Dialogue process processing method and device, equipment, medium and program product

Abstract

The invention provides a dialogue process processing method and device, equipment, a medium and a program product. The dialogue process processing method comprises the steps that in response to a first node selection instruction, multiple selected first process nodes are displayed in a canvas, and the multiple first process nodes comprise at least one of knowledge base nodes used for calling a knowledge base and model nodes used for calling a language model; in response to a first node connection instruction, displaying a connection relationship of the plurality of first process nodes in the canvas; according to the connection relationship of the plurality of first process nodes, generating a dialogue process; and responding to the received first dialogue information, and generating reply information of the first dialogue information based on the dialogue process. The development process of the dialogue process is simplified, and the accuracy of the reply information generated in the dialogue process is improved.

Description

Technical Field

[0001] The present disclosure relates to the field of computer technology, and in particular to a dialog flow processing method, device, equipment, medium and program product. Background Art

[0002] With the development of computer and artificial intelligence technology, the function of computer-user dialogue has been widely used in customer service, voice assistant, intelligent chat and other scenarios. In related technologies, in order to realize the above dialogue function, developers need to write a lot of code for different requirements of the dialogue function, which is complicated and time-consuming, resulting in high development costs. Summary of the invention

[0003] The present disclosure provides a dialogue process processing method, apparatus, device, medium and program product. The embodiments of the present application simplify the development process of the dialogue process and improve the accuracy of the reply information generated during the dialogue process.

[0004] According to a first aspect of the present disclosure, a method for processing a dialog flow is provided, the method comprising: in response to a first node selection instruction, displaying a plurality of selected first flow nodes in a canvas, wherein the plurality of first flow nodes include at least one of a knowledge base node for calling a knowledge base and a model node for calling a language model; in response to a first node connection instruction, displaying a connection relationship between the plurality of first flow nodes in the canvas; generating a dialog flow according to the connection relationship between the plurality of first flow nodes; and in response to receiving a first dialog message, generating reply information for the first dialog message based on the dialog flow.

[0005] According to a second aspect of the present disclosure, there is provided a dialog process processing device, the device comprising: a node selection module, configured to display a plurality of selected first process nodes in a canvas in response to a first node selection instruction, wherein the plurality of first process nodes include at least one of a knowledge base node for calling a knowledge base and a model node for calling a language model; a node connection module, configured to display a connection relationship between the plurality of first process nodes in the canvas in response to the first node connection instruction; a business process generation module, configured to generate a dialog process according to the connection relationship between the plurality of first process nodes; and a business process execution module, configured to generate reply information for the first dialog information based on the dialog process in response to receiving the first dialog information.

[0006] According to a third aspect of the present disclosure, an electronic device is provided, comprising: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to execute the dialog flow processing method of the first aspect and its possible implementation method by executing the executable instructions.

[0007] According to a fourth aspect of the present disclosure, a storage medium is provided, in which a computer program is stored. When the computer program is executed by a processor, the dialog flow processing method of the first aspect and possible implementation methods thereof are implemented.

[0008] According to a fifth aspect of the present disclosure, a computer program product is provided, including a computer program, wherein when the computer program is executed by a processor, the dialog flow processing method of the first aspect and possible implementation methods thereof are implemented.

[0009] The technical solution disclosed in this disclosure has the following beneficial effects:

[0010] The embodiment of the present application provides a dialogue processing flow. By selecting a first process node on a canvas and establishing a connection relationship between the selected multiple first process nodes, a dialogue flow composed of first process nodes can be generated, and reply information for the first dialogue information can be generated through the dialogue flow. Since the entire process of establishing a dialogue flow based on the canvas is visualized, the user only needs to select the process node without writing a large amount of code, which simplifies the development process of the dialogue function, reduces the development time, and thus improves the efficiency of dialogue processing. In addition, the first process node provided by the canvas includes a knowledge base node for calling a knowledge base and a model node for calling a language model. Therefore, when the generated dialogue flow processes the first dialogue information, it can combine the knowledge base and the language model to analyze and understand the first dialogue information, so that the generated reply information can be more compatible with the first dialogue information, and can reply to the first dialogue information more accurately. In addition, the visual editing and development method based on the canvas does not require developers to master professional code-related knowledge, which reduces the technical threshold for developers and has a wider scope of application. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 A flowchart showing a method for processing a dialog flow in this exemplary embodiment;

[0012] FIG. 2A to FIG. 2D A schematic diagram showing a dialog flow in this exemplary embodiment;

[0013] Figure 3 A flowchart of obtaining second dialogue information in this exemplary embodiment is shown;

[0014] Figure 4 A flowchart of building a knowledge base in this exemplary embodiment is shown;

[0015] Figure 5 A schematic diagram showing a knowledge base process in this exemplary embodiment;

[0016] Figure 6A schematic diagram showing a method of implementing a dialogue process in this exemplary embodiment is shown;

[0017] Figure 7 A schematic diagram showing a node editing interface in this exemplary embodiment;

[0018] Figure 8 A schematic diagram showing a single-step debugging in this exemplary embodiment;

[0019] 9A to 9C A schematic diagram showing a setting interface of a process node in this exemplary embodiment;

[0020] Fig.10 A structural block diagram of a dialog flow processing device in this exemplary embodiment is shown;

[0021] Fig.11 An electronic device for implementing the above-mentioned dialog flow processing method in this exemplary embodiment is shown. DETAILED DESCRIPTION

[0022] Exemplary embodiments of the present disclosure will be described more fully hereinafter with reference to the accompanying drawings.

[0023] The accompanying drawings are schematic diagrams of the present disclosure and are not necessarily drawn to scale. Some of the block diagrams shown in the accompanying drawings may be functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software form, or in hardware modules or integrated circuits, or in networks, processors or microcontrollers. The embodiments can be implemented in various forms and should not be construed as being limited to the examples set forth herein. The features, structures or characteristics described in the present disclosure may be combined in one or more embodiments in any suitable manner. In the description below, many specific details are provided to provide a full description of the embodiments of the present disclosure. However, those skilled in the art should appreciate that one or more specific details may be omitted when implementing the technical solution of the present disclosure, or one or more specific details may be replaced by other methods, components, devices, steps, etc.

[0024] In related technologies, in order to realize the dialogue function, developers need to write a lot of code for different aspects of the dialogue function, such as writing relevant code for content parsing, query logic, etc. The whole process is complicated and time-consuming, resulting in high development costs. Due to the differences between different dialogue scenarios, dialogue functions for different scenarios usually need to be developed independently, that is, a complete development process needs to be executed at the code level for each dialogue scenario, further increasing the complexity and development cost of the development process. In addition, if the dialogue scenario requirements change, developers need to modify the underlying code and redeploy related services to implement updates, which increases workload and maintenance costs.

[0025] In view of one or more of the above problems, an exemplary embodiment of the present disclosure provides a dialog flow processing method.

[0026] Figure 1 An exemplary process of a dialog flow processing method is shown, which may include the following steps:

[0027] Step S110, in response to the first node selection instruction, displaying a plurality of selected first process nodes in the canvas, wherein the plurality of first process nodes include at least one of a knowledge base node for calling a knowledge base and a model node for calling a language model;

[0028] Step S120, in response to the first node connection instruction, displaying the connection relationship of the plurality of first process nodes in the canvas;

[0029] Step S130, generating a dialog flow according to the connection relationship of the plurality of first flow nodes;

[0030] Step S140: in response to receiving the first dialogue information, generating reply information of the first dialogue information based on the dialogue process.

[0031] exist Figure 1 In the method, by selecting a first process node on the canvas and establishing a connection relationship between the selected multiple first process nodes, a dialogue process composed of the first process nodes can be generated, and reply information for the first dialogue information can be generated through the dialogue process. Since the entire process of establishing a dialogue process based on the canvas is visual, the user only needs to select the process node, and there is no need to write a lot of code, which simplifies the development process of the dialogue function, reduces the development time, and thus improves the efficiency of dialogue processing. In addition, the first process node provided by the canvas includes a knowledge base node for calling the knowledge base and a model node for calling the language model. Therefore, when the generated dialogue process processes the first dialogue information, the knowledge base and the language model can be combined to analyze and understand the first dialogue information, so that the generated reply information can be more compatible with the first dialogue information, and can reply to the first dialogue information more accurately. In addition, the visual editing and development method based on the canvas does not require developers to master professional code-related knowledge, which reduces the technical threshold for developers and has a wider scope of application.

[0032] refer to Figure 1 In step S110, in response to the first node selection instruction, a plurality of selected first process nodes are displayed in the canvas, wherein the plurality of first process nodes include at least one of a knowledge base node for calling a knowledge base and a model node for calling a language model.

[0033] The canvas is a visual business process design interface. The canvas can provide different process nodes. A process node is essentially a business process instruction, representing a step or unit in the business process. The process nodes are displayed in the form of visual nodes in the canvas, which is convenient for users to edit.

[0034] The user can edit the dialogue flow in a canvas manner, select the flow node required for the dialogue flow in the flow node, the flow node selected by the user for editing the dialogue flow is called the first flow node, and the operation instruction of the user selecting the first flow node is called the first node selection instruction. Exemplarily, the flow node bar of the canvas provides all available flow nodes, and the user can select from these flow nodes when designing the dialogue flow, such as dragging the required flow node to the canvas, thereby determining the first flow node, and displaying the first flow node in the canvas.

[0035] To implement the dialogue process, the user can select multiple first process nodes, which include at least one of a knowledge base node and a model node. The knowledge base node is used to call the knowledge base, such as calling the knowledge base related to the dialogue scenario to obtain the knowledge information required for the dialogue. The model node is used to call the language model, which is a machine learning model for processing natural language or structured language, such as LLM (Large Language Model), which can be a pre-trained model or a model that has been fine-tuned on the dialogue scenario.

[0036] In one embodiment, the first process node may also include but is not limited to one or more of the following process nodes: dialogue input node, knowledge information retrieval node, multi-way recall node, query similar dialogue node, tracing node, chunk expansion node, dialogue output node, etc.

[0037] A dialogue input node is a functional unit or module that receives user input information. It is the starting point of the dialogue process and is responsible for obtaining the content passed in by the user through the text input box (in the text dialogue scenario) or the voice input device (in the voice dialogue scenario), providing raw data for subsequent dialogue processing.

[0038] The knowledge information retrieval node is a key part of a dialogue system or knowledge management system. It is mainly responsible for finding knowledge and information related to user input in the knowledge base or other information sources. The purpose of this node is to quickly and accurately locate content that can help answer user questions or meet user needs through various retrieval strategies and techniques.

[0039] A multi-way recall node is a functional unit used to obtain relevant candidate sets from multiple data sources or multiple recall strategies in the fields of information retrieval and recommendation systems. Its main purpose is to collect candidate items that may be related to user needs as comprehensively as possible, providing rich materials for subsequent sorting and screening.

[0040] The query similarity dialogue node is a functional module used in dialogue systems, question-answering systems, knowledge retrieval systems, etc. to process user queries and find similar existing dialogues. Its main task is to understand user questions (i.e. queries) and find those question-answering contents that are highly similar in semantics, subject or intention in the question-answering knowledge base or related text collection, so as to provide users with accurate and efficient answers.

[0041] The traceability node is a key unit used to track and restore the development process of things or the source of data in many fields such as data processing, knowledge management, and supply chain management. Its main function is to trace the origin of events, the source of data, or the flow of products in the supply chain by recording and analyzing relevant information clues.

[0042] Chunk expansion nodes are components used to expand and enrich the content of specific information blocks in the fields of text processing, knowledge graph construction, information retrieval, etc. It is based on one or more initial information blocks (such as text paragraphs, knowledge entity descriptions, search result fragments, etc.), and adds relevant information in various ways to expand the depth and breadth of these information blocks.

[0043] The dialogue output node is a key part of the dialogue system. It is mainly responsible for presenting the reply content generated by the system to the user in an appropriate form. It is the last link in the interaction between the dialogue system and the user, and its output quality directly affects the user experience.

[0044] Continue to refer Figure 1 In step S120, in response to the first node connection instruction, the connection relationship of multiple first process nodes is displayed in the canvas.

[0045] The first node connection instruction is an operation instruction for connecting the first process node. The user can connect the first process nodes in the canvas, such as pulling a line from a first process node and connecting it to another first process node, or adding a connection line in the canvas and connecting the two ends of the connection line to different first process nodes. According to the user's first node connection instruction, the corresponding connection relationship is displayed in the canvas, thereby editing the order between different first process nodes.

[0046] Continue to refer Figure 1In step S130, a dialog flow is generated according to the connection relationship between the plurality of first flow nodes.

[0047] Through the above-mentioned first node selection instruction and first node connection instruction, the canvas editing for the dialogue process is completed. According to the connection relationship of multiple first process nodes, the first process nodes can be integrated in sequence to form a complete processing flow from the first first process node to the last first process node, that is, the dialogue process. At the program level, the computer can integrate the business process instructions corresponding to these first process nodes according to the connection relationship of multiple first process nodes to form a complete business process instruction for processing dialogues.

[0048] In one implementation, different canvas information may be edited for different business scenarios, including information such as the selected first process node and connection relationship, to generate a dialogue process for different dialogue scenarios.

[0049] Taking text chat or text customer service dialogue scenarios as an example, you can build the following Figure 2A The dialogue process shown in FIG. 1 may include but is not limited to the following first process nodes: dialogue input node, knowledge base node, vector retrieval node, multi-way recall node, model node, dialogue output node, etc. Taking the telephone customer service dialogue scenario as an example, the following can be constructed: Figure 2B The dialogue process shown in the figure may include but is not limited to the following first process nodes: voice input node, voice-to-text node, knowledge base node, vector retrieval node, multi-way recall node, model node, dialogue output node, etc. Optionally, the knowledge base nodes used in the telephone customer service dialogue scenario and the text customer service dialogue scenario can be called against the same knowledge base to reduce the maintenance cost of the knowledge base. Of course, different knowledge bases can also be used to improve the dialogue professionalism of each business scenario. Taking the intelligent operation and maintenance scenario as an example, the following can be constructed: Figure 2C The dialogue flow shown may include but is not limited to the following first flow nodes: dialogue input node, knowledge base node, vector retrieval node, multi-way recall node, model node, operation and maintenance server call node, etc. According to the connection relationship of the first flow nodes, the first flow nodes are integrated to form the dialogue flow of each business scenario.

[0050] FIG. 2A to FIG. 2C The dialog flow shown in the figure is only for illustrative purposes. In specific business scenarios, more complex dialog flows can be created based on actual business development needs, such as Figure 2DThe dialogue flow is shown. The dialog flow includes the following first flow nodes: a dialog input node, which is used to input dialog information to be processed; a query similar dialog node, which is used to query similar dialog information according to the input dialog information; a knowledge base node, which is used to call the knowledge base and obtain instruction information; a full-text retrieval node, which is used to retrieve relevant context according to the input dialog information, the queried similar dialogs and the knowledge information called from the knowledge base; a vector retrieval node, which is used to retrieve knowledge information related to the dialog information by vector matching according to the input dialog information and the knowledge information in the knowledge base; a multi-way recall node, which is used to recall more relevant information according to the knowledge information in the knowledge base, the knowledge information related to the dialog information, the context information, etc.; a block expansion node, which is used to expand the recalled relevant information according to the knowledge information in the knowledge base; a tracing node, which is used to trace and organize the expanded information according to the knowledge information in the knowledge base to make it more organized; a model node, which is used to call a language model to process the prompt information formed by the input dialog information, the knowledge information in the knowledge base, the traced information, etc., to obtain the dialog reply information; and a dialog output node, which is used to output the final complete reply information according to the traced information and the reply information output by the language model. This enables high-quality dialog processing.

[0051] Continue to refer Figure 1 In step S140, in response to receiving the first dialogue information, reply information of the first dialogue information is generated based on the dialogue process.

[0052] The first dialogue information is dialogue information that needs to be replied, such as dialogue information input by users in scenarios such as customer service, voice assistant, and intelligent chat. In response to receiving the first dialogue information, the first dialogue information can be input into the dialogue process, and the first process nodes in the dialogue process are called in sequence for processing according to the order of the first process nodes, and finally the reply information of the first dialogue information is output.

[0053] In one embodiment, reference Figure 3 As shown, the above-mentioned generation of reply information of the first dialogue information based on the dialogue process may include the following steps S310 to S340:

[0054] Step S310: calling a knowledge base through a knowledge base node to obtain first knowledge information from the knowledge base.

[0055] The first knowledge information is knowledge information obtained from the knowledge base, and may be all or part of the knowledge information stored in the knowledge base. For example, the first dialogue information is dialogue information obtained in the first business scenario (such as a customer service scenario related to financial management), and is highly relevant to the first business scenario. The knowledge base is a knowledge base constructed for the first business scenario, and the knowledge information stored therein is all knowledge information highly relevant to the first business scenario. In this case, in step S310, the knowledge base may be called through the knowledge base node to obtain all the knowledge information therein as the first knowledge information. For another example, the first dialogue information is dialogue information obtained in the first business scenario, and is highly relevant to the first business scenario. The knowledge base is a general knowledge base constructed for different business scenarios, and the knowledge information stored therein includes knowledge information related to different business scenarios. In this case, in step S310, the knowledge base may be called through the knowledge base node to obtain knowledge information related to the first business scenario, such as knowledge information under a category related to the first business scenario, or knowledge information retrieved according to keywords of the first business scenario, as the first knowledge information.

[0056] In one implementation, the knowledge base may be pre-built and a knowledge base node may be generated, the knowledge base node including interface information for calling the knowledge base. In step S310, the knowledge base may be called through the interface information, and the first knowledge information may be obtained through a request or the like.

[0057] Step S320: determining second knowledge information matching the first dialogue information from the first knowledge information by matching the first dialogue information with the first knowledge information.

[0058] The second knowledge information may be knowledge information selected from the first knowledge information and semantically related to the first dialogue information. The first dialogue information and the first knowledge information may be converted into structured information, and the structured information may be matched to determine the matching degree, and the first knowledge information with a higher matching degree may be determined as the second knowledge information.

[0059] In one implementation, the above-mentioned matching the first dialogue information with the first knowledge information to determine the second knowledge information matching the first dialogue information from the first knowledge information may include the following steps:

[0060] Extracting features from the first dialogue information to obtain a first feature vector, and extracting features from the first knowledge information to obtain a second feature vector;

[0061] determining a similarity between the first eigenvector and the second eigenvector;

[0062] Second knowledge information matching the first dialogue information is determined from the first knowledge information according to the similarity.

[0063] The first dialogue information may be feature extracted by text encoding or the like to obtain a first feature vector. For example, the first dialogue information may be segmented, and the segmented words obtained may be encoded, and the encodings of different segmented words may form a first feature vector to represent the semantic encoding of the first dialogue information. A similar method may be used to convert the first knowledge information into a second feature vector. Alternatively, the knowledge information stored in the knowledge base is itself structured information such as a vector, and if the first knowledge information is itself in the form of a vector, it may be directly used as the second feature vector.

[0064] The similarity between the first feature vector and the second feature vector can be calculated by using methods such as cosine similarity to represent the matching degree between the first dialogue information and the first knowledge information. Then, the first knowledge information with higher similarity is selected, such as the first knowledge information with similarity reaching a similarity threshold (the similarity threshold can be determined based on experience or specific circumstances), or a certain number of first knowledge information with the highest similarity (such as 10 pieces, or a certain proportion such as 10% of the total number of first knowledge information) as the second knowledge information.

[0065] By extracting the first and second feature vectors and calculating the similarity, the matching degree between the first dialogue information and different first knowledge information can be accurately measured, and then the first knowledge information semantically related to the first dialogue information can be screened out as the second knowledge information, providing accurate and appropriate knowledge information for subsequent dialogue processing, which is conducive to outputting high-quality reply information.

[0066] Step S330: Generate prompt information according to the second knowledge information.

[0067] Among them, the prompt information is information that provides reference and guidance for the processing of the language model, and can be information input to the language model. The second knowledge information can provide reference for the processing of the first dialogue information from the professional knowledge level, and the second knowledge information can be integrated to generate the prompt information. Alternatively, the first dialogue information and the second knowledge information are integrated to generate the prompt information.

[0068] In one implementation, the first dialogue information and the second knowledge information may be combined according to a preset prompt template to obtain prompt information. For example, the prompt template may be as follows:

[0069] “Please refer to the following professional knowledge and provide corresponding reply information for the dialogue information entered by the user;

[0070] Professional knowledge: {second knowledge information};

[0071] The conversation information input by the user is: {first conversation information}."

[0072] Fill the second knowledge information and the first dialogue information into the corresponding positions in the above prompt template to form a complete prompt information. If the first dialogue information has context information, such as the first dialogue information is a sentence in a continuous dialogue, the context information can be added to the prompt information to help the language model better understand the dialogue context.

[0073] Step S340: Process the prompt information by calling the language model through the model node to obtain reply information of the first dialogue information.

[0074] Among them, the prompt information can be input into the language model, and the language model can output the reply information, or the output information of the language model can be further processed, such as converting the structured output information into natural language, or performing certain language optimization or polishing to obtain the reply information.

[0075] In one embodiment, the model node includes interface information for calling the language model, and the prompt information can be sent to the server where the language model is deployed through the interface information. The server uses the language model to process the prompt information and returns the output information of the language model to obtain reply information.

[0076] In one implementation, the above-mentioned invoking a language model by a model node to process the prompt information to obtain the reply information of the first dialogue information may include the following steps:

[0077] Segment the prompt information into words, and encode the obtained word segments to obtain semantic codes corresponding to the prompt information;

[0078] The language model is called through the model node, the semantic code is input into the language model, and the reply information of the first dialogue information is obtained.

[0079] Among them, through word segmentation and encoding, the prompt information can be structured to obtain semantic encoding in the form of vectors, matrices, etc. The semantic encoding is input into the language model, which can extract the feature information in the language encoding more quickly and accurately, and execute the generation process, and finally output structured reply information or convert it into natural language reply information.

[0080] based on Figure 3 In the process of processing the first dialogue information by the dialogue process, the first process nodes are called one by one to realize the complete dialogue processing, and the capabilities of different components such as the first process node, the knowledge base, and the language model are fully utilized to achieve a high processing efficiency. In addition, the knowledge base, the language model and other components are not directly set in the dialogue process, but are used by the dialogue process through calling, which does not affect the calling of these components by other business processes, and realizes the reuse of component capabilities in different business processes.

[0081] In addition to editing the dialogue process, the canvas can also be used to edit other business processes and implement other business functions. For example, it can be used to edit the knowledge base process to configure the knowledge base node. Figure 4 As shown, the dialog flow processing method may further include the following steps S410 to S440:

[0082] Step S410 , in response to the second node selection instruction, displaying a plurality of selected second process nodes in the canvas, wherein the plurality of second process nodes include a knowledge entry node for entering knowledge information into a knowledge base.

[0083] The knowledge base process is a business process used to build, update or maintain the knowledge base. Users can edit the knowledge base process in a canvas manner, select the process node required for the knowledge base process from the process nodes, and the process node selected by the user for editing the knowledge base process is called the second process node, and the operation instruction for the user to select the second process node is called the second node selection instruction. For example, when editing the knowledge base process, the user can drag the required process node to the canvas to implement the second node selection instruction and thereby determine the second process node.

[0084] The second process node includes a knowledge entry node. Among them, the knowledge entry node is used to enter knowledge information into the knowledge base, such as being able to perform write operations on the knowledge base. In addition, the second process node may also include but is not limited to one or more of the following process nodes: a document upload node, used to upload knowledge documents; a diversion node, used to classify knowledge documents according to formats, etc., and assign them to different subsequent process nodes for processing; a document type conversion node, such as a node that can be used to convert non-text documents into text documents; a document direct parsing node, which can parse text information in a document; a document parsing_OCR node, which can parse information in a document through OCR (Optical Character Recognition); a sharding node, used to shard the parsed knowledge information; a structured node, used to convert knowledge information into structured information such as vectors, and so on.

[0085] Step S420 , in response to the second node connection instruction, displaying the connection relationship of multiple second process nodes in the canvas.

[0086] Among them, the second node connection instruction is an operation instruction for connecting the second process node. The user can connect the second process nodes in the canvas, such as pulling a line from a second process node and connecting it to another second process node, or adding a connection line in the canvas and connecting the two ends of the connection line to different second process nodes. According to the user's second node connection instruction, the corresponding connection relationship is displayed in the canvas, thereby editing the order between different second process nodes.

[0087] Step S430: Generate a knowledge base process according to the connection relationship between the plurality of second process nodes.

[0088] The second process nodes can be integrated in sequence according to the connection relationship of the multiple second process nodes to form a complete processing flow from the first second process node (such as a document upload node) to the last second process node (such as a knowledge entry node), that is, a knowledge base flow. At the program level, the computer can integrate the business process instructions corresponding to the second process nodes according to the connection relationship of the multiple second process nodes to form a complete business process instruction for building, updating or maintaining the knowledge base.

[0089] Figure 5 A schematic diagram of a knowledge base process is shown. The knowledge base process may include but is not limited to the following second process nodes: document upload node, diversion node, document type conversion node, document direct parsing node, document parsing_OCR node, fragmentation node, structured node, and knowledge entry node. Among them, the document upload node is used to upload knowledge documents; the diversion node is used to classify knowledge documents according to formats, etc., and assign them to different subsequent process nodes for processing; the document type conversion node is used to convert document types, such as converting non-text documents into text documents; the document direct parsing node is used to parse text information in documents; the document parsing_OCR node is used to parse information in documents by OCR; the fragmentation node is used to fragment the parsed knowledge information; the structured node is used to convert knowledge information into structured information such as vectors; and the knowledge entry node is used to enter structured knowledge information into the knowledge base. According to the connection relationship of these second process nodes, the second process nodes are integrated to form a knowledge base process.

[0090] Step S440, obtaining third knowledge information, processing the third knowledge information based on the knowledge base process to obtain a knowledge base, and generating a knowledge base node for calling the knowledge base.

[0091] The third knowledge information is the knowledge information that needs to be entered into the knowledge base, such as original knowledge information (such as knowledge documents). For example, the third knowledge information can be obtained by searching in a relevant business database according to business needs. Figure 5The knowledge base process shown can go through a series of processing, such as document diversion, parsing, segmentation, and structural conversion, and finally be entered into the knowledge base to construct a knowledge base, or to supplement and update the existing knowledge base to obtain an updated knowledge base.

[0092] When building a knowledge base, a knowledge base node available in the canvas can be generated for calling the knowledge base. For example, interface information for calling the knowledge base can be obtained, the interface information can be encapsulated, and related calling methods (such as request methods) can be encapsulated to form a knowledge base node.

[0093] exist Figure 4 In the method, a canvas-based knowledge base process development method is provided, so that developers can generate knowledge base processes through visual operations on the canvas without writing code, and then build, update or maintain the knowledge base through the knowledge base process. This reduces the development and management costs of the knowledge base and improves efficiency.

[0094] In one embodiment, the third knowledge information may include: special knowledge information that matches the first business scenario to which the dialogue process belongs. The above-mentioned knowledge base may include: a special knowledge base that matches the first business scenario. For example, the first business scenario is a customer service scenario related to financial management, and the dialogue process is used to provide intelligent dialogue and customer service for the first business scenario. The special knowledge information that matches the first business scenario may be professional knowledge information related to financial management. In order to implement the dialogue service of the first business scenario, a special knowledge base that matches the first business scenario may be constructed so that the dialogue process can obtain matching special knowledge information from the knowledge base. Exemplarily, the above-mentioned processing of the third knowledge information based on the knowledge base process to obtain the knowledge base may include the following steps:

[0095] Based on the knowledge base process, the specialized knowledge information is processed and entered into the general knowledge base to build a specialized knowledge base.

[0096] Among them, the general knowledge base is a knowledge base built based on general knowledge information applicable to different dialogue scenarios. For example, the general knowledge base may include knowledge information in general dialogue scenarios, such as knowledge information describing general language habits and dialogue emotion processing, that is, general knowledge information. General knowledge information can help the language model organize language that conforms to people's speaking habits. For example, based on the knowledge base process, general knowledge information can be input into the knowledge base process to obtain a general knowledge base. On the basis of the constructed general knowledge base, the special knowledge information matching the first business scenario can be entered into the general knowledge base through the knowledge base process, such as according to Figure 5The knowledge base process shown in the figure processes the dedicated knowledge information into the general knowledge base after a series of processing, thereby constructing a dedicated knowledge base matching the first business scenario, which includes general knowledge information and dedicated knowledge information matching the first business scenario. Thus, comprehensive and sufficient knowledge information can be provided for dialogue processing, which is conducive to obtaining high-quality reply information.

[0097] Figure 6 A schematic diagram of realizing a dialogue process based on a dedicated knowledge base is shown. Among them, a general knowledge base can be constructed through a knowledge base process, such as inputting general knowledge information into the knowledge base process, and entering it into the newly created knowledge base after processing to construct a general knowledge base. The dedicated knowledge information of a specific business scenario (such as a chat intelligent robot scenario, a customer service intelligent robot, a telephone intelligent robot, an operation and maintenance intelligent robot, etc.) is entered into a general knowledge base through a knowledge base process to form a dedicated knowledge base for a specific business scenario. Furthermore, the dedicated knowledge base can be called through the knowledge base node in the dialogue process to realize dialogue processing, thereby providing corresponding intelligent dialogue services.

[0098] During the editing process of the above-mentioned dialogue process or knowledge base process, the user can select the required process node in the canvas. If the canvas does not provide the process node required by the user, a new process node can be edited and generated. In one embodiment, the dialogue process processing method can also include the following steps:

[0099] In response to the node editing instruction, displaying the first interface and the request parameter in the node editing interface;

[0100] A new process node is generated according to the first interface and the request parameters; the new process node is a visualization node that calls the first interface according to the request parameter request and executes the business process instruction.

[0101] The node editing interface is an interface for editing process nodes. The node editing instruction is an operation instruction for editing and generating new process nodes. In the node editing interface, the user can set the first interface and request parameters. Figure 7 A schematic diagram of a node editing interface is shown, where the user can set a request link, ie, set a first interface, and can set request parameters such as a request method, request header parameters, query parameters, and request body parameters.

[0102] The request method is a way for a client (such as a browser, mobile application, etc.) to specify what operation to perform on server resources when initiating a request to a server. It specifies the type and semantics of the request.

[0103] Request header parameters are key-value pairs contained in the HTTP (Hyper Text Transfer Protocol) request header. They are used to convey various additional information about the request itself to the server, such as some client characteristics, request format requirements, authentication information, etc.

[0104] Query parameters are a set of key-value pairs attached to the URL, which are mainly used to pass additional information to the server so that the server can filter, query related resources or perform specific operations based on this information.

[0105] The request body parameter is the data content contained in the HTTP request body. It is mainly used to pass the data to be processed to the server in POST, PUT, PATCH and other requests.

[0106] The process node for implementing HTTP requests is thus edited. This improves the scalability of the canvas. Users can create process nodes with different functions and inject different component capabilities into the canvas, so that more business processes can be edited based on the canvas.

[0107] For example, the new process node generated by the editing can be used as a general process node, and different business processes can be implemented by setting different first interfaces and request parameters. Figure 7 The process node shown implements the speech-to-text function. Specifically, an access link to a third-party audio-to-text application interface can be added in the node editing interface, and request parameters such as the speech encoding format and the speech conversion type can be set to edit and generate a speech-to-text node.

[0108] When a business process is generated based on the canvas, the business process can be tested and debugged. In one embodiment, the dialog process processing method can also include the following steps:

[0109] inputting test information into the dialogue process;

[0110] Calling the first process node in the dialog process to process the test information, wherein after the first process node completes processing, displaying the current intermediate test result;

[0111] In response to the continue test instruction, the next first process node is continuously called until the final test result is output.

[0112] The test information is information suitable for the business process being tested. For example, for a dialog process, the test information may be dialog information, which may be referred to as second dialog information for easy distinction. After the second dialog information is input into the dialog process, the first process node in the dialog process may be called in sequence to process the second dialog information according to the order in the dialog process.

[0113] During the test process, when each first process node completes the processing, the corresponding intermediate test result can be returned and displayed on the front-end page (such as the test or debugging page) for the debugger to view. In addition, when each first process node completes the processing and returns the intermediate test result, the test process can be paused to wait for the debugger to check the intermediate test result. The front-end page can provide controls such as continue testing and test exception. If the debugger believes that the intermediate test result is normal, he can click the continue testing control to trigger the continue testing instruction. The dialogue process responds to the continue testing instruction and continues to call the next first process node to execute the processing flow. Alternatively, if the debugger does not enter an instruction for a long time (such as 5 minutes) after returning the intermediate test result and pausing the test process, the system determines that the timeout has occurred and automatically triggers the continue testing instruction, so that the dialogue process continues to call the next first process node to execute the processing flow. Thus, the first process nodes are called one by one until the final test result is output. In this way, the single-step debugging method is realized, and the debugger can more easily find which link in the test process has an exception, and locate the first process node that causes the exception, which is convenient for modification.

[0114] Figure 8 The test and debugging process of the dialogue process is shown. Among them, the debugger can start the dialogue process in the debugging mode to realize the start of the single-step debugging process. After starting the single-step debugging process, the first process node can be initialized, the canvas configuration parameters can be checked, the log data generated by the process can be asynchronously sent to the message middleware for storage, and the single-step debugging process can be controlled to be in a suspended state. When the single-step debugging process is in a suspended state, in response to receiving a continue test instruction, the single-step debugging process is switched from a suspended state to a running state, and the current first process node is executed to obtain the corresponding intermediate test result. After obtaining the intermediate test result, the intermediate test result can be asynchronously sent to the message middleware for storage, and the sleep interface can be called to control the single-step debugging process to be in a suspended state through the sleep interface. The intermediate test result is sent to the front-end page for presentation through the message middleware 4 using the communication connection created between the front-end page. The debugger analyzes the abnormal operation problem by viewing the intermediate test results in the front-end page. If there is no abnormality, the continue test control can be clicked to trigger the continue test instruction. The front-end page will send the continue test instruction to the message middleware. When the single-step debugging process is in a paused state, in response to receiving a continue test instruction, the single-step debugging process is switched from the paused state to the running state, and the next first process node is continued to be executed. This loop is executed so that each first process node can complete the single-step debugging.

[0115] In one implementation, the dialog flow processing method may further include the following steps:

[0116] After any first process node returns an intermediate test result, in response to a test exception confirmation instruction, a setting interface of the first process node is displayed; the setting interface is used to adjust the parameters of the first process node.

[0117] Among them, if the user thinks that the intermediate test results are abnormal in the front-end page, he can click the test exception control to trigger the exception confirmation instruction, indicating that there may be an exception in the currently executed first process node, and can jump to the setting interface of the first process node to facilitate the adjustment of the parameters of the first process node.

[0118] like Fig.9A As shown in the figure, a setting interface for process nodes is provided. You can click the corresponding modification control of the parameters to open the Fig. 9B The parameter configuration sub-interface shown in the figure is used to implement the configuration update of the corresponding parameters in the parameter configuration sub-interface, so as to modify and optimize the dialogue process. Optionally, the parameter configuration sub-interface may also include an additional parameter control, which can be used to control the configuration information corresponding to the user-defined parameters, such as Fig. 9C As shown in the provided additional parameter interface, users can adjust additional parameters in the additional parameter interface, which can improve the convenience and flexibility of process node configuration updates to a certain extent.

[0119] The exemplary embodiment of the present disclosure also provides a dialog flow processing device, referring to Fig.10 As shown, the dialog flow processing device 1000 may include the following program modules:

[0120] The node selection module 1010 is configured to display a plurality of selected first process nodes in a canvas in response to a first node selection instruction, wherein the plurality of first process nodes include at least one of a knowledge base node for calling a knowledge base and a model node for calling a language model;

[0121] The node connection module 1020 is configured to display the connection relationship of the plurality of first process nodes in the canvas in response to the first node connection instruction;

[0122] The business process generation module 1030 is configured to generate a dialogue process according to the connection relationship of the plurality of first process nodes;

[0123] The business process execution module 1040 is configured to generate reply information of the first dialogue information based on the dialogue process in response to receiving the first dialogue information.

[0124] In one embodiment, generating reply information for the first dialogue information based on the dialogue process includes: calling the knowledge base through the knowledge base node to obtain first knowledge information from the knowledge base; determining second knowledge information matching the first dialogue information from the first knowledge information by matching the first dialogue information with the first knowledge information; generating prompt information based on the second knowledge information; and calling the language model through the model node to process the prompt information to obtain reply information for the first dialogue information.

[0125] In one embodiment, the matching of the first dialogue information with the first knowledge information to determine the second knowledge information matching the first dialogue information includes: extracting features from the first dialogue information to obtain a first feature vector, and extracting features from the first knowledge information to obtain a second feature vector; determining the similarity between the first feature vector and the second feature vector; and determining the second knowledge information matching the first dialogue information from the first knowledge information according to the similarity.

[0126] In one implementation, the step of calling the language model through the model node to process the prompt information to obtain reply information of the first dialogue information includes: segmenting the prompt information and encoding the obtained segmented words to obtain semantic coding corresponding to the prompt information; calling the language model through the model node, inputting the semantic coding into the language model, and obtaining reply information of the first dialogue information.

[0127] In one embodiment, the dialog process processing device 1000 also includes a testing module, which is configured to: input test information into the dialog process; call the first process node in the dialog process to process the test information, wherein after the first process node completes processing, the current intermediate test result is displayed; in response to a continue testing instruction, continue to call the next first process node until the final test result is output.

[0128] In one embodiment, the node selection module 1010 is further configured to: display multiple selected second process nodes in a canvas in response to a second node selection instruction, wherein the multiple second process nodes include a knowledge entry node for entering knowledge information into a knowledge base; the node connection module 1020 is further configured to: display the connection relationship between the multiple second process nodes in a canvas in response to a second node connection instruction; the business process generation module 1030 is further configured to generate a knowledge base process based on the connection relationship between the multiple second process nodes; the business process execution module 1040 is further configured to: obtain third knowledge information, process the third knowledge information based on the knowledge base process, obtain the knowledge base, and generate the knowledge base node for calling the knowledge base.

[0129] In one embodiment, the dialog process processing device 1000 also includes a node editing module, which is configured to: display a first interface and request parameters in a node editing interface in response to a node editing instruction; generate a new process node based on the first interface and the request parameters; the new process node is a visualization node that calls the first interface according to the request parameter request and executes the business process instruction.

[0130] The specific details of each part of the above-mentioned device have been described in detail in the implementation method of the method part. The undisclosed details can be found in the implementation method of the method part, so they will not be repeated here.

[0131] The exemplary embodiments of the present disclosure also provide a computer program product, which includes a computer program, and when the computer program is executed by a processor, the above-mentioned dialog flow processing method is implemented.

[0132] In one embodiment, the computer program product may be a tangible product containing a computer program, such as a computer-readable storage medium storing a computer program. The readable storage medium may be a storage medium based on electrical, magnetic, optical, electromagnetic, infrared, or other signals, including but not limited to: random access memory (RAM), read-only memory (ROM), magnetic tape, floppy disk, flash memory (Flash), mechanical hard disk (HDD), solid-state drive (SSD), and the like. Exemplarily, the computer program product may be implemented as a non-volatile storage medium storing a computer program, such as a read-only memory, a NAND flash memory, and the like.

[0133] In one embodiment, the computer program product may be an intangible product including a computer program. Exemplarily, the computer program product may be implemented as a virtual digital product, such as a digital file storing an executable file, an installation package, etc. of the computer program.

[0134] The code of the computer program can be written in one or more programming languages. Programming languages ​​such as C language, Java, C++, etc. The program code can be executed entirely on the user computing device, or partially on the user computing device, or as a separate software package, or partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server. In the case of a remote computing device, the remote computing device can be connected to the user computing device through any type of network, such as a local area network (LAN), a wide area network (WAN), etc., or can be connected to an external computing device (e.g., an Internet connection provided by an operator).

[0135] The computer program may be carried or transmitted via electrical, magnetic, optical, electromagnetic, infrared, or other signals. The electronic device may convert the signal carrying the computer program into a digital signal, thereby running the computer program. When the computer program runs on the electronic device, its code is used to enable the electronic device to execute (more specifically, to enable the processor of the electronic device to execute) the method steps of various exemplary embodiments of the present disclosure, such as Figure 1 The method steps shown, etc.

[0136] The above method steps are implemented by a computer program. By selecting a first process node on the canvas and establishing a connection relationship between the selected multiple first process nodes, a dialogue process composed of the first process nodes can be generated, and reply information for the first dialogue information can be generated through the dialogue process. Since the entire process of establishing a dialogue process based on the canvas is visualized, the user only needs to select the process node, without writing a large amount of code, which simplifies the development process of the dialogue function, reduces the development time, and thus improves the efficiency of dialogue processing. In addition, the first process node provided by the canvas includes a knowledge base node for calling the knowledge base and a model node for calling the language model. Therefore, when the generated dialogue process processes the first dialogue information, it can combine the knowledge base and the language model to analyze and understand the first dialogue information, so that the generated reply information can be more compatible with the first dialogue information, and can reply to the first dialogue information more accurately. In addition, the visual editing and development method based on the canvas does not require developers to master professional code-related knowledge, which reduces the technical threshold for developers and has a wider scope of application.

[0137] The exemplary embodiment of the present disclosure also provides an electronic device capable of implementing the above-mentioned dialog flow processing method. The electronic device may include a processor and a memory. The memory stores executable instructions of the processor, such as program code. The processor executes the method in the exemplary embodiment by executing the executable instructions. In addition, the electronic device may also include a display for displaying a graphical user interface.

[0138] Reference below Fig.11 , the electronic device is exemplarily described in the form of a general-purpose computing device. It should be understood that Fig.11 The electronic device 1100 shown is merely an example and should not limit the functions and scope of use of the embodiments of the present disclosure.

[0139] like Fig.11 As shown, the electronic device 1100 may include: a processor 1110 , a memory 1120 , a bus 1130 , an I / O (input / output) interface 1140 , a network adapter 1150 , and a display 1160 .

[0140] The memory 1120 may include a volatile memory, such as a RAM 1121, a cache unit 1122, and may also include a non-volatile memory, such as a ROM 1123. The memory 1120 may also include one or more program modules 1124, such program modules 1124 include but are not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or a combination thereof may include the implementation of a network environment. For example, the program module 1124 may include each module in the above-mentioned device.

[0141] Processor 1110 may include one or more processing units. For example, processor 1110 may include an AP (Application Processor), a modem processor, a GPU (Graphics Processing Unit), an ISP (Image Signal Processor), a controller, an encoder, a decoder, a DSP (Digital Signal Processor), a baseband processor and / or an NPU (Neural-Network Processing Unit) and other processing units.

[0142] The processor 1110 may be configured to execute executable instructions stored in the memory 1120 , such as executing any one or more method steps in the exemplary embodiment.

[0143] By executing the above method steps through the processor 1110, by selecting the first process node on the canvas and establishing the connection relationship between the selected multiple first process nodes, a dialogue process composed of the first process nodes can be generated, and the reply information for the first dialogue information can be generated through the dialogue process. Since the entire process of establishing a dialogue process based on the canvas is visual, the user only needs to select the process node, without writing a large amount of code, which simplifies the development process of the dialogue function, reduces the development time, and thus improves the efficiency of dialogue processing. In addition, the first process node provided by the canvas includes a knowledge base node for calling the knowledge base and a model node for calling the language model. Therefore, when the generated dialogue process processes the first dialogue information, it can combine the knowledge base and the language model to analyze and understand the first dialogue information, so that the generated reply information can be more compatible with the first dialogue information, and can reply to the first dialogue information more accurately. In addition, the visual editing and development method based on the canvas does not require developers to master professional code-related knowledge, reduces the technical threshold for developers, and has a wider scope of application.

[0144] The bus 1130 is used to realize the connection between different components of the electronic device 1100, and may include a data bus, an address bus, and a control bus.

[0145] The electronic device 1100 can communicate with one or more external devices 1200 (eg, a keyboard, a mouse, an external controller, etc.) through the I / O interface 1140 .

[0146] The electronic device 1100 can communicate with one or more networks through the network adapter 1150. For example, the network adapter 1150 can provide mobile communication solutions such as 3G / 4G / 5G, or provide wireless communication solutions such as wireless LAN, Bluetooth, near field communication, etc. The network adapter 1150 can communicate with other modules of the electronic device 1100 through the bus 1130.

[0147] The electronic device 1100 may display a graphical user interface, such as a canvas interface, through the display 1160 .

[0148] although Fig.11 Not shown, other hardware and / or software modules may also be provided in the electronic device 1100, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID (Redundant Arrays of Independent Disks) systems, tape drives, and data backup storage systems.

[0149] It should be noted that, although several modules or units of the device for action execution are mentioned in the above detailed description, this division is not mandatory. In fact, according to the exemplary embodiments of the present disclosure, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided into multiple modules or units to be embodied.

[0150] It will be appreciated by those skilled in the art that various aspects of the present disclosure may be implemented as a system, method or program product. Therefore, various aspects of the present disclosure may be specifically implemented in the following forms, namely: a complete hardware implementation, a complete software implementation (including firmware, microcode, etc.), or an implementation combining hardware and software aspects, which may be collectively referred to herein as a "circuit", "module" or "system". Those skilled in the art will readily think of other embodiments of the present disclosure after considering the specification and practicing the invention disclosed herein. The present application is intended to cover any variations, uses or adaptive changes of the present disclosure, which follow the general principles of the present disclosure and include common knowledge or customary technical means in the art that are not disclosed in the present disclosure. The specification and implementation are intended to be exemplary only, and the true scope and spirit of the present disclosure are indicated by the claims.

[0151] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. A method for processing a dialog flow, characterized in that: The method comprises: In response to the first node selection instruction, displaying a plurality of selected first process nodes in the canvas, wherein the plurality of first process nodes include at least one of a knowledge base node for calling a knowledge base and a model node for calling a language model; In response to the first node connection instruction, displaying the connection relationship of the plurality of first process nodes in the canvas; Generate a dialogue process according to the connection relationship of the plurality of first process nodes; In response to receiving the first dialogue information, reply information of the first dialogue information is generated based on the dialogue flow.

2. The method according to claim 1, characterized in that: The generating reply information of the first dialogue information based on the dialogue process includes: Calling the knowledge base through the knowledge base node to obtain first knowledge information from the knowledge base; matching the first dialogue information with the first knowledge information, and determining second knowledge information matching the first dialogue information from the first knowledge information; generating prompt information according to the second knowledge information; The language model is called by the model node to process the prompt information to obtain reply information of the first dialogue information.

3. The method according to claim 2, characterized in that The matching the first dialogue information with the first knowledge information to determine second knowledge information matching the first dialogue information from the first knowledge information includes: Performing feature extraction on the first conversation information to obtain a first feature vector, and performing feature extraction on the first knowledge information to obtain a second feature vector; determining a similarity between the first feature vector and the second feature vector; Second knowledge information matching the first dialogue information is determined from the first knowledge information according to the similarity.

4. The method according to claim 2, characterized in that: The step of calling the language model through the model node to process the prompt information to obtain reply information of the first dialogue information includes: Segmenting the prompt information, and encoding the obtained segmented words to obtain semantic codes corresponding to the prompt information; The language model is called through the model node, the semantic code is input into the language model, and reply information of the first dialogue information is obtained.

5. The method according to claim 1, characterized in that The method further comprises: inputting test information into the dialogue process; Calling the first process node in the dialog process to process the test information, wherein after the first process node completes processing, displaying the current intermediate test result; In response to the continue test instruction, the next first process node is continuously called until the final test result is output.

6. The method according to claim 1, characterized in that The method further comprises: In response to the second node selection instruction, displaying a plurality of selected second process nodes in the canvas, wherein the plurality of second process nodes include a knowledge entry node for entering knowledge information into a knowledge base; In response to the second node connection instruction, displaying the connection relationship of the plurality of second process nodes in the canvas; Generate a knowledge base process according to the connection relationship of the plurality of second process nodes; The third knowledge information is acquired, and the third knowledge information is processed based on the knowledge base process to obtain the knowledge base, and the knowledge base node for calling the knowledge base is generated.

7. The method according to claim 1, characterized in that The method further comprises: In response to the node editing instruction, displaying the first interface and the request parameter in the node editing interface; A new process node is generated according to the first interface and the request parameters; the new process node is a visualization node that calls the first interface according to the request parameter request and executes the business process instruction.

8. A dialog flow processing device, characterized in that: The device comprises: A node selection module is configured to display a plurality of selected first process nodes in a canvas in response to a first node selection instruction, wherein the plurality of first process nodes include at least one of a knowledge base node for calling a knowledge base and a model node for calling a language model; a node connection module, configured to display the connection relationship of the plurality of first process nodes in the canvas in response to a first node connection instruction; A business process generation module, configured to generate a dialogue process according to the connection relationship of the plurality of first process nodes; The business process execution module is configured to generate reply information of the first dialogue information based on the dialogue process in response to receiving the first dialogue information.

9. An electronic device, characterized in that: include: processor; as well as A memory, configured to store executable instructions of the processor; The processor is configured to perform the method of any one of claims 1 to 7 by executing the executable instructions.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method described in any one of claims 1 to 7 are implemented.

Images