Business processing method and device based on human-computer interaction, equipment and storage medium

Abstract

Embodiments of the present application provide a business processing method and device based on human-computer interaction, an electronic device and a storage medium. The method is applied to an electronic device, at least one business agent is deployed in the electronic device, the business agent is used for processing a corresponding business, at least one logic block of a business step is encapsulated in the business agent, the business step and the logic block correspond to each other, and the logic block represents the processing logic of the business step. The method comprises: obtaining context information input by a user in a current dialogue turn, and obtaining current business state information; the business state information represents the progress of business processing; determining a target agent from the at least one business agent according to the context information and the current business state information, and determining a logic block to be called from the target agent; updating the current business state information according to the logic block to be called, and outputting reply information to the user; and the reply information represents the processing result of the logic block to be called.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a business processing method, apparatus, device and storage medium based on human-computer interaction. Background Technology

[0002] In the fintech field, the processing of various businesses involves complex multi-step processes. Users need to input business elements step by step through multiple rounds of dialogue, such as transaction time, amount, merchant information, etc. The system needs to dynamically track the context and guide users to complete the business objectives.

[0003] Currently, errors are prone to occur during human-computer interaction business processing. Therefore, there is an urgent need for an efficient and intelligent business processing method to improve the accuracy and efficiency of business processing. Summary of the Invention

[0004] This application provides a business processing method, apparatus, device, and storage medium based on human-computer interaction to improve the efficiency and accuracy of business processing.

[0005] In a first aspect, embodiments of this application provide a business processing method based on human-computer interaction. This method is applied to an electronic device, which deploys at least one business intelligence agent. The business intelligence agent is used to process corresponding business processes. Each business intelligence agent encapsulates at least one logic block representing a business step. Each business step corresponds one-to-one with the logic block, and the logic block represents the processing logic of the business step. The method includes:

[0006] Obtain the context information input by the user in the current dialogue round, and obtain the current business status information; wherein, the business status information represents the progress of business processing;

[0007] Based on the context information and the current business status information, a target intelligent agent is determined from the at least one business intelligent agent, and a logic block to be invoked is determined from the target intelligent agent;

[0008] Based on the logic block to be invoked, update the current business status information and output a response message to the user; wherein the response message represents the processing result of the logic block to be invoked.

[0009] Secondly, embodiments of this application provide a business processing device based on human-computer interaction. This device is applied to an electronic device, which has at least one business intelligence agent deployed therein. The business intelligence agent is used to process corresponding business processes. Each business intelligence agent encapsulates at least one logic block of a business step, with each business step corresponding one-to-one with the logic block. The logic block represents the processing logic of the business step. The device includes:

[0010] The information acquisition unit is used to acquire contextual information input by the user in the current dialogue round, and to acquire current business status information; wherein, the business status information represents the progress of business processing.

[0011] The target determination unit is configured to determine a target intelligent agent from the at least one business intelligent agent based on the context information and the current business status information, and to determine the logic block to be invoked from the target intelligent agent;

[0012] The information update unit is used to update the current business status information according to the logic block to be invoked, and output reply information to the user; wherein the reply information represents the processing result of the logic block to be invoked.

[0013] Thirdly, embodiments of this application provide an electronic device, including: a memory and a processor;

[0014] The memory stores computer-executed instructions;

[0015] The processor executes computer execution instructions stored in the memory, causing the processor to perform the implementation method described in the first aspect above.

[0016] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the embodiments described in the first aspect above.

[0017] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the implementation methods described in the first aspect above.

[0018] This application provides a business processing method, apparatus, device, and storage medium based on human-computer interaction. Each business can correspond to a business intelligence agent for processing the corresponding business. The processing of each business involves multiple business steps, and each business process corresponds to a logic block. The business intelligence agent encapsulates multiple logic blocks, and each logic block represents the processing logic of the business step. In one dialogue round, the user can input information multiple times. All information input by the user can be obtained as context information, and the current business status information can be obtained, which represents the progress of business processing. Based on the context information and the current business status information, the type of business that needs to be processed is determined, that is, the target intelligence agent can be determined from multiple business intelligence agents. The target intelligence agent contains multiple logic blocks. The logic block to be invoked is determined from the target intelligence agent, the invoked logic block is executed, the response information to the user and the new business status information are obtained, the business status information is updated, and the response information is output to the user, completing the processing of the current step. By pre-building multiple business intelligence agents to handle different business tasks, each business intelligence agent encapsulates multiple logic blocks. When business processing is required, the target intelligence agent is determined first, and then the target logic block is determined, making the processing flow clear. Furthermore, by combining context and processing status during the processing, the efficiency and accuracy of processing are improved, and the user's human-computer interaction experience is enhanced. Attached Figure Description

[0019] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0020] Figure 1 A flowchart illustrating a business processing method based on human-computer interaction provided in an embodiment of this application;

[0021] Figure 2 A flowchart illustrating a business processing method based on human-computer interaction provided in an embodiment of this application;

[0022] Figure 3 This is a schematic diagram of the logical design of the intelligent agent provided in the embodiments of this application;

[0023] Figure 4 A flowchart illustrating a business processing method based on human-computer interaction provided in an embodiment of this application;

[0024] Figure 5 A flowchart illustrating a business processing method based on human-computer interaction provided in an embodiment of this application;

[0025] Figure 6 A schematic diagram of the structure of a business processing device based on human-computer interaction provided in an embodiment of this application;

[0026] Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0027] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0028] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application.

[0029] First, let me explain the terms used in this application:

[0030] Business intelligence agent: An independent unit encapsulated with business processing logic. Each business intelligence agent corresponds to a specific business, such as dispute negotiation, evidence supplementation, account inquiry, etc.

[0031] Logic block: Also known as workflow, it refers to a logical module that encapsulates business steps, contains functions used for processing, and is responsible for processing specific business steps;

[0032] Token: The basic unit of text processing when inputting data into a large model. For example, one token is approximately equal to one or two Chinese characters.

[0033] Human-computer interaction technology is increasingly being used in fields such as financial services and customer support, and multi-turn dialogue systems have become an important tool for improving user experience and business processing efficiency. In complex business scenarios, systems need to gradually collect business elements input by users through multiple rounds of dialogue in order to achieve specific business objectives.

[0034] In financial business scenarios such as error dispute resolution, users need to continuously handle multiple independent and potentially dynamically switching business processes within a single conversation session, such as "order adjustment application," "order cancellation appeal," and "negotiation and settlement." This leads to blurred business boundaries. When a user intends to switch from "order adjustment application" to "order cancellation appeal," the model is prone to incorrectly reusing fields from previous business processes in the current business, causing logical conflicts, resulting in increased response latency and a higher illusion rate. In other words, in scenarios involving multiple business types and allowing switching between them, it is currently difficult to effectively manage the state information of different businesses, leading to state confusion and information loss during business processing. Simultaneously, the number of business elements stored in the session continuously increases, causing excessive consumption of storage resources. These problems are particularly prominent in businesses like error dispute resolution, which have strong rules, multiple branches, and high fault tolerance requirements. There is an urgent need for a technical solution that can maintain business isolation while achieving lightweight state expression and cross-business collaborative scheduling.

[0035] This application provides a business processing method, apparatus, device, and storage medium based on human-computer interaction, which is applied in the field of artificial intelligence technology to improve the efficiency and accuracy of business processing.

[0036] It should be noted that the data in this embodiment is not specific to any particular user and does not reflect the personal information of any particular user. It should also be noted that the data in this embodiment comes from a publicly available dataset.

[0037] The collection, storage, use, processing, transmission, provision, and disclosure of user personal information involved in the technical solution of this application all comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0038] To enable readers to have a deeper understanding of the implementation principles of this application, the embodiments are further detailed.

[0039] Figure 1 This is a flowchart illustrating a human-computer interaction-based business processing method according to an embodiment of this disclosure. The method can be executed by a human-computer interaction-based business processing device. The method is applied to an electronic device, which deploys at least one business intelligence agent. The business intelligence agent is used to process corresponding business processes. Each business intelligence agent encapsulates at least one logic block of a business step, with each business step corresponding to a logic block, and the logic block representing the processing logic of the business step. For example... Figure 1 As shown, the method includes the following steps:

[0040] S101. Obtain the context information input by the user in the current dialogue round, and obtain the current business status information; wherein, the business status information represents the progress of business processing.

[0041] For example, electronic devices refer to terminal devices with computing capabilities and human-computer interaction interfaces, including but not limited to smartphones, tablets, desktop computers, dedicated financial operation terminals, or cloud server clusters. Their core function is to support and schedule various intelligent agents, and they are not limited to hardware form factors; they can also be deployed on edge computing nodes or in hybrid cloud architectures. Electronic devices achieve lifecycle management, communication routing, and state persistence for business intelligent agents through the operating system layer or middleware framework, ensuring resource isolation and secure concurrent execution between intelligent agents.

[0042] A business intelligence agent is a software-defined autonomous agent unit, essentially an executable module that encapsulates specific business domain knowledge, a rule engine, a state machine, and API (Application Programming Interface) call capabilities. Each business intelligence agent is strictly bound to a unique business type, such as an "error and order adjustment agent" or a "refund and appeal agent." Internally, a business intelligence agent can have a pre-defined business process graph, which expresses the sequential constraints and conditional branching logic between business steps. Graph nodes represent business steps, and each step maps to an independent logic block. Logic blocks are atomic callable function units whose input interface receives structured parameters, such as field names and values ​​defined in JSON Schema, and whose output interface returns standard-formatted business status information and natural language response text. Logic blocks can be implemented based on a rule engine, a lightweight LLM (Large Language Model) fine-tuning model, or traditional scripts; different implementation methods can be dynamically selected according to business complexity.

[0043] Business status information is organized in a multi-dimensional and structured manner, and includes at least business elements, business summaries, and business details. Business elements can store the fields required for the current business and their values ​​in key-value pairs. For example, "transaction_id" is a business element with the value "ABC123"; "dispute_reason" is a business element with the value "duplicate_charge". The business summary is a short, machine-parseable string following a unified template. For example, the business summary could be "Executing [Business Agent ID], currently at [Step ID], [Business Element ID] has been entered, [Business Element ID] still needs to be entered". The business summary does not contain the specific values ​​of the business elements, only serving as a progress indicator. Business details are an enhanced version of the summary, explicitly embedding the current values ​​of each field, such as "Executing the order-requesting agent, currently at step 2, Merchant ID (SH001) and transaction amount (299.00) have been entered, transaction time still needs to be entered". Business details can be used to support fine-grained decisions within each step. These three elements together constitute a complete snapshot of the business status information, supporting serialized storage and cross-round transmission.

[0044] The contextual information input by the user in the current dialogue round can include the natural language text actively sent by the user in this round, such as, "I want to appeal this order cancellation; the transaction time was yesterday afternoon at 3 PM." Within a single dialogue round, the user can input multiple times, and the contextual information can include all information entered by the user, including responses from the human-computer interaction. After preprocessing by the front-end, the contextual information retains its original semantics and key entities, without undergoing summary compression or feature extraction, and is directly fed into the subsequent intelligent agent decision chain as raw input, ensuring semantic integrity.

[0045] Each time feedback is sent to the user, the business status information can be updated, and the current business status is the latest information. The current business status information can be read from a local memory cache or a distributed key-value database. If it's the first interaction, it's initialized as an empty object, and a default business status slot is automatically created. The business status information can use an incremental synchronization mechanism, that is, only a subset of fields that have changed since the last round are transmitted, further reducing network overhead and memory usage.

[0046] S102. Based on the context information and the current business status information, determine the target intelligent agent from at least one business intelligent agent, and determine the logic block to be invoked from the target intelligent agent.

[0047] For example, contextual information and current business status information are used as joint input features and fed into the decision model. The decision model can be a lightweight classifier, a small language model, or a rule matching engine. It determines which type of business needs to be processed. Each type of business corresponds to a business agent, thus identifying the target agent. The process of determining the target agent reflects the isolation of business processes. For example, it can iterate through all registered business agents, each with its own explanation describing the required input data. For instance, for the order request agent, the explanation describes handling order requests initiated by merchants, involving transaction voucher uploads and time limit verification. The explanations of each business agent are semantically matched with the current context information and business status information, and the business agent with the highest matching degree is selected as the target agent. The matching algorithm supports multiple metrics such as cosine similarity and BERT (Bidirectional Encoder Representations from Transformers) embedding distance, and can be flexibly switched according to the computing power of the deployment environment.

[0048] Each business agent encapsulates multiple logic blocks. The logic block currently needed is determined from the target agent and designated as the logic block to be invoked. That is, the business step to be executed is identified, and the logic block corresponding to that step is chosen as the logic block to be invoked. This can be done based on the current business state information and the current node of the target agent's internal state machine. For example, the business summary in the business state information explicitly indicates the current step number, such as "Step 2." This number directly indexes the corresponding node in the agent's DAG (Directed Acyclic Graph), thus locking onto a unique logic block.

[0049] S103. Based on the logic block to be invoked, update the current business status information and output a response message to the user; wherein, the response message represents the processing result of the logic block to be invoked.

[0050] For example, the logic block to be invoked is called to execute the business processing function encapsulated within that logic block. The input parameters of the logic block can be composed of at least two parts: first, the stored business elements extracted from the current business state information; and second, newly added business elements extracted in real time from the context information of the current dialogue round. For example, the business element "yesterday afternoon at 3 PM" can be extracted using a NER (Named Entity Recognition) model. During the execution of the logic block, business rules can be validated on the input business elements, such as checking whether the transaction time is within the dispute window, detecting missing fields, and querying related data. In this embodiment, the processing logic within the logic block is not specifically limited.

[0051] After execution, the logic block can update the business status information. For example, the logic block outputs a status update package, which is atomically written into the structure of the business status information. The status update package can include the processing result of the current step, the business elements used in the processing, the specific values ​​of the business elements, the next business steps to be executed, and the business elements to be used in the next step. Based on the status update package, business element fields in the business status information can be overwritten or appended, and the business summary can also be refreshed according to a preset template, such as updating from "Step 2" to "Step 3". The business details are updated synchronously with the field values ​​and prompt text. The updated business status information takes effect immediately for the next use. The logic block can generate and output response information based on the processing result, providing feedback to the user through human-computer interaction. The response information can be natural language text generated by the logic block, and its content strictly corresponds to the processing result. For example, if the business elements are complete and the verification passes, a prompt guiding the next operation is output, such as "The voucher has been received, please confirm the appeal reason"; if there are missing business elements, structured feedback is output, such as "Please provide a screenshot link of the transaction voucher". Reply messages can be pushed to user terminals via a unified messaging gateway, supporting various presentation formats such as rich text, button cards, or voice broadcast.

[0052] In this embodiment, business domain isolation is achieved on a business intelligence agent basis, fundamentally eliminating state pollution caused by the mixing of multiple business elements. By combining context information and business state information, the target intelligence agent and the logic block to be invoked can be accurately identified, avoiding processing flow chaos and improving processing efficiency and accuracy.

[0053] This application provides a business processing method based on human-computer interaction. Each business can correspond to a business agent for processing the corresponding business. The processing of each business involves multiple business steps, and each business process corresponds to a logic block. The business agent encapsulates multiple logic blocks, which represent the processing logic of the business steps. In one dialogue round, the user can input information multiple times. All information input by the user can be obtained as context information, and the current business status information can be obtained, which represents the progress of business processing. Based on the context information and the current business status information, the type of business that needs to be processed is determined, that is, the target agent can be determined from multiple business agents. The target agent contains multiple logic blocks. The logic block to be invoked is determined from the target agent, the invoked logic block is executed, the response information to the user and the new business status information are obtained, the business status information is updated, and the response information is output to the user, completing the processing of the current step. By pre-building multiple business intelligence agents to handle different business tasks, each business intelligence agent encapsulates multiple logic blocks. When business processing is required, the target intelligence agent is determined first, and then the target logic block is determined, making the processing flow clear. Furthermore, by combining context and processing status during the processing, the efficiency and accuracy of processing are improved, and the user's human-computer interaction experience is enhanced.

[0054] Figure 2 A flowchart illustrating a business processing method based on human-computer interaction provided in this application embodiment is shown below. Figure 2 As shown, this embodiment, based on the above embodiments, provides a detailed description of a business processing method based on human-computer interaction, which includes:

[0055] S201. Obtain the context information input by the user in the current dialogue round, and obtain the current business status information; wherein, the business status information represents the progress of business processing.

[0056] S202. Based on the context information and the current business status information, determine the target description information based on the routing agent; wherein, the target description information represents the description information of the business to be processed by the user.

[0057] For example, an electronic device deploys a routing agent, which routes to a service agent based on service status information and context information. Each service agent corresponds to service description information, which characterizes the description of the service for which the service agent corresponds.

[0058] A routing agent is a lightweight agent module independently deployed in electronic devices. Its core function is to perform cross-business dimension intent recognition and scheduling decisions. It does not directly participate in specific business logic processing, but only undertakes semantic understanding, intent abstraction, and matching guidance tasks. The routing agent can be implemented using a miniaturized language model or a hybrid rule-and-vector engine. As the decision model in the above embodiments, the routing agent's input consists of the user input text of the current dialogue turn, i.e., context information and current business state information. Specifically, it can include summary information from the context information and business state information; the summary information is the business summary.

[0059] Based on contextual information and current business status information, the routing agent can determine a structured, semantically concise natural language description, i.e., the target description information. The target description information is not a restatement of the original context, but rather a high-level business semantic expression after semantic compression and intent enhancement. For example, combining the user input "I want to check yesterday's rejected transaction and also submit the dispute materials" with the summary information "Error query completed, dispute submission process not started" in the current business status information, the target description information "User intends to initiate UnionPay error dispute submission process, associated with previous query results" is abstracted to generate "User intends to initiate UnionPay error dispute submission process, associated with previous query results." In this embodiment, the routing agent can also be configured as a multimodal input interface, supporting the reception of speech-to-text, OCR (Optical Character Recognition) results, or structured form fields as supplementary contextual information.

[0060] In this embodiment, the service status information includes summary information, which represents a summary of the current processing status of the service; based on the context information and the current service status information, the target description information is determined by the routing agent, including: inputting the context information and the current summary information into the routing agent to obtain the target description information.

[0061] Specifically, business status information is a structured data set dynamically maintained during multiple rounds of human-computer interaction, used to characterize the complete execution progress and context snapshot of a single business instance. Business status information includes at least summary information, detail information, and business element information. The summary information is a business summary that describes the current execution status of the business, for example, "Executing business agent 2, currently executing step 1, element 1 has been input, and user input of element 2 is still required." The summary information is generated and updated by the target business agent after each logical block execution. Generation methods include, but are not limited to, semantic compression of the current business elements and step execution results based on LLM, or generation using a preset template.

[0062] The current summary information specifically refers to the summary information in the latest updated business status information associated with the current dialogue round. It originates from the updated business status information output after the previous round of logic block calls and can be stored in memory or a local key-value database, offering low-latency read characteristics.

[0063] Contextual information refers to the natural language text entered by the user in the current dialogue turn, including but not limited to user questions, instructions, confirmation responses, correction feedback, or interruption requests. Contextual information can directly reflect the user's immediate intent and is a key dynamic input for routing decisions.

[0064] The routing agent is a independently deployed, lightweight, large-model inference service unit. Its core function is to perform cross-business intent recognition and target matching, without participating in the execution of specific business logic. In this four-fold strength framework, no specific limitations are placed on the model architecture of the large model. The input to the routing agent can include context information and current summary information. That is, by inputting the context information and current summary information into the routing agent, target description information is obtained, thereby determining what business the user currently wants to process. The target description information can be an intermediate semantic representation output by the routing agent, facilitating subsequent routing to the target agent based on the target description information.

[0065] The beneficial effect of this setup is that the conciseness of the summary information directly determines the total number of input tokens for the routing agent, making the input length of the routing agent stable and controllable, thereby affecting the routing response latency and the probability of model illusion. The collaborative input of summary information and contextual information avoids the state ambiguity caused by relying solely on the original dialogue, and avoids the intent drift caused by relying solely on static business descriptions, thus improving the efficiency and accuracy of business processing.

[0066] S203. Based on the target description information and the business description information corresponding to each business agent, determine the target agent from among the business agents.

[0067] For example, business description information is metadata pre-configured statically or dynamically registered for each business agent, stored in the form of natural language phrases or embedded vectors, used to characterize the essential attributes and applicable boundaries of the business carried by the business agent. For instance, the business description information for the "Error Dispute Submission Agent" could be "Handling cardholder complaints caused by merchant settlement anomalies, supporting material upload, liability determination option selection, timeliness verification, and automatic submission to the clearing system"; the business description information for the "Error Query Agent" could be "Retrieving and displaying all error transaction details under a user's name within a specified time range that are in the 'pending' or 'processing' state, supporting multi-dimensional filtering by transaction time, amount, card number, and error type." Business description information can be stored in a local knowledge base, an embedded vector database, or a distributed configuration center, and supports hot updates.

[0068] The system can search for the business description information that best matches the target description information from the business description information of each business agent, and then identify the business agent corresponding to that business description information as the target agent. For example, the system can perform word segmentation and matching on the target description information and the business description information, and select the business agent corresponding to the business description information with the most matched words as the target agent.

[0069] In this embodiment, the routing agent and various business agents form a hierarchical collaborative architecture. The routing agent resides in the scheduling layer, focusing on "what business to do," while the business agents reside in the execution layer, focusing on "how to perform a certain step of the business." This separation of responsibilities design enables the system to have good scalability. When adding a new business, only the new business agent and its business description information need to be registered, without modifying the routing agent code. When upgrading the routing strategy, there is no need to retrain or reconstruct the business agents. This achieves accurate capture and efficient distribution of user intent in multi-business concurrent scenarios, improving the accuracy of business processing.

[0070] In this embodiment, the target intelligent agent is determined from each business intelligent agent based on the target description information and the business description information corresponding to each business intelligent agent. This includes: determining the semantic similarity between the target description information and the business description information corresponding to each business intelligent agent for each business intelligent agent; and determining the target intelligent agent from each business intelligent agent based on the semantic similarity between each business intelligent agent.

[0071] Specifically, target description information refers to a natural language text description that represents the essence of the business that the user's intent in this round is directed to, generated by the routing agent based on the context information of the current dialogue round and the summary information in the current business state. Business description information corresponding to the business agent refers to a natural language text description that is pre-configured for each business agent and is either manually calibrated or automatically extracted from a business knowledge graph.

[0072] After obtaining the target description information, the semantic similarity between the target description information and the business description information corresponding to each business agent can be calculated. For example, both can be mapped to the same semantic vector space, and the cosine similarity between the embedded vectors can be calculated to obtain the semantic similarity between the target description information and the business description information. In this embodiment, the method of calculating the similarity is not specifically limited.

[0073] The target agent is determined from among the business agents based on their semantic similarity. For example, all business agents can be sorted in descending order of semantic similarity, and the agent with the highest similarity can be selected as the target agent. If multiple business agents have the same highest similarity, a secondary sort can be performed based on a preset business priority strategy, such as sorting according to the user's historical preferences.

[0074] The beneficial effect of this setup is that, in human-computer interaction scenarios where multiple services coexist, subjective intent recognition is transformed into objective semantic distance measurement, which solves the problem of misjudgment of business routing caused by relying on manual rules or shallow keyword matching, and improves the accuracy and robustness of target intelligent agent selection.

[0075] Figure 3 This is a schematic diagram of the logical design of an intelligent agent. Figure 3 In this system, two business intelligence agents are pre-defined: Business Intelligence Agent 1 and Business Intelligence Agent 2. Different business operations require different steps, each corresponding to a different logic block, and each logic block requires different business elements. For example, the business operation corresponding to Business Intelligence Agent 1 requires three steps. Step 1 in Business Intelligence Agent 1 corresponds to logic block 1, requiring elements 1 and 2; Step 2 corresponds to logic block 2, requiring element 3; Step 3 corresponds to logic block 3, requiring elements 4 and 5; while Step 3 in Business Intelligence Agent 2 corresponds to logic block 4, requiring elements 7 and 8. The routing intelligence agent receives user input and determines the business intelligence agent to be invoked, for example, it can route to either Business Intelligence Agent 1 or Business Intelligence Agent 2.

[0076] S204. Determine the logic block to be invoked from the target intelligent agent.

[0077] S205. Based on the logic block to be invoked, update the current business status information and output a response message to the user; wherein, the response message represents the processing result of the logic block to be invoked.

[0078] This application provides a business processing method based on human-computer interaction. Each business can correspond to a business agent for processing the corresponding business. The processing of each business involves multiple business steps, and each business process corresponds to a logic block. The business agent encapsulates multiple logic blocks, which represent the processing logic of the business steps. In one dialogue round, the user can input information multiple times. All information input by the user can be obtained as context information, and the current business status information can be obtained, which represents the progress of business processing. Based on the context information and the current business status information, the type of business that needs to be processed is determined, that is, the target agent can be determined from multiple business agents. The target agent contains multiple logic blocks. The logic block to be invoked is determined from the target agent, the invoked logic block is executed, the response information to the user and the new business status information are obtained, the business status information is updated, and the response information is output to the user, completing the processing of the current step. By pre-building multiple business intelligence agents to handle different business tasks, each business intelligence agent encapsulates multiple logic blocks. When business processing is required, the target intelligence agent is determined first, and then the target logic block is determined, making the processing flow clear. Furthermore, by combining context and processing status during the processing, the efficiency and accuracy of processing are improved, and the user's human-computer interaction experience is enhanced.

[0079] Figure 4 A flowchart illustrating a business processing method based on human-computer interaction provided in this application embodiment is shown below. Figure 4 As shown, this embodiment, based on the above embodiments, provides a detailed description of a business processing method based on human-computer interaction, which includes:

[0080] S401. Obtain the context information input by the user in the current dialogue round, and obtain the current business status information; wherein, the business status information represents the progress of business processing.

[0081] S402. Based on the context information and the current business status information, determine the target agent from at least one business agent.

[0082] S403. Based on the context information and current details, determine the target step information based on the target agent; wherein, the target step information represents the descriptive information of the business step to be executed.

[0083] For example, the business status information includes detailed information, which represents the current processing status of the business and the data being processed. Each business step's logical block corresponds to its own step description information, which represents a description of the processing logic for that business step.

[0084] Contextual information refers to multimodal input content such as natural language text, structured parameters, speech recognition transcription results, or key fields extracted by image OCR, entered by the user in the current dialogue turn. Detailed information, also known as business details, is a detailed description of the current processing status and contains more information than the summary information. Detailed information explains the specific execution status of the current business, adding various input elements compared to the summary information. For example, detailed information could be: "Executing business agent 2, currently executing step 1, element 1 has been input (value of element 1), and the user still needs to input element 2."

[0085] The target agent is the business agent that best matches the current user intent, as determined by the routing agent, and possesses a unique business identifier. By inputting contextual information and current details into the target agent, target step information can be obtained. This target step information is a semantic description text or vector representation dynamically generated by the target agent's internal step reasoning engine, representing the description of the business steps to be executed next.

[0086] In this embodiment, the content of the target step information and the operation intent expressed in the context information are related. Figure 1 For example, if a user says "I'd like to upload the receipt again," the target step information should point to "Evidence Upload" rather than "Submit for Review." The content of the target step information is also compatible with the business stage, completed elements, and blocking conditions recorded in the current details information. For example, if the details information shows "Receipt not uploaded" and "Current step status = Pending trigger," then the target step information includes the semantics of "Trigger evidence upload workflow."

[0087] In this embodiment, based on the context information and the current details information, the target step information is determined based on the target agent, including: determining the historical agent; wherein, the historical agent is the target agent previously determined in the current dialogue round; if the historical agent is different from the target agent, the details information corresponding to the target agent is obtained; and based on the context information, the current details information, and the details information corresponding to the target agent, the target step information is determined based on the target agent.

[0088] Specifically, the historical agent refers to the target business agent determined and selected by the routing agent in the most recent interaction of the current dialogue round; that is, the previous target agent. The identifier or unique ID of each determined target agent is persistently stored in the session context cache as an anchor point for state tracking. During system initialization or the first round of dialogue without prior state, the historical agent is an empty value or a preset placeholder; in this case, the historical agent is different from the newly determined target agent by default.

[0089] Determining whether the historical agent and the target agent are the same involves determining whether they belong to the same business agent. For example, this can be done using the business agent's ID.

[0090] If the historical agent differs from the target agent, it indicates that when the historical agent was the target agent, the business state information at that time was updated; that is, the current business state information is the updated business state information. For the newly identified target agent, obtain the detailed information corresponding to the target agent. For example, pre-store the detailed information corresponding to the target agent, or preset default detailed information, and use it as the detailed information of the target agent.

[0091] The context information, current details, and details corresponding to the target agent are input into the target agent to obtain the target step information. In this embodiment, the processing logic of the target agent is not specifically limited. For example, the three types of information can be aligned at the field level first, and then the preconditions of which steps have been met can be determined based on the business step dependency graph. Finally, the target step information is output.

[0092] The beneficial effects of this setup are that the identification of historical intelligent agents constitutes a prerequisite for service switchover detection, and the comparison of their unique identifiers ensures the certainty of the judgment. Obtaining detailed information about the target intelligent agent is a key action in cross-service context inheritance, and the fusion of three sources of information ensures the robustness of the target step selection. This achieves accurate, stable, and reproducible determination of target step information in multi-service dynamic switching scenarios, improving system real-time performance and user experience consistency while ensuring the integrity of business processing.

[0093] In this embodiment, obtaining the detailed information corresponding to the target intelligent agent includes: searching for the detailed information associated with the target intelligent agent stored in a preset data table; if the detailed information associated with the target intelligent agent is not stored in the preset data table, then the default detailed information is obtained, which is the detailed information corresponding to the target intelligent agent.

[0094] Specifically, the default data table is a relational or key-value table maintained by the structured state persistent storage module. Its structure includes at least three types of fields: a business agent identifier field (used to uniquely identify each business agent), a business status information field (used to store the latest business status information of the business agent), and an update timestamp field (recording the millisecond-level timestamp of the last write or refresh of the business status information). This default data table can be deployed on a local embedded database, an in-memory database, or a distributed state storage service, supporting high-concurrency read and atomic write operations.

[0095] The system retrieves detailed information associated with the target intelligent agent from a pre-defined data table. For example, it can initiate a primary key query operation on the pre-defined data table based on the unique identifier of the target intelligent agent. If the corresponding record is found, the detailed information can be directly parsed from the corresponding business status information to obtain the detailed information associated with the target intelligent agent.

[0096] If the target agent is not associated with any detailed information in the preset data table, it means that the target agent is being selected for the first time in this round of the session. At this time, there is no record of the agent in the preset data table, or the agent may have been recorded before, but its corresponding record has been cleared, so the corresponding detailed information cannot be found.

[0097] Default details information is preset, for example, the default details information is empty. If the preset data table does not contain details information associated with the target agent, the default details information is directly obtained and used as the details information corresponding to the target agent.

[0098] The beneficial effect of this setting is that when switching services, the detailed information of the new target intelligent agent can be obtained quickly and effectively. This effectively solves the problems of process interruption, element extraction failure or aggravation of large model illusion caused by state loss in multi-service dynamic switching scenarios, and significantly improves the startup robustness, context recovery efficiency and cross-service collaboration stability of the human-computer interaction system.

[0099] In this embodiment, the preset data table includes multiple data rows; obtaining the current business status information includes: if the current business status information does not exist, the current business status information is set to a preset value, and a new data row is constructed in the preset data table; wherein the new data row stores the current business status information.

[0100] Specifically, the preset data table can be a persistent / semi-persistent state storage structure with a clear row structure, field semantics, and lifecycle management capabilities. Each data row corresponds to a state snapshot of an independent business instance, including but not limited to fields such as the business agent's identifier, business agent type, current summary information, current details information, field values ​​of each business element, creation timestamp, and last update timestamp.

[0101] In step S401, it is necessary to obtain the current business status information. This can be obtained directly from a preset timetable, using the business status information with the most recent update timestamp as the current business status information. If the preset data table is empty, it means that the user has entered information for the first time in the current dialogue round, and the current business status information does not exist in the preset data table.

[0102] A preset value is configured, which can be empty. If it is determined that the current business status information does not exist in the preset data table, the current business status information can be directly set to the preset value. That is, the current business status information includes empty summary information, empty detail information, empty business element information, etc.

[0103] If it is determined that there is no current business status information, a new data row can be created in the preset data table as a new data row, and the current business status information can be recorded in the new data row. At this time, the target intelligent agent has not yet been determined, therefore, the new data row does not contain the identifier of the associated business intelligent agent.

[0104] The advantage of this setup is that, in the absence of historical states, it automatically triggers a standardized initialization process, assigns a dedicated data row carrier to each newly started business session, and ensures that the system is always at an operable, traceable, and auditable operating baseline, avoiding increased maintenance costs caused by fragmented state storage.

[0105] In this embodiment, the method further includes: after determining the target intelligent agent from at least one business intelligent agent, adding the target intelligent agent to the new data row and storing it in association with the current running status information; after updating the current business status information according to the logic block to be invoked, replacing the current business status information in the new data row with the updated business status information.

[0106] Specifically, a new data row refers to a newly constructed structured record unit in a preset data table for the current dialogue round. This data row is persistently stored in a local memory cache or a lightweight embedded database in the form of key-value pairs or relational database row records, supporting atomic write operations. The identifier of the target agent can be a globally unique string or a business semantic name, used to accurately point to a specific business agent instance deployed in the electronic device.

[0107] After step S402 is completed, and the target agent is determined from at least one business agent based on the context information and the current business status information, the identifier of the target agent is obtained. Based on the identifier, the target agent can be added to a new data row; that is, the identifier of the target agent can be written into the new data row, thus associating the target agent with the current business status information.

[0108] After executing step S405, the current business status information is updated according to the logic block to be called. The updated business status information is then obtained, and the current business status information in the newly added data row is replaced with the updated business status information. That is, the data row only stores the latest business status information.

[0109] During human-computer interaction, users can switch services. When switching to a new service, the target agent is identified. If the business status information of the target agent does not exist in the preset data table, a new data row can be added to the preset data table while obtaining the default details. The identifier and preset value of the target agent are associated and stored in the new data row, which facilitates the subsequent updating of the business status information in the new data row.

[0110] The beneficial effect of this setup is that it achieves strong consistency assurance throughout the entire process of business intelligence agent selection and state change, adapts to typical scenarios of multi-business switching, and significantly improves the reliability and maintainability of human-computer interaction.

[0111] In this embodiment, the method further includes: if the number of data rows in the preset data table reaches a preset threshold, then redundant rows are determined from each data row and deleted based on the update time of each data row in the preset data table.

[0112] Specifically, the system can count the number of data rows in the current preset data table in real time or at set intervals. A preset threshold is set, and when the number of data rows is greater than or equal to the preset threshold, the process of identifying and cleaning up redundant rows is triggered. Redundant rows are data rows that need to be deleted.

[0113] Each data row in the preset data table is associated with an update timestamp, representing the last update time of the data row. All data rows are traversed, and the last update time recorded in each data row is read. Based on the update time of each data row, one or more data rows that need to be deleted are identified as redundant rows, and thus removed from the preset data table. For example, the data row with the oldest update time can be designated as a redundant row. Alternatively, the time difference between the update time of a data row and the current time can be calculated; if the time difference is greater than a preset difference threshold, the data row is identified as a redundant row.

[0114] The beneficial effect of this setting is that, during multi-business human-computer interaction tasks, when the number of business status data rows accumulated in the preset data table reaches the preset resource protection threshold, the system automatically identifies the stale status data that has not been called or updated by the business intelligence agent for the longest time based on the last update time of each data row, and performs controllable logical or physical deletion operations on it, which significantly reduces the peak memory usage, shortens the average time of status retrieval, and improves the long-term stability of the system.

[0115] In this embodiment, target step information is determined based on the target intelligent agent according to context information, current details information, and details information corresponding to the target intelligent agent. This includes: determining target information based on the current details information and details information corresponding to the target intelligent agent; wherein, target information represents the current details information and details information corresponding to the target intelligent agent; and determining target step information based on the target intelligent agent according to context information and target information.

[0116] Specifically, if a business switch is confirmed, the current details information and the details information corresponding to the target agent can be obtained. The current details information is the last updated details information of the historical agent. Combining the current details information and the details information corresponding to the target agent, the target information is obtained. The target information is an intermediate semantic representation generated after the above two types of details information are fused. Its essence is the enhanced context input required for the target agent's decision-making. The fusion methods include but are not limited to: (a) Field-level merging: taking the latest value for fields with the same name and retaining the union of fields with different names; (b) Semantic alignment concatenation: mapping the two pieces of details information into 768-dimensional vectors through a lightweight encoder, and then generating a joint embedding vector through weighted averaging or attention concatenation; (c) Structured summary generation: calling a small language model in a prompting engineering manner, taking the two pieces of details information as input, and generating a natural language summary of no more than 128 characters, for example, "Currently processing order request, transaction date and card number have been provided; previous error appeal has filled in the appeal reason 'merchant has not shipped', but has not yet been submitted."

[0117] Contextual and target information are input into the target agent, and target step information is obtained through joint reasoning. Target step information can be a short natural language description, such as "Please upload a screenshot of the transaction receipt." This target step information directly drives subsequent logic block calls, and its accuracy depends on the collaborative modeling capability of context and target information.

[0118] The beneficial effect of this setup is that, in business switching scenarios, by fusing the current details information with the details information corresponding to the target intelligent agent to generate target information, and using it together with the current context information as input to the target intelligent agent, the system can accurately understand the cross-business state transition path, thereby accurately identifying the business steps to be executed, improving the robustness of step decision-making in the multi-business collaborative processing process, and thus enhancing the coherence and naturalness of human-computer interaction.

[0119] In this embodiment, the target step information is determined based on the target agent according to the context information and the current details information, including: if the historical agent is the same as the target agent, the current details information is determined as the target information; and the target step information is determined based on the target agent according to the context information and the target information.

[0120] Specifically, if the historical agent is the same as the target agent, it means that no service switch has been performed. If it is determined that no service switch has been performed, the current details can be directly identified as the target information.

[0121] Contextual and target information are input into the target agent, specifically the contextual information and current details, to obtain the target step information. In other words, the historical agent's recognition results determine whether to directly input current details; this decision directly constrains the data source for the target information and ultimately leads to the generation of the target step information.

[0122] The advantage of this setup is that it enables precise simplification of the context construction process in continuous processing scenarios where business operations do not switch, thereby improving the real-time performance and stability of continuous business processing.

[0123] S404. Based on the target step information and the step description information corresponding to the logical blocks of each business step in the target intelligent agent, determine the logical blocks to be invoked.

[0124] For example, the step description information is metadata pre-configured for each business step logic block, representing the processing logic of the business step. When the target step information is determined, it can be matched with the step description information corresponding to the logic blocks of each business step in the target agent to determine the logic block to be invoked.

[0125] For example, a two-level matching mechanism can be adopted. The first level is semantic coarse screening, where the target step information is input into the semantic tag set of the step description information for keyword matching or vector similarity calculation, filtering out multiple logical blocks with high similarity to form a candidate logical block set. The second level is constraint fine matching, which traverses the candidate logical block set and verifies whether the input parameter types of each logical block in the set can cover the field types that can be extracted from the context information, and whether the pre-validation rules are compatible with the business element status recorded in the current details information. For example, if the "disputed amount" field already exists in the details information and is not empty, then logical blocks that require input amount are excluded. Finally, the logical block with the highest unique matching degree and complete constraint satisfaction is selected as the logical block to be called. In this embodiment, when multiple logical blocks satisfy the constraints, a preset priority strategy can be used, such as descending order by version number or ascending order by average response time; or when there is no completely matching logical block, a manual review can be prompted.

[0126] This embodiment achieves accurate, robust, and scalable dynamic scheduling of business steps within the target intelligent body in multi-service hybrid dialogue scenarios. It solves the problem of inaccurate step recognition caused by relying solely on summary text. It can accurately identify fine-grained operation intentions such as "supplement materials", "withdraw application", and "switch to negotiation mode" raised by users in complex error processes, significantly improving the efficiency and accuracy of business processing and enhancing user experience.

[0127] S405. Based on the logic block to be invoked, update the current business status information and output a response message to the user; wherein, the response message represents the processing result of the logic block to be invoked.

[0128] For example, if no business switching occurs, the context information and current details can be input into the logic block to be invoked. After calculation by the logic block, the updated business status information and response information are obtained. For instance, by invoking the logic block to be invoked, the business element to be confirmed is determined. The business element represents the type of information required by the logic block when performing business processing. If the context information and current details do not contain the element value of the business element to be confirmed, feedback information is sent to the user to prompt the user to input the element value of the business element to be confirmed. Based on the element value of the business element to be confirmed input by the user, the current business status information is updated.

[0129] In this embodiment, updating the current business status information according to the logic block to be invoked includes: determining the business element to be confirmed by invoking the logic block to be invoked; wherein, the business element represents the type of information required by the logic block when performing business processing; if the context information and target information do not contain the element value of the business element to be confirmed, then feedback information is sent to the user; wherein, the feedback information is used to prompt the user to input the element value of the business element to be confirmed; and updating the current business status information according to the element value of the business element to be confirmed input by the user.

[0130] Specifically, the logic block to be invoked refers to the executable functional unit selected from the target business intelligence agent that corresponds to the current business step. It encapsulates the complete business rules, verification logic, and data transformation logic for that step. Business elements are the fields required by the logic block during execution. For example, in an error dispute handling scenario, for the "submit a request for a dispute" step, its business elements may include: dispute number, original transaction time, merchant name, dispute reason code, and certificate file hash value. Each business element has a unique identifier, data type, constraints (such as length, format, and value range), semantic tags (such as "transaction time" and "responsible party"), and whether default filling is supported.

[0131] Each logic block is configured with the types of business elements it requires. By calling the logic block to be called, all the necessary business elements can be identified as those to be confirmed. For example, this can be achieved by reading the description file embedded within the logic block.

[0132] The context information and target information are processed through semantic recognition to determine the business elements contained within them. In this embodiment, the semantic recognition algorithm is not specifically limited. If all the business elements to be confirmed can be found in the context information and target information, their values ​​can be input into the logic block to be invoked. The logic block processes these elements to obtain the processing results of the business steps, and then updates the business status information and generates response information based on the processing results.

[0133] For any business element to be confirmed, if neither the context information nor the target information contains the element value of that business element, feedback information can be sent to the user through human-computer interaction. This feedback information prompts the user to input the element value of the business element to be confirmed. That is, the feedback information is a user-oriented natural language prompt, such as, "Please provide the original transaction time involved in this order adjustment (format: YYYY-MM-DD HH:MM:SS)". In this embodiment, the feedback information can be configured with a hierarchical guidance strategy, prompting only the most urgent element at a time, and triggering the confirmation of the next element only after the user responds, thus avoiding information overload.

[0134] Users respond to feedback information, for example, by directly entering text, clicking preset options, or uploading files. Upon receiving user feedback, and confirming the values ​​of all pending business elements, the system can begin processing the business steps and obtain updated business status information.

[0135] The advantage of this setup is that it proactively identifies the necessary input elements before executing any business logic block and performs a completeness check based on currently available information. When a missing item is detected, it accurately locates the missing item and generates semantic and context-aware interactive feedback to guide the user to complete the task with minimal cognitive load, thereby improving the reliability, predictability, and user guidance efficiency of business processing in multi-business mixed dialogue scenarios.

[0136] This application provides a business processing method based on human-computer interaction. Each business can correspond to a business agent for processing the corresponding business. The processing of each business involves multiple business steps, and each business process corresponds to a logic block. The business agent encapsulates multiple logic blocks, which represent the processing logic of the business steps. In one dialogue round, the user can input information multiple times. All information input by the user can be obtained as context information, and the current business status information can be obtained, which represents the progress of business processing. Based on the context information and the current business status information, the type of business that needs to be processed is determined, that is, the target agent can be determined from multiple business agents. The target agent contains multiple logic blocks. The logic block to be invoked is determined from the target agent, the invoked logic block is executed, the response information to the user and the new business status information are obtained, the business status information is updated, and the response information is output to the user, completing the processing of the current step. By pre-building multiple business intelligence agents to handle different business tasks, each business intelligence agent encapsulates multiple logic blocks. When business processing is required, the target intelligence agent is determined first, and then the target logic block is determined, making the processing flow clear. Furthermore, by combining context and processing status during the processing, the efficiency and accuracy of processing are improved, and the user's human-computer interaction experience is enhanced.

[0137] Figure 5 A flowchart illustrating a business processing method based on human-computer interaction provided in this application embodiment is shown below. Figure 5 As shown, this embodiment, based on the above embodiments, provides a detailed description of a business processing method based on human-computer interaction, which includes:

[0138] S501, User Input.

[0139] S502. Determine if the business status exists, that is, determine if there is current business status information. If not, proceed to step S503; if yes, proceed to step S504.

[0140] S503. Add business status information to the preset data table and delete business status information that has not been updated for a long time as needed.

[0141] S504. The routing agent routes to the required business agent, i.e. the target agent, based on the business summary and context information.

[0142] S505. Determine whether to switch services. If yes, execute S506; otherwise, execute S507.

[0143] S506. Obtain detailed information about the target agent.

[0144] S507. Combining the current details information and the details information corresponding to the target agent, determine the target information, which is the details information that ultimately needs to be processed.

[0145] S508: The target intelligent agent calls the logic block of the business steps based on the target information and context information.

[0146] S509. The logic block of the business step updates the business status information.

[0147] Figure 6 This is a schematic diagram of a human-computer interaction-based business processing device provided in an embodiment of this application. The device is applied to an electronic device, which deploys at least one business intelligence agent. The business intelligence agent is used to process corresponding business processes. Each business intelligence agent encapsulates at least one logic block representing a business step. Each business step corresponds one-to-one with a logic block, and the logic block represents the processing logic of the business step. For example... Figure 6 As shown, the business processing device 60 based on human-computer interaction provided in this embodiment includes:

[0148] The information acquisition unit 601 is used to acquire contextual information input by the user in the current dialogue round, and to acquire current business status information; wherein, the business status information represents the progress of business processing.

[0149] The target determination unit 602 is used to determine a target intelligent agent from at least one business intelligent agent based on context information and current business state information, and to determine the logic block to be invoked from the target intelligent agent.

[0150] The information update unit 603 is used to update the current business status information according to the logic block to be invoked, and output reply information to the user; wherein, the reply information represents the processing result of the logic block to be invoked.

[0151] In one possible implementation, the electronic device deploys a routing agent, which routes to a service agent based on service status information and context information; the service agent corresponds to service description information, which characterizes the description information of the service corresponding to the service agent; the target determination unit 602 is specifically used for:

[0152] Based on contextual information and current business status information, the target description information is determined by the routing agent; whereby the target description information represents the description information of the business that the user needs to process.

[0153] Based on the target description information and the business description information corresponding to each business agent, the target agent is determined from among the business agents.

[0154] In one possible implementation, the business status information includes summary information, which represents a summary of the current processing status of the business; the target determination unit 602 is specifically used for:

[0155] The context information and the current summary information are input into the routing agent to obtain the target description information.

[0156] In one possible implementation, the target determination unit 602 is specifically used for:

[0157] For each business intelligent agent, determine the semantic similarity between the target description information and the business description information corresponding to the business intelligent agent.

[0158] The target agent is determined from among the business agents based on the semantic similarity of each business agent.

[0159] In one possible implementation, the business status information includes detailed information, which represents the current processing status of the business and the data being processed; the logical blocks of the business steps correspond to step description information, which represents a description of the processing logic for the business steps; the target determination unit 602 is specifically used for:

[0160] Based on the context information and current details, the target step information is determined based on the target agent; whereby the target step information represents the descriptive information of the business step to be executed.

[0161] Based on the target step information and the step description information corresponding to the logical blocks of each business step in the target intelligent agent, the logical blocks to be invoked are determined.

[0162] In one possible implementation, the target determination unit 602 is specifically used for:

[0163] Identify the historical agent; where the historical agent is the target agent identified in the previous dialogue round.

[0164] If the historical agent is different from the target agent, obtain the detailed information corresponding to the target agent;

[0165] Based on the context information, the current details, and the details corresponding to the target agent, the target step information is determined.

[0166] In one possible implementation, the target determination unit 602 is specifically used for:

[0167] Retrieve detailed information associated with the target agent from the preset data table;

[0168] If the preset data table does not contain detailed information associated with the target agent, then the default detailed information is obtained, which is the detailed information corresponding to the target agent.

[0169] In one possible implementation, the preset data table includes multiple data rows; the information acquisition unit 601 is specifically used for:

[0170] If the current business status information does not exist, the current business status information is set to a preset value, and a new data row is created in the preset data table; the new data row stores the current business status information.

[0171] One possible implementation also includes:

[0172] The storage unit is used to add the target agent to the new data row after the target agent is determined from at least one business agent, and to store it in association with the current running status information.

[0173] The replacement unit is used to replace the current business status information in the newly added data row with the updated business status information after updating the current business status information according to the logic block to be called.

[0174] One possible implementation also includes:

[0175] The deletion unit is used to determine redundant rows from each data row in the preset data table and delete them if the number of data rows in the preset data table reaches a preset threshold.

[0176] In one possible implementation, the target determination unit 602 is specifically used for:

[0177] Based on the current detailed information and the detailed information corresponding to the target intelligent agent, the target information is determined; whereby the target information represents the current detailed information and the detailed information corresponding to the target intelligent agent.

[0178] Based on contextual and target information, target step information is determined using the target agent.

[0179] In one possible implementation, the target determination unit 602 is specifically used for:

[0180] If the historical agent is the same as the target agent, then the current details are determined as the target information;

[0181] Based on contextual and target information, target step information is determined using the target agent.

[0182] In one possible implementation, the information update unit 603 is specifically used for:

[0183] By invoking the logic block to be invoked, the business elements to be confirmed are determined; whereby the business elements represent the types of information required by the logic block when performing business processing.

[0184] If the context information and target information do not contain the element value of the business element to be confirmed, a feedback message is sent to the user; the feedback message is used to prompt the user to enter the element value of the business element to be confirmed.

[0185] Update the current business status information based on the element value of the business element to be confirmed input by the user.

[0186] This embodiment provides a business processing device based on human-computer interaction, which can execute the methods provided in the above-described method embodiments. Its implementation principle and technical effects are similar, and will not be described in detail here.

[0187] Figure 7 This is a schematic diagram of the structure of an electronic device provided in this application. Figure 7 As shown, the electronic device 70 provided in this embodiment includes at least one processor 701 and a memory 702. Optionally, the device 70 further includes a communication component 703. The processor 701, memory 702, and communication component 703 are connected via a bus 704.

[0188] In a specific implementation, at least one processor 701 executes computer execution instructions stored in memory 702, causing at least one processor 701 to perform the above-described method.

[0189] The specific implementation process of processor 701 can be found in the above method embodiments, and its implementation principle and technical effect are similar. It will not be repeated here.

[0190] In the above embodiments, it should be understood that the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor.

[0191] The memory may include random access memory (RAM) and may also include non-volatile memory (NVM), such as at least one disk storage device.

[0192] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, the buses shown in the accompanying drawings are not limited to a single bus or a single type of bus.

[0193] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0194] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the above-described method.

[0195] The aforementioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The readable storage medium can be any available medium accessible to a general-purpose or special-purpose computer.

[0196] An exemplary readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can reside in an Application Specific Integrated Circuit (ASIC). Alternatively, the processor and the readable storage medium can exist as discrete components in the device.

[0197] The division of units is merely a logical functional division; in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0198] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0199] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0200] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0201] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0202] Finally, it should be noted that other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This invention is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein, and is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.

Claims

1. A business processing method based on human-computer interaction, characterized in that, The method is applied to an electronic device, which deploys at least one business intelligence agent. The business intelligence agent is used to process corresponding business processes. Each business intelligence agent encapsulates at least one logic block of a business step, with each business step corresponding one-to-one with the logic block. The logic block represents the processing logic of the business step. The method includes: Obtain the context information input by the user in the current dialogue round, and obtain the current business status information; wherein, the business status information represents the progress of business processing; Based on the context information and the current business status information, a target intelligent agent is determined from the at least one business intelligent agent, and a logic block to be invoked is determined from the target intelligent agent; Based on the logic block to be invoked, update the current business status information and output a response message to the user; wherein the response message represents the processing result of the logic block to be invoked.

2. The method according to claim 1, characterized in that, The electronic device is equipped with a routing agent, which is used to route to a service agent based on service status information and context information; the service agent corresponds to service description information, which represents the description information of the service corresponding to the service agent. Determining the target agent from the at least one business agent based on the context information and the current business state information includes: Based on the context information and the current service status information, the target description information is determined by the routing agent; wherein the target description information represents the description information of the service that the user needs to process. Based on the target description information and the business description information corresponding to each business agent, the target agent is determined from each business agent.

3. The method according to claim 2, characterized in that, The business status information includes summary information, which represents a summary of the current processing status of the business. Based on the context information and the current service status information, and using the routing agent, target description information is determined, including: The context information and the current summary information are input into the routing agent to obtain the target description information.

4. The method according to claim 2, characterized in that, Based on the target description information and the business description information corresponding to each business agent, the target agent is determined from each business agent, including: For each business intelligent agent, determine the semantic similarity between the target description information and the business description information corresponding to the business intelligent agent. The target agent is determined from the business agents based on the semantic similarity of each business agent.

5. The method according to claim 1, characterized in that, The business status information includes detailed information, which represents the current processing status of the business and the data being processed; the logical blocks of the business steps correspond to step description information, which represents the description information of the processing logic of the business steps; Determining the logic block to be invoked from the target agent includes: Based on the context information and current details, target step information is determined based on the target agent; wherein, the target step information represents the descriptive information of the business step to be executed; Based on the target step information and the step description information corresponding to the logical blocks of each business step in the target intelligent agent, the logical block to be invoked is determined.

6. The method according to claim 5, characterized in that, Based on the context information and current details, and based on the target agent, target step information is determined, including: Identify the historical agent; wherein, the historical agent is the target agent previously identified in the current dialogue round; If the historical agent is different from the target agent, then obtain the detailed information corresponding to the target agent; Based on the context information, the current details information, and the details information corresponding to the target agent, the target step information is determined based on the target agent.

7. The method according to claim 6, characterized in that, Obtain detailed information about the target agent, including: Retrieve detailed information associated with the target intelligent agent from a preset data table; If the preset data table does not contain any detailed information associated with the target intelligent agent, then the default detailed information is obtained, which is the detailed information corresponding to the target intelligent agent.

8. The method according to claim 7, characterized in that, The preset data table includes multiple data rows; obtaining the current business status information includes: If the current business status information does not exist, the current business status information is set to a preset value, and a new data row is created in the preset data table; wherein the new data row stores the current business status information.

9. The method according to claim 8, characterized in that, Also includes: After determining the target agent from the at least one business agent, the target agent is added to the new data row and stored in association with the current running status information; After updating the current business status information according to the logic block to be invoked, the current business status information in the newly added data row is replaced with the updated business status information.

10. The method according to claim 8, characterized in that, Also includes: If the number of data rows in the preset data table reaches a preset threshold, redundant rows are identified and deleted based on the update time of each data row in the preset data table.

11. The method according to claim 6, characterized in that, Based on the context information, the current details information, and the details information corresponding to the target agent, the target step information is determined, including: Target information is determined based on the current details and the details corresponding to the target agent; wherein, the target information represents the current details and the details corresponding to the target agent. Based on the context information and the target information, target step information is determined based on the target agent.

12. The method according to claim 6, characterized in that, Based on the context information and current details, and based on the target agent, target step information is determined, including: If the historical agent is the same as the target agent, then the current details are determined as the target information; Based on the context information and the target information, target step information is determined based on the target agent.

13. The method according to claim 11 or 12, characterized in that, Based on the logic block to be invoked, update the current business status information, including: By invoking the logic block to be invoked, the business elements to be confirmed are determined; wherein, the business elements represent the types of information required by the logic block when performing business processing; If the context information and the target information do not contain the element value of the business element to be confirmed, then feedback information is sent to the user; wherein, the feedback information is used to prompt the user to input the element value of the business element to be confirmed; The current business status information is updated based on the element value of the business element to be confirmed input by the user.

14. A business processing device based on human-computer interaction, characterized in that, The device is applied to an electronic device, which deploys at least one business intelligence agent. The business intelligence agent is used to process corresponding business processes. Each business intelligence agent encapsulates at least one logic block representing a business step. Each business step corresponds one-to-one with a logic block, and each logic block represents the processing logic of a business step. The device includes: The information acquisition unit is used to acquire contextual information input by the user in the current dialogue round, and to acquire current business status information; wherein, the business status information represents the progress of business processing. The target determination unit is configured to determine a target intelligent agent from the at least one business intelligent agent based on the context information and the current business status information, and to determine the logic block to be invoked from the target intelligent agent; The information update unit is used to update the current business status information according to the logic block to be invoked, and output reply information to the user; wherein the reply information represents the processing result of the logic block to be invoked.

15. An electronic device, characterized in that, include: Memory, processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory, causing the processor to perform the method as described in any one of claims 1-13.

16. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any one of claims 1-13.

17. A computer program product, characterized in that, Includes a computer program that, when executed by a processor, implements the method described in any one of claims 1-13.

Images