Software operation method, device and equipment based on intelligent agent and readable storage medium

Abstract

The invention discloses a software operation method, device and equipment based on an intelligent agent and a readable storage medium. The method comprises the steps that initialization parameters of the intelligent agent are configured; in response to a query instruction executed by a user in the configured agent, obtaining a to-be-answered task, preprocessing the to-be-answered task based on the agent, and generating a standard task field; performing complexity calculation on the standard task field through a task classification function, and determining a task level of the standard task field; judging whether the task level of the standard task field needs to call an external tool or not; if yes, a target external tool corresponding to the task level is determined through the matching degree of the standard task field and a preset tool, and target answer information of the standard task field is generated based on the target external tool; if not, target answer information of the standard task field is generated based on a local knowledge base and an internal algorithm, so that software operation needed by the user is achieved. Through interaction between the intelligent agent and the real environment, the task processing efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the field of computer application technology, and in particular to an agent-based software operation method, device, equipment and readable storage medium. Background Art

[0002] In the current office environment, computer technology has been involved in all aspects of work. Among them, software operation, as one of the indispensable skills, is widely used in many fields such as office, design, and development. Mastering the basic knowledge of software operation is crucial to improving work efficiency.

[0003] The main operating buttons of conventional software operating systems include open, click text, click icon, page up and down, input, check, return, exit, etc. The activation of each button requires the manual operation of engineers to upload the required data information and perform formula calculations. This process is time-consuming and prone to errors.

[0004] There is currently no effective solution to the technical problem that manual software operation in the existing technology leads to low efficiency and accuracy. Summary of the invention

[0005] The purpose of the present invention is to provide a software operation method, device, equipment and readable storage medium based on an intelligent agent, which can solve the technical problems of low efficiency and accuracy caused by manual software operation in the prior art.

[0006] One aspect of the present invention provides a software operation method based on an intelligent agent, the method comprising: configuring initialization parameters of the intelligent agent, wherein the initialization parameters include framework model parameters, demand calling tools, dialogue management parameters, dialogue state values, and query template fields; in response to a query instruction executed by a user in a configured intelligent agent, obtaining tasks to be answered, preprocessing the tasks to be answered based on the intelligent agent, and generating standard task fields; calculating the complexity of the standard task fields through a task classification function, and determining the task level of the standard task fields, wherein the task levels include simple, medium, and complex levels; determining whether the task level of the standard task fields requires calling external tools; if so, determining the target external tool corresponding to the task level through the matching degree between the standard task fields and the preset tools, and generating target answer information of the standard task fields based on the target external tool; if not, generating the target answer information of the standard task fields based on a local knowledge base and an internal algorithm to implement the software operation required by the user.

[0007] Optionally, target answer information of the standard task field is generated based on the target external tool, including: sorting the target external tools to generate a tool processing sequence; starting from the first tool, processing the standard task field; using the output result of the previous tool as the input data of the next tool in the order corresponding to the tool processing sequence, and repeating this step until the processing result of the last tool in the tool processing sequence is obtained; and determining the processing result of the last tool in the tool processing sequence as the target answer information of the standard task field.

[0008] Optionally, target answer information of a standard task field is generated based on a target external tool, including: generating a topological structure through the content correlation of the standard task field, wherein the topological structure includes a root node and child nodes, and each child node corresponds to a subtask; determining subtasks that can be executed in parallel in the topological structure; simultaneously calling external tools corresponding to the subtasks that can be executed in parallel to calculate the subtasks; and summarizing the calculation results of all subtasks of the standard task field to generate the target answer information of the standard task field.

[0009] Optionally, the standard task field includes multiple subtask description fields, and the target external tool corresponding to the task level is determined by the matching degree between the standard task field and the preset tool, including: converting the subtask description field through the trained semantic representation model to generate a subtask semantic vector; calculating the similarity score between the subtask semantic vector and the semantic vector of the preset tool description, and screening out the preset tools with a similarity score higher than a preset threshold as candidate tools; calculating the question-answer matching score between the subtask semantic vector and the semantic vector of the candidate tool description, and screening out the candidate tool with the highest question-answer matching score as the target external tool.

[0010] Optionally, the task to be answered is preprocessed based on the intelligent agent to generate a standard task field, including: concatenating the task to be answered with the query template field to generate a sentence to be processed; performing word segmentation on the sentence to be processed using a word segmentation module to obtain multiple feature words; converting the feature words into feature word vectors, and inputting the feature word vectors into an encoder to generate a vector representation containing context information; decoding the vector representation containing context information to generate a standard task field.

[0011] Optionally, the task classification function is:

[0012] Among them, X is the semantic expression degree value of the word segmentation in the standard task field, and Y is the word segmentation statistics in the standard task field. is the balance factor of word segmentation semantics, is the balancing factor of the word segmentation statistics.

[0013] Another aspect of the present invention provides a software operation device based on an intelligent agent, which includes: a configuration module for configuring the initialization parameters of the intelligent agent, wherein the initialization parameters include framework model parameters, demand calling tools, dialogue management parameters, dialogue state values, and query template fields; a preprocessing module for responding to query instructions executed by a user in a configured intelligent agent, obtaining tasks to be answered, preprocessing the tasks to be answered based on the intelligent agent, and generating standard task fields; a task level determination module for calculating the complexity of the standard task field through a task classification function to determine the task level of the standard task field, wherein the task level includes simple level, medium level and complex level; a tool determination module for determining whether the task level of the standard task field requires calling an external tool; a first generation module for, if yes, determining the target external tool corresponding to the task level through the matching degree between the standard task field and the preset tool, and generating target answer information of the standard task field based on the target external tool; and a second generation module for, if no, generating the target answer information of the standard task field based on the local knowledge base and internal algorithm to implement the software operation required by the user.

[0014] Optionally, the first generation module is also used to: sort the target external tools to generate a tool processing sequence; start from the first tool to process the standard task field; use the output result of the previous tool as the input data of the next tool in the order corresponding to the tool processing sequence, and repeat this step until the processing result of the last tool in the tool processing sequence is obtained; determine the processing result of the last tool in the tool processing sequence as the target answer information of the standard task field.

[0015] Another aspect of the present invention provides a computer device, which includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the agent-based software operation method of any of the above-mentioned embodiments when executing the computer program.

[0016] Another aspect of the present invention provides a computer storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the agent-based software operation method of any of the above embodiments is implemented. Further, the computer-readable storage medium may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function, etc.; the data storage area may store data created according to the use of a blockchain node, etc.

[0017] The present invention automatically generates a coherent and executable operation plan based on user questions through the interaction between the intelligent agent and the real environment, and sends it to the back-end program, which initiates a response without the need to perceive the operation interface, effectively reducing the difficulty of understanding and reasoning large models and making it easy to implement. Through the one-to-one correspondence between task levels and tools, appropriate tools are selected to process each task, thereby improving the efficiency of processing user task requirements. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Moreover, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings: Figure 1 An optional flow chart of the agent-based software operation method provided in the first embodiment of the present invention is shown; Figure 2 shows a structural block diagram of an agent-based software operating device provided in Embodiment 2 of the present invention; and Figure 3 A block diagram of a computer device suitable for implementing an agent-based software operation method provided in Embodiment 3 of the present invention is shown. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical scheme and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0020] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0021] Embodiment 1 This embodiment provides a software operation method based on an agent. Figure 1 A flow chart of the agent-based software operation method is shown, as shown in Figure 1As shown, the agent-based software operation method may include steps S101 to S106, wherein: Step S101, configuring the initialization parameters of the agent, wherein the initialization parameters include framework model parameters, demand calling tools, dialogue management parameters, dialogue state values, and query template fields; Before implementing software operations, a series of initialization settings are required for the system to ensure smooth and efficient interaction.

[0022] Specifically, 1) configure the large language model. The framework model parameters can be the LLM model framework; in the startup phase, the system loads and configures a high-level language processing model designed for user interaction, which is responsible for understanding the user's input task and generating answers that match the task, thereby solving user needs.

[0023] 2) Integration of data processing tools and services. The required calling tools can be database access components, network search components, and third-party API interfaces. According to the requirements that may be involved in the interaction, the intelligent experience will preset a series of external tools and services that may need to be called.

[0024] 3) Setting of dialogue management parameters: The dialogue management parameters can be the context length of the dialogue, the threshold of user intent recognition, and the dialogue strategy.

[0025] 4) Initialization of the conversation state. At the beginning of each conversation, the intelligent assistant will initialize the conversation state and clear the historical conversation records to ensure that the new conversation is not affected by the previous interaction, thus providing users with a brand new interactive experience.

[0026] 5) Define ask (query template). This query template is the strategy for the agent to process user queries. It automatically integrates preset queries into the original query, subdivides complex tasks into multiple clear subtasks, and can be applied to query requirements in different scenarios.

[0027] The implementation of this step enables the intelligent system to effectively and accurately understand user needs, enabling it to solve these subtasks one by one, and finally summarize the answers to each subtask to form a comprehensive and in-depth response to the original query.

[0028] Step S102, in response to a query instruction executed by a user in a configured agent, obtaining tasks to be answered, preprocessing the tasks to be answered based on the agent, and generating standard task fields; The initial task input by the user may contain unclear semantics, typos, and reversed logic, which are not easy to be effectively recognized by the intelligent agent. Therefore, it is necessary to preprocess the tasks to be answered and generate standard task fields to facilitate the rapid recognition of the intelligent agent.

[0029] Step S103, performing complexity calculation on the standard task field through a task classification function to determine the task level of the standard task field, wherein the task level includes simple level, medium level and complex level; The task classification function may be a random forest, a support vector machine, a decision tree, a logistic regression, or other algorithms that can perform data classification, which are not limited here.

[0030] In actual application scenarios, the needs of each user are different, some are easy to handle while others are not. By calculating the complexity of user needs and classifying standard task fields, we can handle different user needs in a targeted manner, thereby improving the efficiency of software operations.

[0031] Step S104, determining whether the task level of the standard task field requires calling an external tool; In order to improve the accuracy of processing user query requirements, the present invention divides tasks into three levels according to complexity and tailors processing strategies for each level: 1) Simple-level tasks: These tasks completely rely on the knowledge base and processing power of the large language model itself. At this level, the model operates independently, directly parses the semantics of the query, and quickly generates appropriate responses without the help of any external tools.

[0032] 2) Intermediate-level tasks: When processing such tasks, the model will call an external tool to assist in completing the task. First, the model will deeply analyze the query content and accurately identify the required external tool. Then, the model uses the tool to obtain the necessary supplementary information and combines this information with its own processing capabilities to generate a comprehensive response.

[0033] 3) Complex tasks: These tasks require the model to call external tools in multiple steps, and each call involves collaboration between different tools. The model plays the role of a strategist in this process. It will plan and execute the calls to each tool in an orderly manner to ensure that the output of each step can be effectively passed to the next link. Ultimately, the model will integrate the results of all steps to form a comprehensive and accurate final response.

[0034] Step S105: if yes, determine the target external tool corresponding to the task level by the matching degree between the standard task field and the preset tool, and generate the target answer information of the standard task field based on the target external tool; When a task needs to be solved with the help of external tools, the model needs to accurately identify and call the appropriate tools to quickly and specifically solve user needs.

[0035] Step S106: if not, generate target answer information of the standard task field based on the local knowledge base and internal algorithm to implement the software operation required by the user.

[0036] When the system determines that the task requirements raised by the user are simple, there is no need to call external tools. The processing flow is usually as follows: a) Direct answer: The model immediately generates answers using its internal algorithms and knowledge base without complex queries or external data retrieval. b) Cache queries and outputs: To improve efficiency and optimize future interactive experiences, the system stores the user's standard task fields and the output of the agent model in the history. This caching mechanism helps the model learn and improve its answers, especially when faced with common or repeated questions. c) Serving the next answer: By referring to the history and cached data, the model can respond to similar queries faster while providing more personalized and accurate answers.

[0037] Preferably, the standard task field includes a plurality of subtask description fields.

[0038] Determining the target external tool corresponding to the task level by the matching degree between the standard task field and the preset tool includes steps A1 to A3, wherein: Step A1, converting the subtask description field through the trained semantic representation model to generate a subtask semantic vector; The semantic representation model can be a NLP network framework.

[0039] Step A2, calculating the similarity score between the subtask semantic vector and the semantic vector of the preset tool description, and screening out the preset tools with the similarity score higher than a preset threshold as candidate tools; The selection of candidate tools can also be performed by filtering out the Top-K tool information with the highest similarity scores.

[0040] The similarity is the degree of similarity between the subtask semantics and the text described by the preset tool.

[0041] Step A3, calculating the question-answer matching score between the subtask semantic vector and the semantic vector of the candidate tool description, and selecting the candidate tool with the highest question-answer matching score as the target external tool.

[0042] The question-answer matching degree is the similarity between the semantics of the subtask and the meaning of the candidate usage tool.

[0043] The final selected target external tool is called, and the results returned by the target external tool are integrated into the reasoning process of the model to finally generate a complete answer. The target external tool is used in the predefined tool list. The results returned by the tool will be added to the functional description of the tool to guide the subsequent reasoning of the model.

[0044] Preferably, step S105 may include steps S1051 to S1054, wherein: Step S1051, sorting the target external tools to generate a tool processing sequence; Step S1052, starting from the first tool, processing the standard task field; Step S1053, according to the order corresponding to the tool processing sequence, the output result of the previous tool is used as the input data of the next tool, and this step is repeated until the processing result of the last tool in the tool processing sequence is obtained; Step S1054: determining that the processing result of the last tool in the tool processing sequence is the target answer information of the standard task field.

[0045] Serial collaboration refers to calling multiple tools in sequence, with the output of each tool serving as the input of the next tool. Serial collaboration can be implemented in the infer function: maintain a tool call queue and dynamically adjust the queue based on the model's response. Call the tools in the queue one by one and pass the results to the next tool, thus achieving multi-tool collaboration for complex problems.

[0046] Preferably, step S105 may further include steps S1051' to S1054', wherein: Step S1051', generating a topological structure through the content relevance of the standard task field, wherein the topological structure includes a root node and child nodes, and each child node corresponds to a subtask; Step S1052', determining subtasks that can be executed in parallel in the topological structure; Step S1053', simultaneously calling external tools corresponding to subtasks that can be executed in parallel to calculate the subtasks; Step S1054', summarizing the calculation results of all subtasks of the standard task field to generate target answer information of the standard task field.

[0047] Design a parallel processing framework based on the topological structure of the task: Identify subtasks that can be executed in parallel. Assign subtasks to different tools and execute them in parallel. Finally, integrate the results of all tools to generate the final answer. This greatly improves the efficiency of software operation.

[0048] Preferably, step S102 may include steps S1021 to S1024, wherein: Step S1021, concatenating the task to be answered and the query template field to generate a statement to be processed; Step S1022, using a word segmentation module to perform word segmentation processing on the sentence to be processed to obtain a plurality of feature words; The user's unanswered tasks and pre-defined questions are concatenated and passed to the word segmentation module, which divides them into independent words or symbols, called tokens. For example, "What is the weather like today?" will be divided into ["What", "is", "the", "weather", "like", "today", "?"].

[0049] Step S1023, converting the feature word into a feature word vector, and inputting the feature word vector into an encoder to generate a vector representation containing context information; Each token is converted into a vector representation to obtain a vector sequence, which contains the semantic information of the token, so that the model can understand the relationship between words. The vector sequence is input into the encoder of the model, which converts the input token sequence into a vector representation containing contextual information.

[0050] Step S1024: decode the vector representation containing the context information to generate a standard task field.

[0051] The output vector of the encoder is input into the decoder of the model to generate the final text result. The final output subtask list format, such as [subtask1, subtask2, ...], not only makes the task decomposition process organized, but also facilitates monitoring and evaluation of the processing of each subtask, ensuring that the information obtained by the user is accurate.

[0052] Preferably, the task classification function is:

[0053] Among them, X is the semantic expression degree value of the word segmentation in the standard task field, and Y is the word segmentation statistics in the standard task field. is the balance factor of word segmentation semantics, is the balance factor of the word segmentation statistics. The level score of a simple task is S1, which is less than or equal to the first preset threshold; the level score of a medium-level task is S2, which is greater than the first preset threshold and less than or equal to the second preset threshold; the level score of a complex task is S3, which is greater than the second preset threshold. By effectively dividing the task levels, targeted processing of tasks of different complexity is achieved.

[0054] This embodiment uses the intelligent agent to interact with the real environment, automatically generates a coherent and executable operation plan based on the user's questions, and sends it to the back-end program, which initiates a response without the need to perceive the operation interface, effectively reducing the difficulty of understanding and reasoning large models and making it easy to implement. Through the one-to-one correspondence between task levels and tools, appropriate tools are selected to handle each task, improving the efficiency of handling user task requirements.

[0055] Embodiment 2 Embodiment 2 of the present invention further provides an agent-based software operation device, which corresponds to the agent-based software operation method provided in the above embodiment 1. The corresponding technical features and technical effects are not described in detail in this embodiment, and the relevant parts can refer to the above embodiment 1. Specifically, Figure 2 FIG. 4 shows a structural block diagram of the agent-based software operation device. Figure 2 As shown, the agent-based software operation device 200 includes a configuration module 201, a preprocessing module 202, a task level determination module 203, a tool determination module 204, a first generation module 205 and a second generation module 206, wherein: Configuration module 201, used to configure the initialization parameters of the agent, wherein the initialization parameters include framework model parameters, demand calling tools, dialogue management parameters, dialogue state values, and query template fields; The preprocessing module 202 is connected to the configuration module 201 and is used to respond to the query instruction executed by the user in the configured agent, obtain the task to be answered, preprocess the task to be answered based on the agent, and generate the standard task field; The task level determination module 203 is connected to the preprocessing module 202 and is used to perform complexity calculation on the standard task field through the task classification function to determine the task level of the standard task field, wherein the task level includes simple level, medium level and complex level; The tool determination module 204 is connected to the task level determination module 203 and is used to determine whether the task level of the standard task field requires calling an external tool; The first generation module 205 is connected to the tool determination module 204 and is used to determine the target external tool corresponding to the task level through the matching degree between the standard task field and the preset tool, and generate the target answer information of the standard task field based on the target external tool; The second generating module 206 is connected to the first generating module 205 and is used to generate target answer information of the standard task field based on the local knowledge base and internal algorithm to implement the software operation required by the user.

[0056] Optionally, the first generation module is also used to: sort the target external tools to generate a tool processing sequence; start from the first tool to process the standard task field; use the output result of the previous tool as the input data of the next tool in the order corresponding to the tool processing sequence, and repeat this step until the processing result of the last tool in the tool processing sequence is obtained; determine the processing result of the last tool in the tool processing sequence as the target answer information of the standard task field.

[0057] Optionally, the first generation module is also used to: generate a topological structure through the content correlation of the standard task field, wherein the topological structure includes a root node and child nodes, and each child node corresponds to a subtask; determine the subtasks that can be executed in parallel in the topological structure; simultaneously call external tools corresponding to the subtasks that can be executed in parallel to calculate the subtasks; summarize the calculation results of all subtasks of the standard task field to generate target answer information for the standard task field.

[0058] Optionally, the first generation module is also used to: convert the subtask description field through the trained semantic representation model to generate a subtask semantic vector; calculate the similarity score between the subtask semantic vector and the semantic vector of the preset tool description, and screen out the preset tools with a similarity score higher than a preset threshold as candidate tools; calculate the question-answer matching score between the subtask semantic vector and the semantic vector of the candidate tool description, and screen out the candidate tool with the highest question-answer matching score as the target external tool.

[0059] Optionally, the preprocessing module is also used to: concatenate the task to be answered and the query template field to generate a sentence to be processed; use the word segmentation module to perform word segmentation on the sentence to be processed to obtain multiple feature words; convert the feature words into feature word vectors, and input the feature word vectors into the encoder to generate a vector representation containing context information; decode the vector representation containing context information to generate a standard task field.

[0060] Optionally, the task classification function is:

[0061] Among them, X is the semantic expression degree value of the word segmentation in the standard task field, and Y is the word segmentation statistics in the standard task field. is the balance factor of word segmentation semantics, is the balancing factor of the word segmentation statistics.

[0062] Embodiment 3 Figure 3 The block diagram of a computer device suitable for implementing the agent-based software operation method provided in the third embodiment of the present invention is shown. In this embodiment, the computer device 300 can be a smart phone, tablet computer, laptop computer, desktop computer, rack server, blade server, tower server or cabinet server (including an independent server or a server cluster composed of multiple servers) that executes a program. Figure 3 As shown, the computer device 300 of this embodiment includes at least but is not limited to: a memory 301, a processor 302, and a network interface 303 that can be interconnected through a system bus. It should be noted that Figure 3 Only computer device 300 having components 301 - 303 is shown, but it should be understood that implementing all of the components shown is not a requirement, and more or fewer components may alternatively be implemented.

[0063] In this embodiment, the memory 303 includes at least one type of computer-readable storage medium, and the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, disk, optical disk, etc. In some embodiments, the memory 301 can be an internal storage unit of the computer device 300, such as a hard disk or memory of the computer device 300. In other embodiments, the memory 301 can also be an external storage device of the computer device 300, such as a plug-in hard disk equipped on the computer device 300, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. Of course, the memory 301 can also include both the internal storage unit of the computer device 300 and its external storage device. In this embodiment, the memory 301 is generally used to store the operating system and various application software installed on the computer device 300, such as the program code of the software operation method based on the intelligent agent, etc.

[0064] In some embodiments, the processor 302 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chips. The processor 302 is generally used to control the overall operation of the computer device 300. For example, it performs control and processing related to data interaction or communication with the computer device 300. In this embodiment, the processor 302 is used to run the program code of the steps of the agent-based software operation method stored in the memory 301.

[0065] In this embodiment, the agent-based software operation method stored in the memory 301 may also be divided into one or more program modules and executed by one or more processors (processor 302 in this embodiment) to complete the present invention.

[0066] The network interface 303 may include a wireless network interface or a wired network interface, and the network interface 303 is generally used to establish a communication link between the computer device 300 and other computer devices. For example, the network interface 303 is used to connect the computer device 300 to an external terminal through a network, and to establish a data transmission channel and a communication link between the computer device 300 and the external terminal. The network can be a wireless or wired network such as an intranet, the Internet, the Global System of Mobile communication (GSM), Wideband Code Division Multiple Access (WCDMA), 4G network, 5G network, Bluetooth, Wi-Fi, etc.

[0067] Embodiment 4 This embodiment also provides a computer-readable storage medium, including flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, disk, optical disk, server, App application store, etc., on which a computer program is stored, and when the computer program is executed by a processor, the steps of the software operation method based on the intelligent agent are implemented.

[0068] Obviously, those skilled in the art should understand that the modules or steps of the above-mentioned embodiments of the present invention can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, and optionally, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order from that here, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. In this way, the embodiments of the present invention are not limited to any specific combination of hardware and software.

[0069] It should be noted that the serial numbers of the embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0070] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method.

[0071] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A software operation method based on an agent, characterized in that: The method comprises: Configuring initialization parameters of the agent, wherein the initialization parameters include framework model parameters, demand calling tools, dialogue management parameters, dialogue state values, and query template fields; In response to a query instruction executed by a user in a configured agent, a task to be answered is obtained, and the task to be answered is preprocessed based on the agent to generate a standard task field; Performing complexity calculation on the standard task field through a task classification function to determine the task level of the standard task field, wherein the task level includes a simple level, a medium level and a complex level; Determine whether the task level of the standard task field requires calling an external tool; If yes, determining the target external tool corresponding to the task level by the matching degree between the standard task field and the preset tool, and generating the target answer information of the standard task field based on the target external tool; If not, target answer information for the standard task field is generated based on the local knowledge base and internal algorithms to implement the software operation required by the user.

2. The method according to claim 1, characterized in that The generating target answer information of the standard task field based on the target external tool includes: Sorting the target external tools to generate a tool processing sequence; Using the first tool as a starting point, the standard task fields are processed; According to the order corresponding to the tool processing sequence, the output result of the previous tool is used as the input data of the next tool, and this step is repeated until the processing result of the last tool in the tool processing sequence is obtained; Determine that the processing result of the last tool in the tool processing sequence is the target answer information of the standard task field.

3. The method according to claim 1, wherein generating target answer information of the standard task field based on the target external tool comprises: Generate a topological structure through the content relevance of the standard task field, wherein the topological structure includes a root node and child nodes, and each child node corresponds to a subtask; Determining subtasks that can be executed in parallel in the topology; At the same time, external tools corresponding to subtasks that can be executed in parallel are called to calculate the subtasks; The calculation results of all subtasks of the standard task field are summarized to generate target answer information of the standard task field.

4. The method according to claim 1, characterized in that: The standard task field includes a plurality of subtask description fields, and the target external tool corresponding to the task level is determined by the matching degree between the standard task field and the preset tool, including: The subtask description field is converted by using the trained semantic representation model to generate a subtask semantic vector; Calculating a similarity score between the subtask semantic vector and a semantic vector of a preset tool description, and selecting preset tools with similarity scores higher than a preset threshold as candidate tools; The question-answer matching score between the subtask semantic vector and the semantic vector of the candidate tool description is calculated, and the candidate tool with the highest question-answer matching score is screened out as the target external tool.

5. The method according to claim 1, characterized in that The preprocessing of the task to be answered based on the agent to generate a standard task field includes: The task to be answered is concatenated with the query template field to generate a statement to be processed; Using a word segmentation module to perform word segmentation processing on the sentence to be processed to obtain a plurality of feature words; Convert the feature word into a feature word vector, and input the feature word vector into an encoder to generate a vector representation containing context information; The vector representation containing the context information is decoded to generate a standard task field.

6. The method according to any one of claims 1 to 5, wherein the task classification function is: in, X is the semantic expression value of the word segmentation in the standard task field, and Y is the word segmentation statistics in the standard task field. is the balance factor of word segmentation semantics, is the balancing factor of the word segmentation statistics.

7. A software operation device based on an intelligent agent, characterized in that: The device comprises: A configuration module, used to configure the initialization parameters of the agent, wherein the initialization parameters include framework model parameters, demand calling tools, dialogue management parameters, dialogue state values, and query template fields; A preprocessing module, for obtaining tasks to be answered in response to a query instruction executed by a user in a configured agent, preprocessing the tasks to be answered based on the agent, and generating standard task fields; A task level determination module, used for performing complexity calculation on the standard task field through a task classification function to determine the task level of the standard task field, wherein the task level includes simple level, medium level and complex level; A tool determination module, used to determine whether the task level of the standard task field requires calling an external tool; A first generating module is used to determine the target external tool corresponding to the task level by the matching degree between the standard task field and the preset tool, and generate the target answer information of the standard task field based on the target external tool; The second generating module is used to generate target answer information of the standard task field based on the local knowledge base and internal algorithm to achieve the software operation required by the user.

8. The device according to claim 7, characterized in that The first generating module is also used for: Sorting the target external tools to generate a tool processing sequence; Using the first tool as a starting point, the standard task fields are processed; According to the order corresponding to the tool processing sequence, the output result of the previous tool is used as the input data of the next tool, and this step is repeated until the processing result of the last tool in the tool processing sequence is obtained; Determine that the processing result of the last tool in the tool processing sequence is the target answer information of the standard task field.

9. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method according to any one of claims 1 to 6 when executing the computer program.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images