Cross-platform task intelligent collaboration method and device based on natural language and storage medium

Abstract

The invention provides a cross-platform task intelligent collaboration method and device based on a natural language and a storage medium, and the method comprises the steps: receiving a natural language operation instruction input by a user, analyzing the natural language operation instruction to obtain a plurality of tool operation instructions, and enabling the plurality of tool operation instructions to form a tool call chain, the plurality of tool operation instructions are cross-platform operation instructions; calling Web application APIs, local application programs and / or cloud services corresponding to the multiple cross-platform tool operation instructions at the cloud end based on the sequence of the tool call chain to execute corresponding operations; the method comprises the steps of monitoring execution states of a plurality of cross-platform tool operation instructions in real time, sending out warning information when the execution states are abnormal, intelligently selecting an alternative scheme for execution or switching to man-machine cooperative execution based on the abnormal type, and displaying the execution states in a WEBUI interface in real time. The execution efficiency of the natural language operation instruction is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of the integration of artificial intelligence and automated data processing, and particularly relates to a cross-platform task intelligent collaboration method, device, and storage medium based on natural language. Background Art

[0002] Common methods for automatic human-computer interaction processing in the prior art are as follows: Traditional RPA (Robotic Process Automation) systems, and their defects are: ◦ Dependent on preset process rules (such as UiPath, Automation Anywhere), professional developers are required to write scripts, and cannot respond to users' dynamic requirements in real time; ◦ For cross-system data interaction, manual definition of data mapping rules is required (such as field matching from Excel to the ERP system), and data structure changes cannot be automatically adapted; ◦ Typical case: A certain enterprise needs to manually export data from CRM and manually import it into the BI tool to generate reports every month, taking up to 8 hours per month.

[0003] Conversational AI tools: ◦ Existing products (such as ChatGPT plugins, Microsoft Copilot) only support operations within a single application (such as document generation or email writing), lacking cross-platform task orchestration capabilities; ◦ Users need to manually switch between multiple tools. For example, first generate an Excel chart through a conversation, and then manually copy it into a PPT, resulting in low efficiency and prone to errors.

[0004] The main reason for the above defects is the lack of a standardized mapping mechanism between the semantic results output by the natural language understanding (NLU) module and the tool API calls. For example, a user instruction "Compare last year's and this year's sales data" needs to be converted into multiple action chains such as CRM data query, Excel calculation, and visualization tool rendering. The prior art cannot automatically generate coherent instructions. Summary of the Invention

[0005] In view of one or more of the above technical defects in the prior art, the present invention proposes the following technical solutions.

[0006] A cross-platform task intelligent collaboration method based on natural language, the method comprising: A conversion step of receiving a natural language operation instruction input by a user, parsing the natural language operation instruction to obtain a plurality of tool operation instructions, and forming a tool call chain with the plurality of tool operation instructions, wherein the plurality of tool operation instructions are cross-platform operation instructions; Scheduling step: Based on the order of the tool call chain in the cloud, call the Web application APIs, native applications, and / or cloud services corresponding to the multiple cross-platform tool operation instructions to execute corresponding operations; Monitoring step: Real-time monitor the execution status of multiple cross-platform tool operation instructions, send a warning message when the execution status is abnormal, intelligently select an alternative solution to execute or switch to human-machine collaborative execution based on the type of exception, and display the execution status in real-time on the WEBUI interface.

[0007] Furthermore, the operation of the conversion step is as follows: Use the Bi-LSTM+CRF model to perform intent recognition on the natural language operation instruction to obtain the action subject, object, and constraint conditions of the operation instruction, and perform context-dependent analysis on the natural language operation instruction through the knowledge graph to associate historical tasks, and output the user operation intent in JSON format; Select available tools that can implement the user operation from the tool library based on the user operation intent, and dynamically evaluate the execution efficiency of the tool combination based on the Q-learning algorithm to select the tool combination with the highest execution efficiency ranking or use the trained graph neural network to determine the tool combination; Convert the user operation intent into operation instructions for each tool in the tool combination based on the atomic capability descriptions of each tool stored in the tool metadata database, perform dependency analysis on the operation instructions of each tool based on the action subject, object, and constraint conditions, and form a tool call chain for the operation instructions of each tool based on the results of the dependency analysis.

[0008] Furthermore, the graph construction method of the graph neural network is as follows: Use the function of implementing a tool operation instruction as a node of the graph neural network, and all the historical execution parameters of the software that can implement this tool operation instruction as the feature vector of this node. If there is a dependency relationship between two tool operation instructions, there is an edge between these two tool operation instructions. For example, a tool operation instruction is to generate a chart, and the corresponding software for this tool operation instruction is excel, pdf, and ppt. The historical execution success rate and execution time of the three software form a vector as the feature vector of this node.

[0009] Furthermore, the calculation method of the eigenvalue of each node in the graph of the graph neural network is as follows: + ; Among them, represents the eigenvalue of the i-th node, represents the number of software available for executing the tool operation instruction corresponding to the i-th node, represents The maximum value in the historical execution success rate of a software, denotes the minimum value in the historical execution time of a software, i≥2, ≥1.

[0010] Furthermore, the weight of an edge in the graph of the graph neural network is defined as: if there is a dependency relationship between the tool operation instructions corresponding to two nodes, the weight of the edge between these two nodes is 1, otherwise it is 0.

[0011] Furthermore, the JSON-LD (Linked Data) standard is used to define the data schema for cross-platform data operations, and when the target tool data structure changes, the mapping rule update is automatically triggered to implement the automatic conversion of the data source to the target tool data.

[0012] Furthermore, in the monitoring step, a graph database is used to record the task execution trajectories corresponding to each tool operation instruction, construct a data lineage graph, and persistently store the execution status.

[0013] Furthermore, the user can operate the tool or adjust the tool call chain in the WEBUI interface.

[0014] The present invention also proposes a cross-platform task intelligent collaboration device based on natural language, and the device includes: A conversion unit that receives the natural language operation instructions input by the user, parses the natural language operation instructions to obtain a plurality of tool operation instructions, and forms a tool call chain with the plurality of tool operation instructions, wherein the plurality of tool operation instructions are cross-platform operation instructions; A scheduling unit that sequentially calls the Web application APIs, local application programs, and / or cloud services corresponding to the plurality of cross-platform tool operation instructions in the cloud to execute corresponding operations; A monitoring unit that real-time monitors the execution status of the plurality of cross-platform tool operation instructions, issues a warning message when the execution status is abnormal, and intelligently selects an alternative solution to execute or switches to human-machine collaborative execution based on the type of the abnormality, and displays the execution status in real time on the WEBUI interface.

[0015] Further, the operations of the conversion unit are as follows: Use the Bi-LSTM+CRF model to identify the action subject, object, and constraint conditions of the operation instruction from the natural language operation instruction, and perform context-dependent analysis on the natural language operation instruction through the knowledge graph to associate historical tasks, and output the user operation intention in JSON format; Select available tools that can implement the user operation from the tool library based on the user operation intention, and dynamically evaluate the execution efficiency of the tool combination based on the Q-learning algorithm to select the tool combination with the highest execution efficiency ranking or use the trained graph neural network to determine the tool combination; Convert the user operation intention into the operation instructions of each tool in the tool combination based on the atomic capability descriptions of each tool stored in the tool metadata database, perform dependency analysis on the operation instructions of each tool based on the action subject, object, and constraint conditions, and form a tool call chain from the operation instructions of each tool based on the results of the dependency analysis.

[0016] Further, the graph composition method of the graph neural network is as follows: Use the function of implementing an operation instruction of a tool as a node of the graph neural network, and all the historical execution parameters of the software that can implement the operation instruction of the tool as the feature vector of the node. If there is a dependency relationship between two tool operation instructions, there is an edge between these two tool operation instructions. For example, an operation instruction of a tool is to generate a chart, and the corresponding software for this tool operation instruction is excel, pdf, and ppt. The historical execution success rates and execution times of the three software are used. Therefore, the historical execution success rates and execution times of the three software form a vector as the feature vector of the node.

[0017] Further, the calculation method of the eigenvalue of each node in the graph of the graph neural network is as follows: + ; Among them, represents the eigenvalue of the i-th node, represents the number of software available for executing the tool operation instruction corresponding to the i-th node, represents the maximum value among the historical execution success rates of software, represents the minimum value among the historical execution times of software, i≥2,

[0018] Further, the weight of the edge in the graph of the graph neural network is defined as: If there is a dependency relationship between the tool operation instructions corresponding to two nodes, the weight of the edge between these two nodes is 1, otherwise it is 0.

[0019] Furthermore, the JSON-LD (Linked Data) standard is used to define the data schema for cross-platform data operations. When the data structure of the target tool changes, the mapping rule update is automatically triggered to achieve the automatic conversion of the data source to the target tool data.

[0020] Furthermore, in the monitoring unit, a graph database is used to record the task execution trajectories corresponding to each tool operation instruction, construct a data lineage graph, and persistently store the execution status.

[0021] Furthermore, the user can operate the tool or adjust the tool call chain in the WEBUI interface.

[0022] The present invention also proposes a computer-readable storage medium, on which computer program code is stored. When the computer program code is executed by a computer, the method described above is executed.

[0023] The technical effects of the present invention are as follows: A cross-platform task intelligent collaboration method, device, and storage medium based on natural language according to the present invention. The method includes: a conversion step S101, receiving a natural language operation instruction input by a user, parsing the natural language operation instruction to obtain a plurality of tool operation instructions, and forming a tool call chain with the plurality of tool operation instructions. Among them, the plurality of tool operation instructions are cross-platform operation instructions; a scheduling step S102, based on the order of the tool call chain in the cloud, calling Web application APIs, local application programs, and / or cloud services corresponding to the plurality of cross-platform tool operation instructions to execute corresponding operations; a monitoring step S103, real-time monitoring the execution status of the plurality of cross-platform tool operation instructions, issuing a warning message when the execution status is abnormal, and intelligently selecting an alternative solution to execute or switching to human-machine collaborative execution based on the type of abnormality, and displaying the execution status in real time on the WEBUI interface. The present invention aims to solve the problem that the operation instructions in existing AI products can only be executed in one application, and for executing multiple instruction operations, especially cross-platform operations, manual operations are required. The present invention creatively proposes a cross-platform task intelligent collaboration method. First, the natural language operation instruction input by the user is parsed to obtain a plurality of tool operation instructions, and the plurality of tool operation instructions are formed into a tool call chain. Among them, the plurality of tool operation instructions are cross-platform operation instructions. Then, based on the order of the tool call chain in the cloud, the Web application APIs, local application programs, and / or cloud services corresponding to the plurality of cross-platform tool operation instructions are called to execute corresponding operations, the execution status of the plurality of cross-platform tool operation instructions is real-time monitored, a warning message is issued when the execution status is abnormal, and an alternative solution is intelligently selected to execute or switched to human-machine collaborative execution based on the type of abnormality, and the execution status is displayed in real time on the WEBUI interface. That is, the present invention realizes cross-platform tool calls by constructing a cross-platform tool chain, can real-time monitor the execution status of tool operation instructions, and can also intelligently select an alternative solution to execute or switch to human-machine collaborative execution based on the type of abnormality, thereby improving the execution efficiency of natural language operation instructions. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Other features, objects, and advantages of the present application will become more apparent by reading the detailed description of the non-limiting embodiments with reference to the following drawings.

[0025] Figure 1 is a flowchart of a cross-platform task intelligent collaboration method based on natural language according to an embodiment of the present invention.

[0026] Figure 2 is a structural diagram of a cross-platform task intelligent collaboration device based on natural language according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0027] The present application will be further described in detail below in conjunction with the accompanying drawings and embodiments. It can be understood that the specific embodiments described herein are only used to explain the related invention, rather than limiting the invention. In addition, it should be noted that, for the sake of convenience of description, only the parts related to the relevant invention are shown in the drawings.

[0028] It should be noted that, without conflict, the embodiments in the present application and the features in the embodiments can be combined with each other. The present application will be described in detail below with reference to the drawings and embodiments.

[0029] Figure 1 A cross-platform task intelligent collaboration method based on natural language of the present invention is shown. The method includes: Conversion step S101: Receive a natural language operation instruction input by a user, parse the natural language operation instruction to obtain a plurality of tool operation instructions, and form a tool call chain with the plurality of tool operation instructions. Among them, the plurality of tool operation instructions are cross-platform operation instructions; Scheduling step S102: Based on the order of the tool call chain in the cloud, call the Web application APIs, local application programs, and / or cloud services corresponding to the plurality of cross-platform tool operation instructions to execute corresponding operations; Monitoring step S103: Real-time monitor the execution status of the plurality of cross-platform tool operation instructions, send a warning message when the execution status is abnormal, and intelligently select an alternative solution to execute or switch to human-machine collaborative execution based on the type of abnormality, and display the execution status in real time on the WEBUI interface.

[0030] In order to solve the problem that operation instructions in existing AI products can only be executed in one application, and manual operations are required for executing multiple instruction operations, especially cross-platform operations, the present invention creatively proposes a cross-platform task intelligent collaboration method. First, the natural language operation instructions input by the user are parsed to obtain multiple tool operation instructions, and the multiple tool operation instructions are combined into a tool call chain. Among them, the multiple tool operation instructions are cross-platform operation instructions. Then, based on the order of the tool call chain in the cloud, the Web application APIs, native application programs, and / or cloud services corresponding to the multiple cross-platform tool operation instructions are called to execute corresponding operations, the execution status of the multiple cross-platform tool operation instructions is monitored in real time, and a warning message is sent when the execution status is abnormal, and an alternative solution is intelligently selected based on the type of the exception to execute or switch to human-machine collaborative execution, and the execution status is displayed in real time on the WEBUI interface. That is, by constructing a cross-platform tool chain, the present invention realizes cross-platform tool calls, can monitor the execution status of tool operation instructions in real time, and can also intelligently select alternative solutions based on the type of the exception to execute or switch to human-machine collaborative execution, thereby improving the execution efficiency of natural language operation instructions, which is the important inventive concept of the present invention.

[0031] In the present invention, intelligently selecting an alternative solution to execute based on the type of the exception can be, for example, if the originally selected execution software has an error and an alarm message is reported, the optimal alternative solution can be selected based on the historical success rate. For example, if the originally scheduled execution software for a tool instruction is PowerBI, and an error occurs during the execution of this software, then it can be automatically switched to Excel with a historical success rate of 98% for chart generation. This is an innovation of the present invention, which can ensure that the execution of the user's natural language operation instructions is not interrupted and the correctness of the results is ensured. Of course, it is also possible to switch to the manual operation model after the alarm is issued and execute in a human-machine collaborative manner.

[0032] The real-time display of the execution status on the WEBUI interface can be as follows: ▪ Data has been exported from CRM (30%) ▪ Generating a chart in Excel (60%) ▪ Preparing to send an email (80%) ▪ If a certain step gets stuck (such as Excel crashing), the system will highlight it in red as a warning message.

[0033] The key to the present invention is to determine a highly feasible and reliable tool chain. In one embodiment, the operation of the conversion step S101 is as follows: Use a Bi-LSTM+CRF model to perform intent recognition on the natural language operation instruction to obtain the action subject, object, and constraint conditions of the operation instruction, and perform context-dependent analysis on the natural language operation instruction through a knowledge graph to associate historical tasks, and output the user operation intent in JSON format. For example, for the user's original instruction ("Generate the Q3 sales report and send it by email to the management"), use the Bi-LSTM+CRF model to identify the action subject ("Generate"), the object ("sales report"), and the constraint conditions ("Q3", "management"), and associate historical tasks through the knowledge graph (for example, "sales report" needs to be associated with the CRM data source and the previous month's report template); Select available tools that can implement the user operation from the tool library based on the user operation intent, and dynamically evaluate the execution efficiency of the tool combination based on the Q-learning algorithm to select the tool combination with the highest execution efficiency ranking or use the trained graph neural network to determine the tool combination. The Q-learning algorithm is a commonly used combination evaluation algorithm and a relatively traditional method. Its effect is fast execution, but the effect is slightly poor; Based on the atomic ability descriptions of each tool stored in the tool meta-database, convert the user operation intent into operation instructions for each tool in the tool combination, perform dependency analysis on the operation instructions of each tool based on the action subject, object, and constraint conditions, and form a tool call chain from the operation instructions of each tool based on the results of the dependency analysis.

[0034] To improve the problem of the slightly poor effect of the Q-learning algorithm, the present invention also proposes to use the trained graph neural network to determine the tool combination, so that the determined tool combination is more accurate, and then based on the atomic ability descriptions of each tool stored in the tool meta-database, convert the user operation intent into operation instructions for each tool in the tool combination, perform dependency analysis on the operation instructions of each tool based on the action subject, object, and constraint conditions, and form a tool call chain from the operation instructions of each tool based on the results of the dependency analysis. That is, the finally generated tool call chain is the best tool combination, which improves the efficiency and accuracy of the execution of the natural language operation instruction input by the user. This is the important inventive concept of the present invention.

[0035] In one embodiment, the graph composition method of the graph neural network is as follows: Use the function of implementing a tool operation instruction as a node of the graph neural network, and all the software historical execution parameters that can implement this tool operation instruction as the feature vector of this node. If there is a dependency relationship between two tool operation instructions, then there is an edge between these two tool operation instructions. For example, one tool operation instruction is to generate a chart, and the corresponding software for this tool operation instruction includes Excel, PDF, and PPT. The historical execution success rate and execution time of the three software are used. Therefore, the historical execution success rate and execution time of the three software form a vector as the feature vector of this node.

[0036] In one embodiment, the calculation method of the eigenvalue of each node in the graph of the graph neural network is as follows: + ; Wherein, represents the eigenvalue of the i-th node, represents the number of software available for executing the tool operation instruction corresponding to the i-th node, represents the maximum value among the historical execution success rates of software, represents the minimum value among the historical execution times of software, i ≥ 2,

[0037] In one embodiment, the weight of the edge in the graph of the graph neural network is defined as follows: If there is a dependency relationship between the tool operation instructions corresponding to two nodes, then the weight of the edge between these two nodes is 1, otherwise it is 0.

[0038] In the present invention, a specific graph composition method of the graph neural network is proposed. Use a tool operation instruction as a node, and all the software historical execution parameters that can implement this tool operation instruction as the feature vector of this node. If there is a dependency relationship between two tool operation instructions, then there is an edge between these two tool operation instructions. Calculate the node eigenvalue based on the historical execution success rate and execution time of the software, and a specific calculation formula is proposed. Through simulation calculation, the success rate of the tool combination determined by this method is 8% higher than that of the traditional Q-learning, and the execution time is reduced by 4%. This is another important inventive concept of the present invention.

[0039] In one embodiment, the JSON-LD (Linked Data) standard is used to define a data schema for cross-platform data operations. When the data structure of the target tool changes, the mapping rule update is automatically triggered to achieve the automatic conversion of the data source to the target tool data. Thus, the technical defect in the background art that 'the cross-system data format differences (such as the JSON data of CRM and the table structure of Excel) lead to manual intervention in the intermediate processing and the interruption rate of the automated process is as high as over 40%' is solved.

[0040] In one embodiment, in the monitoring step, a graph database is used to record the task execution trajectories corresponding to each tool operation instruction, construct a data lineage graph, and persistently store the execution status. That is, all operation records of the task (corresponding to a natural language operation instruction input by a user) are saved for easy viewing or collaboration in the future. For example, the system will record the source of each piece of data. For example, which module in the CRM a certain number in a bar chart comes from, just like attaching a 'birth certificate' to the data. After each task is completed, the system will save a'snapshot'. If someone in the team makes a mistake, it can be restored to the previous version with one click. After the task is completed, the system automatically generates a report, which mainly includes: which tools are used, how long it takes, whether there are errors, and where the final result is saved.

[0041] In one embodiment, the user can operate the tool or adjust the tool call chain in the WEBUI interface. For example, if the user thinks the color of the bar chart is not good-looking, they can directly open Excel in the WEBUI to modify the color, and the system will automatically save the modification and continue with the subsequent steps.

[0042] Figure 2 A cross-platform task intelligent collaboration device based on natural language of the present invention is shown, and the device includes: A conversion unit 201, which receives the natural language operation instruction input by the user, parses the natural language operation instruction to obtain a plurality of tool operation instructions, and forms a tool call chain with the plurality of tool operation instructions, wherein the plurality of tool operation instructions are cross-platform operation instructions; A scheduling unit 202, which calls the Web application APIs, local application programs, and / or cloud services corresponding to the plurality of cross-platform tool operation instructions based on the order of the tool call chain in the cloud to execute corresponding operations; A monitoring unit 203, which monitors the execution status of the plurality of cross-platform tool operation instructions in real time, issues a warning message when the execution status is abnormal, and intelligently selects an alternative solution to execute or switches to human-machine collaborative execution based on the type of the abnormality, and displays the execution status in real time on the WEBUI interface.

[0043] In order to solve the problem that the operation instructions in existing AI products can only be executed in one application, and manual operations are required to execute multiple instruction operations, especially cross-platform operations, the present invention creatively proposes a cross-platform task intelligent collaboration method. First, the natural language operation instructions input by the user are parsed to obtain multiple tool operation instructions, and the multiple tool operation instructions are combined into a tool call chain. Among them, the multiple tool operation instructions are cross-platform operation instructions. Then, based on the order of the tool call chain in the cloud, the Web application APIs, local application programs, and / or cloud services corresponding to the multiple cross-platform tool operation instructions are called to execute corresponding operations, and the execution status of the multiple cross-platform tool operation instructions is monitored in real time. When the execution status is abnormal, a warning message is sent, and an alternative solution is intelligently selected based on the type of abnormality to execute or switch to human-machine collaborative execution, and the execution status is displayed in real time on the WEBUI interface. That is, the present invention realizes cross-platform tool calls by constructing a cross-platform tool chain, can monitor the execution status of tool operation instructions in real time, and can also intelligently select alternative solutions based on the type of abnormality to execute or switch to human-machine collaborative execution, thereby improving the execution efficiency of natural language operation instructions. This is the important inventive concept of the present invention.

[0044] In the present invention, intelligently selecting an alternative solution to execute based on the type of abnormality can be, for example, if the originally selected execution software fails and an alarm message is reported, the optimal alternative solution can be selected based on the historical success rate. For example, if the originally scheduled execution software for a tool instruction is Power BI and the software fails during execution, it can be automatically switched to Excel with a historical success rate of 98% for chart generation. This is an innovation of the present invention, which can ensure that the execution of the user's natural language operation instructions is not interrupted and the correctness of the results is ensured. Of course, it can also be switched to the manual operation model after the alarm is issued and executed in a human-machine collaborative manner.

[0045] The real-time display of the execution status on the WEBUI interface can be: ▪ Data has been exported from CRM (30%) ▪ Generating a chart in Excel (60%) ▪ Preparing to send an email (80%) ▪ If a certain step gets stuck (such as Excel crashing), the system will highlight it in red as a warning message.

[0046] The key of the present invention lies in determining a tool chain with high feasibility and high reliability. In one embodiment, the operation of the conversion unit 201 is as follows: using a Bi-LSTM + CRF model to identify the action subject, object, and constraint conditions of the natural language operation instruction, and performing context-dependent analysis on the natural language operation instruction through a knowledge graph to associate historical tasks, and outputting the user operation intention in JSON format. For example, for the user's original instruction ("Generate the Q3 sales report and send it by email to the management"), use the Bi-LSTM + CRF model to identify the action subject ("Generate"), object ("sales report"), and constraint conditions ("Q3", "management"), and associate historical tasks through the knowledge graph (such as "sales report" needs to be associated with the CRM data source and the previous month's report template); select available tools that can implement the user operation from the tool library based on the user operation intention, and dynamically evaluate the execution efficiency of the tool combination based on the Q-learning algorithm to select the tool combination with the highest execution efficiency ranking or determine the tool combination using the trained graph neural network. The Q-learning algorithm is a commonly used combination evaluation algorithm and a relatively traditional method, and its effect is fast execution but slightly poor effect; convert the user operation intention into operation instructions for each tool in the tool combination based on the atomic ability descriptions of each tool stored in the tool metadata database, perform dependency analysis on the operation instructions of each tool based on the action subject, object, and constraint conditions, and form a tool call chain from the operation instructions of each tool based on the results of the dependency analysis.

[0047] To improve the problem of the slightly poor effect of the Q-learning algorithm, the present invention also proposes to determine the tool combination using the trained graph neural network, so that the determined tool combination is more accurate. Then, based on the atomic ability descriptions of each tool stored in the tool metadata database, convert the user operation intention into operation instructions for each tool in the tool combination, perform dependency analysis on the operation instructions of each tool based on the action subject, object, and constraint conditions, and form a tool call chain from the operation instructions of each tool based on the results of the dependency analysis. That is, the finally generated tool call chain is the best tool combination, which improves the execution efficiency and accuracy of the natural language operation instruction input by the user. This is the important inventive concept of the present invention.

[0048] In one embodiment, the graph construction method of the graph neural network is as follows: Use the function of implementing a tool operation instruction as a node of the graph neural network, and all the software historical execution parameters that can implement the tool operation instruction as the feature vector of this node. If there is a dependency relationship between two tool operation instructions, then there is an edge between these two tool operation instructions. For example, a tool operation instruction is to generate a chart, and the corresponding software for this tool operation instruction has Excel, PDF, and PPT. The historical execution success rate and execution time of the three software, therefore, the historical execution success rate and execution time of the three software form a vector as the feature vector of this node.

[0049] In one embodiment, the calculation method of the eigenvalue of each node in the graph of the graph neural network is as follows: + ; Wherein, represents the eigenvalue of the i-th node, represents the number of software available for executing the tool operation instruction corresponding to the i-th node, represents the maximum value among the historical execution success rates of software, represents the minimum value among the historical execution times of software, i ≥ 2,

[0050] In one embodiment, the weight of the edge in the graph of the graph neural network is defined as: If there is a dependency relationship between the tool operation instructions corresponding to two nodes, then the weight of the edge between these two nodes is 1, otherwise it is 0.

[0051] In the present invention, a specific graph construction method of the graph neural network is proposed. Use a tool operation instruction as a node, and all the software historical execution parameters that can implement the tool operation instruction as the feature vector of this node. If there is a dependency relationship between two tool operation instructions, then there is an edge between these two tool operation instructions, and calculate the node eigenvalue based on the historical execution success rate and execution time of the software, and propose a specific calculation formula. Through simulation calculation, the success rate of the tool combination determined by this method is 8% higher than that of the traditional Q-learning, and the execution time is reduced by 4%. This is another important inventive concept of the present invention.

[0052] In one embodiment, the JSON-LD (Linked Data) standard is used to define a data schema for cross-platform data operations. When the data structure of the target tool changes, the mapping rule update is automatically triggered to achieve the automatic conversion of data from the data source to the target tool. Thus, the technical defect in the background art, that is, 'the cross-system data format differences (such as the JSON data of CRM and the table structure of Excel) lead to manual intervention in the intermediate processing, and the interruption rate of the automated process is as high as more than 40%', is solved.

[0053] In one embodiment, in the monitoring step, a graph database is used to record the task execution trajectories corresponding to each tool operation instruction, construct a data lineage graph, and persistently store the execution status. That is, all operation records of the task (corresponding to a natural language operation instruction input by a user) are saved for easy viewing or collaboration later. For example, the system will record the source of each piece of data. For example, which module in the CRM a certain number in a bar chart comes from, just like attaching a 'birth certificate' to the data. After each task is completed, the system saves a'snapshot'. If someone in the team makes a wrong modification, it can be restored to the previous version with one click. After the task ends, the system automatically generates a report, mainly including: which tools are used, how long it takes, whether there are errors, and where the final result is saved.

[0054] In one embodiment, the user can operate the tool or adjust the tool call chain in the WEBUI interface. For example, if the user thinks the color of the bar chart is not good-looking, they can directly open Excel in the WEBUI to modify the color, and the system will automatically save the modification and continue with the subsequent steps.

[0055] Finally, it should be noted that although the embodiments of the present invention are described for cross-platform and cross-software, its method can be fully used for the execution of operation instructions within the same system.

[0056] In one embodiment of the present invention, a computer storage medium is proposed. A computer program is stored on the computer storage medium. When the computer program on the computer storage medium is executed by a processor, the above method is implemented. The computer storage medium can be a hard disk, DVD, CD, flash memory, and other memories.

[0057] For the convenience of description, the above device is described by dividing it into various units according to functions. Of course, when implementing the present application, the functions of each unit can be realized in the same or multiple software and / or hardware. From the description of the above embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus a necessary general hardware platform. Based on such an understanding, the technical solution of this application, in essence, or the part that makes a contribution to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) to execute the devices described in various embodiments or some parts of the embodiments of this application.

[0058] Finally, it should be noted that the above embodiments are only used to illustrate rather than limit the technical solutions of the present invention. Although the present invention has been described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: the present invention can still be modified or equivalently replaced, and any modification or partial replacement without departing from the spirit and scope of the present invention shall be covered by the scope of the claims of the present invention.

Claims

1. A cross-platform task intelligent collaboration method based on natural language, characterized in that, The method includes: A conversion step that receives a natural language operation instruction input by a user, parses the natural language operation instruction to obtain a plurality of tool operation instructions, and forms the plurality of tool operation instructions into a tool call chain, where the plurality of tool operation instructions are cross-platform operation instructions; A scheduling step that, based on the order of the tool call chain in the cloud, calls Web application APIs, native applications, and / or cloud services corresponding to the plurality of cross-platform tool operation instructions to perform corresponding operations; A monitoring step that monitors the execution status of the plurality of cross-platform tool operation instructions in real time, issues a warning message when the execution status is abnormal, intelligently selects an alternative solution to execute or switches to human-machine collaborative execution based on the type of the abnormality, and displays the execution status in real time on a WEBUI interface; Among them, the operation of the conversion step is as follows: Use a Bi-LSTM+CRF model to perform intent recognition on the natural language operation instruction to obtain the action subject, object, and constraint conditions of the operation instruction, and perform context-dependent analysis on the natural language operation instruction through a knowledge graph to associate historical tasks, and output the user operation intent in JSON format; Select available tools that can implement the user operation from a tool library based on the user operation intent, and dynamically evaluate the execution efficiency of tool combinations based on the Q-learning algorithm to select the tool combination with the highest execution efficiency ranking or use a trained graph neural network to determine the tool combination; Based on the atomic ability descriptions of each tool stored in a tool metadata database, convert the user operation intent into operation instructions for each tool in the tool combination, perform dependency analysis on the operation instructions of each tool based on the action subject, object, and constraint conditions, and form the operation instructions of each tool into a tool call chain based on the results of the dependency analysis.

2. The method according to claim 1, wherein Use the JSON-LD (Linked Data) standard to define a data schema for cross-platform data operations, and when the data structure of the target tool changes, automatically trigger the update of the mapping rule to achieve the automatic conversion of the data source to the target tool data.

3. The method according to claim 2, characterized in that, In the monitoring step, use a graph database to record the task execution trajectories corresponding to each tool operation instruction, construct a data lineage graph, and persistently store the execution status.

4. The method according to claim 3, wherein The user can operate the tools or adjust the tool call chain in the WEBUI interface.

5. An intelligent cross-platform task collaboration device based on natural language, characterized in that, The device includes: A conversion unit that receives a natural language operation instruction input by a user, parses the natural language operation instruction to obtain a plurality of tool operation instructions, and forms the plurality of tool operation instructions into a tool call chain, where the plurality of tool operation instructions are cross-platform operation instructions; A scheduling unit that, based on the order of the tool call chain in the cloud, calls Web application APIs, native applications, and / or cloud services corresponding to the plurality of cross-platform tool operation instructions to perform corresponding operations; The monitoring unit monitors the execution status of multiple cross-platform tool operation instructions in real time, issues a warning message when the execution status is abnormal, intelligently selects an alternative solution for execution or switches to human-machine collaborative execution based on the type of exception, and displays the execution status in real time on the WEBUI interface. Among them, the operations of the conversion unit are as follows: Use the Bi-LSTM+CRF model to identify the action subject, object, and constraints of the operation instruction from the natural language operation instruction, and perform context-dependent analysis on the natural language operation instruction through the knowledge graph to associate historical tasks, and output the user operation intention in JSON format; Select available tools that can implement the user operation from the tool library based on the user operation intention, and dynamically evaluate the execution efficiency of the tool combination based on the Q-learning algorithm to select the tool combination with the highest execution efficiency ranking or use the trained graph neural network to determine the tool combination; Convert the user operation intention into the operation instructions of each tool in the tool combination based on the atomic ability descriptions of each tool stored in the tool metadata database, perform dependency analysis on the operation instructions of each tool based on the action subject, object, and constraints, and form a tool call chain for the operation instructions of each tool based on the results of the dependency analysis.

6. The device according to claim 5, characterized in that, Use the JSON-LD (Linked Data) standard to define the data schema for cross-platform data operations, and when the data structure of the target tool changes, automatically trigger the update of the mapping rules to achieve the automatic conversion of the data source to the target tool data.

7. The device according to claim 6, characterized in that, Use a graph database in the monitoring unit to record the task execution trajectories corresponding to each tool operation instruction, construct a data lineage graph, and persistently store the execution status.

8. A computer storage medium, characterized in that, A computer program is stored on the computer storage medium, and when the computer program on the computer storage medium is executed by a processor, the method according to any one of claims 1-4 is implemented.

Images