays the target conversation robot created in advance by the target user in the cloud assistant management platform 102 in the user interface of the target user;

[0096] In response to the target user adding an instruction to the knowledge base of the target conversational robot in the digital employee platform 101, configuring the knowledge base included in the knowledge base adding instruction as a question-and-answer knowledge base of the target conversational robot;

[0097] In response to the function configuration instruction of the target user for the target dialogue robot in the digital employee platform 101, automatically generate a question-answering logic matching the function configuration instruction according to the function description information and / or function module identifier contained in the function configuration instruction;

[0098] In response to the target user importing the question-and-answer knowledge base and the question-and-answer logic into the natural language processing engine in the digital employee platform 101, the original dialogue model deployed in the natural language processing engine is trained according to the imported question-and-answer knowledge base and the question-and-answer logic, and the trained original dialogue model is configured as the dialogue model of the target dialogue robot to obtain a configured target dialogue robot.

[0099] In an optional implementation, the target user terminal 110 is used to create the target conversational robot in the cloud assistant management platform by the following method:

[0100] In response to a robot creation request from the target user in the cloud assistant management platform 102 , displaying a robot creation page in the user interface of the target user;

[0101] According to the multiple attribute parameters selected by the target user in the robot creation page, a target conversation robot matching the multiple attribute parameters is automatically created; wherein the multiple attribute parameters include: robot type, question and answer type, and open channels.

[0102] In an optional implementation manner, the target user terminal 110 is configured to determine the function configuration instruction by:

[0103] In response to an access operation of the target user on the zero-code development interface in the digital employee platform 101, displaying the zero-code development interface in the user interface of the target user;

[0104] In response to the target user's selection operation on the target function module identifier in the zero-code development interface, the selected target function module identifier is determined as the function module identifier included in the function configuration instruction.

[0105] In an optional implementation manner, the target user terminal 110 is configured to determine the function configuration instruction by:

[0106] In response to an access operation of the target user on the zero-code development interface in the digital employee platform 101, displaying the zero-code development interface in the user interface of the target user;

[0107] In response to an input operation of the target user on a target input box in the zero-code development interface, the text information input by the target user in the target input box is determined as the function description information included in the function configuration instruction.

[0108] In an optional implementation, when the original dialogue model deployed in the natural language processing engine is trained according to the imported question-answer knowledge base and the question-answer logic, the target user terminal 110 is used to:

[0109] When the question-answering type of the target conversational robot belongs to the knowledge base type, extracting a plurality of question-answering pairs from the document data contained in the question-answering knowledge base as candidate conversational training data; wherein the question-answering pairs are composed of a set of matching question information and reply information;

[0110] For each question-answer pair, input the question information in the question-answer pair into the original dialogue model to obtain predicted response information for the question information output by the original dialogue model;

[0111] The original dialogue model is trained according to the semantic loss between the predicted reply information and the reply information in the question-answer pair and the logical loss between the question information in the question-answer pair and the predicted reply information and the question-answer logic until the original dialogue model reaches convergence.

[0112] In an optional implementation, when the original dialogue model deployed in the natural language processing engine is trained according to the imported question-answer knowledge base and the question-answer logic, the target user terminal 110 is used to:

[0113] When the question and answer type of the target conversational robot belongs to the document type, the document data contained in the question and answer knowledge base is used as training sample data;

[0114] For each of the training sample data, the training sample data is input into the original dialogue model, question information is randomly extracted from the training sample data as question sample information through the original dialogue model, and predicted response sample information for the question sample information output by the original dialogue model is obtained;

[0115] The original dialogue model is trained according to the semantic loss between the predicted response sample information and the real response information corresponding to the question sample information in the training sample data and the logical loss between the question sample information and the predicted response sample information and the question-answering logic until the original dialogue model reaches convergence.

[0116] In an optional implementation, after obtaining the configured target dialogue robot, the target user terminal 110 is further used to:

[0117] In response to the target user configuring the target conversation robot to be open to the target user group in the cloud assistant application 103 in the digital employee platform 101, receiving user question information in the target user group from the cloud assistant application 103;

[0118] Inputting the user question information into the dialogue model configured by the target dialogue robot, and obtaining robot reply information of the dialogue model to the user question information;

[0119] The robot reply information is sent to the target user group through the target dialogue robot.

[0120] Based on the same inventive concept, the present application also provides an electronic device corresponding to the construction method of the above-mentioned conversational robot. Since the principle of solving the problem by the electronic device in the embodiment of the present application is similar to the above-mentioned construction method in the embodiment of the present application, the implementation of the electronic device can refer to the implementation of the above-mentioned construction method, and the repeated parts will not be repeated.

[0121] like Figure 6 As shown, an embodiment of the present application provides an electronic device 600 for executing the method for constructing a conversational robot in the present application, wherein the electronic device 600 includes a memory 601, a processor 602, and a computer program stored in the memory 601 and executable on the processor 602, wherein the memory 601 and the processor 602 are communicatively connected via a bus, and the processor 602 implements the steps of the method for constructing a conversational robot when executing the computer program.

[0122] Specifically, the above-mentioned memory 601 and processor 602 can be general-purpose memory and processor, which are not specifically limited here. When the processor 602 runs the computer program stored in the memory 601, it can execute the above-mentioned method for constructing the conversational robot.

[0123] Corresponding to the construction method of the conversational robot in the present application, an embodiment of the present application also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is run by a processor, the steps of the construction method of the above-mentioned conversational robot are executed.

[0124] Specifically, the storage medium can be a general storage medium, such as a mobile disk, a hard disk, etc. When the computer program on the storage medium is run, the above-mentioned method for constructing a conversational robot can be executed.

[0125] In the embodiments provided in the present application, it should be understood that the disclosed systems and methods can be implemented in other ways. The system embodiments described above are merely schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, and the indirect coupling or communication connection of the system or unit can be electrical, mechanical or other forms.

[0126] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0127] In addition, each functional unit in the embodiments provided in the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0128] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0129] It should be noted that similar numbers and letters represent similar items in the following figures. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. In addition, the terms "first", "second", "third", etc. are only used to distinguish the description and are not to be understood as indicating or implying relative importance.

[0130] Finally, it should be noted that the above-described embodiments are only specific implementation methods of the present application, which are used to illustrate the technical solutions of the present application, rather than to limit them. The protection scope of the present application is not limited thereto. Although the present application is described in detail with reference to the above-mentioned embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above-mentioned embodiments within the technical scope disclosed in the present application, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application. They should all be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. A method for constructing a conversational robot, characterized in that: The target user terminal is applied to the construction system, wherein the construction system further comprises: a digital employee platform, a cloud assistant management platform and a cloud assistant application; wherein the construction method comprises: In response to a target user's login access request to the digital employee platform, displaying a target conversation robot pre-created by the target user in the cloud assistant management platform in the user interface of the target user; In response to the target user adding an instruction to a knowledge base of the target conversational robot in the digital employee platform, configuring the knowledge base included in the knowledge base adding instruction as a question-and-answer knowledge base of the target conversational robot; In response to a function configuration instruction of the target user for the target conversational robot in the digital employee platform, automatically generating a question-and-answer logic matching the function configuration instruction according to function description information and / or function module identifier contained in the function configuration instruction; In response to the target user importing the question-and-answer knowledge base and the question-and-answer logic into the natural language processing engine in the digital employee platform, the original dialogue model deployed in the natural language processing engine is trained according to the imported question-and-answer knowledge base and the question-and-answer logic, and the trained original dialogue model is configured as the dialogue model of the target dialogue robot to obtain a configured target dialogue robot.

2. The construction method according to claim 1, characterized in that: The target conversational robot is created in the cloud assistant management platform by the following method: In response to a robot creation request by the target user in the cloud assistant management platform, displaying a robot creation page in a user interface of the target user; According to the multiple attribute parameters selected by the target user in the robot creation page, a target conversation robot matching the multiple attribute parameters is automatically created; wherein the multiple attribute parameters include: robot type, question and answer type, and open channels.

3. The construction method according to claim 1, characterized in that: The function configuration instruction is determined by: In response to an access operation of the target user on the zero-code development interface in the digital employee platform, displaying the zero-code development interface in the user interface of the target user; In response to the target user's selection operation on the target function module identifier in the zero-code development interface, the selected target function module identifier is determined as the function module identifier included in the function configuration instruction.

4. The construction method according to claim 1, characterized in that: The function configuration instruction is determined by: In response to an access operation of the target user on the zero-code development interface in the digital employee platform, displaying the zero-code development interface in the user interface of the target user; In response to an input operation of the target user on a target input box in the zero-code development interface, the text information input by the target user in the target input box is determined as the function description information included in the function configuration instruction.

5. The construction method according to claim 1, characterized in that: The training of the original dialogue model deployed in the natural language processing engine according to the imported question-answer knowledge base and the question-answer logic includes: When the question-answering type of the target conversational robot belongs to the knowledge base type, extracting a plurality of question-answering pairs from the document data contained in the question-answering knowledge base as candidate conversational training data; wherein the question-answering pairs are composed of a set of matching question information and reply information; For each question-answer pair, input the question information in the question-answer pair into the original dialogue model to obtain predicted response information for the question information output by the original dialogue model; The original dialogue model is trained according to the semantic loss between the predicted reply information and the reply information in the question-answer pair and the logical loss between the question information in the question-answer pair and the predicted reply information and the question-answer logic until the original dialogue model reaches convergence.

6. The construction method according to claim 1, characterized in that: The training of the original dialogue model deployed in the natural language processing engine according to the imported question-answer knowledge base and the question-answer logic includes: When the question and answer type of the target conversational robot belongs to the document type, the document data contained in the question and answer knowledge base is used as training sample data; For each of the training sample data, the training sample data is input into the original dialogue model, question information is randomly extracted from the training sample data as question sample information through the original dialogue model, and predicted response sample information for the question sample information output by the original dialogue model is obtained; The original dialogue model is trained according to the semantic loss between the predicted response sample information and the real response information corresponding to the question sample information in the training sample data and the logical loss between the question sample information and the predicted response sample information and the question-answering logic until the original dialogue model reaches convergence.

7. The construction method according to claim 1, characterized in that: After obtaining the configured target dialogue robot, the construction method further includes: In response to the target user configuring the target conversation robot to be open to the target user group in the cloud assistant application in the digital employee platform, receiving user question information in the target user group from the cloud assistant application; Inputting the user question information into the dialogue model configured by the target dialogue robot, and obtaining robot reply information of the dialogue model to the user question information; The robot reply information is sent to the target user group through the target dialogue robot.

8. A system for constructing a conversational robot, characterized in that: The construction system includes: a target user terminal corresponding to the target user, a digital employee platform, a cloud assistant management platform and a cloud assistant application; wherein the target user terminal is used to: In response to a target user's login access request to the digital employee platform, displaying a target conversation robot pre-created by the target user in the cloud assistant management platform in the user interface of the target user; In response to the target user adding an instruction to a knowledge base of the target conversational robot in the digital employee platform, configuring the knowledge base included in the knowledge base adding instruction as a question-and-answer knowledge base of the target conversational robot; In response to a function configuration instruction of the target user for the target conversational robot in the digital employee platform, automatically generating a question-and-answer logic matching the function configuration instruction according to function description information and / or function module identifier contained in the function configuration instruction; In response to the target user importing the question-and-answer knowledge base and the question-and-answer logic into the natural language processing engine in the digital employee platform, the original dialogue model deployed in the natural language processing engine is trained according to the imported question-and-answer knowledge base and the question-and-answer logic, and the trained original dialogue model is configured as the dialogue model of the target dialogue robot to obtain a configured target dialogue robot.

9. An electronic device, characterized in that: include: A processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor and the memory communicate via the bus, and when the machine-readable instructions are executed by the processor, the steps of the method for constructing a conversational robot as described in any one of claims 1 to 7 are performed.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which, when executed by a processor, executes the steps of the method for constructing a conversational robot as described in any one of claims 1 to 7.

Images