RPA process aided design method, device and equipment for large model collaborative knowledge graph

Abstract

The embodiment of the invention discloses an RPA process aided design method, device and equipment for a large model collaborative knowledge graph. The method comprises the steps of obtaining a completed RPA code of a user, and extracting operation information and parameter information in the completed RPA code; performing user intention recognition according to the operation information and the parameter information to obtain at least one intention operation category and an intention parameter corresponding to each intention operation category; for any intention operation category, determining the similarity between the intention operation category and each graph control in a preset knowledge graph according to the intention operation category and the corresponding intention parameter; and recommending the graph control corresponding to the similarity meeting a preset similarity condition so as to assist a user in designing an RPA process. On the basis, the intent recognition mechanism and the knowledge graph recommendation mechanism are combined, the needed graph control is recommended to the user, the user can directly select the graph control for subsequent design, and therefore the development and design efficiency of the RPA process is improved, and the development difficulty is reduced.

Description

Technical Field

[0001] The embodiments of the present application relate to the technical field of RPA process design, and in particular, to an RPA process assisted design method, device, and equipment for collaborative knowledge graph of large models. Background Art

[0002] With the development of technology, for some repetitive business process tasks, the software technology of robotic process automation (RPA) is used to automatically complete by simulating the interaction between human users and application programs (such as clicking, inputting, copying and pasting, etc.). It significantly improves work efficiency and reduces the time and error rate of manual operations.

[0003] Currently, the development and design of RPA processes usually rely on manually writing code to achieve. Manually writing RPA code requires high professionalism, has low efficiency and is prone to problems. Summary of the Invention

[0004] The embodiments of the present application provide an RPA process assisted design method, device, and equipment for collaborative knowledge graph of large models to improve the development and design efficiency of RPA processes, reduce the development difficulty, and improve the development accuracy.

[0005] In a first aspect, the embodiments of the present application provide an RPA process assisted design method for collaborative knowledge graph of large models, and the method includes:

[0006] Obtain the completed RPA code of the user, and extract the operation information and parameter information in the completed RPA code;

[0007] Perform user intention recognition based on the operation information and parameter information to obtain at least one intention operation category and the respective intention parameters corresponding to each intention operation category;

[0008] For any intention operation category, determine the similarity between the intention operation category and each graph control in the preset knowledge graph according to the intention operation category and the corresponding intention parameters; the graph control is a code control pre-stored in the preset knowledge graph for realizing the RPA goal;

[0009] Recommend the graph controls corresponding to the similarities that meet the preset similarity conditions to assist the user in designing the RPA process.

[0010] In a second aspect, the embodiments of the present application provide an RPA process assisted design device for collaborative knowledge graph of large models, including:

[0011] An extraction module, configured to obtain the completed RPA code of the user and extract the operation information and parameter information in the completed RPA code;

[0012] An intention recognition module, configured to perform user intention recognition according to operation information and parameter information, so as to obtain at least one intention operation category and corresponding intention parameters for each intention operation category;

[0013] A similarity comparison module, configured to, for any intention operation category, determine the similarity between the intention operation category and each graph control in a preset knowledge graph according to the intention operation category and the corresponding intention parameters; the graph control is a code control stored in advance in the preset knowledge graph for realizing the RPA target;

[0014] A recommendation assistance module, configured to recommend the graph controls corresponding to the similarities that meet the preset similarity conditions to assist the user in designing the RPA process.

[0015] In a third aspect, an embodiment of the present application further provides an RPA process assisted design device for a large model collaborative knowledge graph. The RPA process assisted design device for a large model collaborative knowledge graph includes:

[0016] One or more processors;

[0017] A storage device, configured to store one or more programs,

[0018] When the one or more programs are executed by the one or more processors, the one or more processors implement the RPA process assisted design method for a large model collaborative knowledge graph provided in any embodiment of the present application.

[0019] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium, on which a computer program is stored, and characterized in that when the program is executed by a processor, it implements the RPA process assisted design method for a large model collaborative knowledge graph provided in any embodiment of the present application.

[0020] In the technical solution of the embodiment of the present application, the completed RPA code of the user is obtained, and the operation information and parameter information in the completed RPA code are extracted; according to the operation information and parameter information, user intention recognition is performed to obtain at least one intended operation category and the intended parameters corresponding to each intended operation category; for any intended operation category, the similarity between the intended operation category and each graph control in the preset knowledge graph is determined according to the intended operation category and the corresponding intended parameters; the graph control is a code control pre-stored in the preset knowledge graph for realizing the RPA target; the graph controls corresponding to the similarities that meet the preset similarity conditions are recommended to assist the user in designing the RPA process. Based on this, the present application combines the intention recognition mechanism and the knowledge graph recommendation mechanism, and during the process of the user designing the RPA process, recommends the required graph controls to the user. The user can directly select the recommended graph controls for the subsequent process design, thereby improving the development and design efficiency of the RPA process, reducing the development difficulty, and improving the development accuracy. Description of the Drawings

[0021] Figure 1 It is a schematic flowchart of the RPA process assisted design method of the large model collaborative knowledge graph provided by Embodiment 1 of the present application;

[0022] Figure 2 It is a schematic structural diagram of an RPA process assisted design device of a large model collaborative knowledge graph provided by Embodiment 2 of the present application;

[0023] Figure 3 It is a schematic structural diagram of an RPA process assisted design device of a large model collaborative knowledge graph provided by Embodiment 3 of the present application. Detailed Embodiment

[0024] The present application will be further described in detail below with reference to the drawings and embodiments. It can be understood that the specific embodiments described herein are only used to explain the present application, rather than limiting the present application. In addition, it should be noted that for the sake of description, only parts related to the present application are shown in the drawings rather than all the structures.

[0025] Embodiment 1

[0026] Figure 1 It is a schematic flowchart of the RPA process assisted design method of the large model collaborative knowledge graph provided by Embodiment 1 of the present application. As Figure 1 shown, the RPA process assisted design method of the large model collaborative knowledge graph provided in this embodiment can be implemented based on an RPA process assisted design device of the large model collaborative knowledge graph equipped with a video acquisition end, and specifically may include the following steps:

[0027] Step 101: Obtain the completed RPA code of the user, and extract the operation information and parameter information from the completed RPA code.

[0028] In this step, the completed RPA code of the user refers to the robot process automation code that the user has currently written. To further reduce the professionalism required in design, at the initial stage of design, the knowledge graph can recommend some graph controls corresponding to the starting operations for the user to directly select and use.

[0029] Among them, the starting operation can be determined according to the user's historical design. Since different users are in different industries, the starting operations of the RPA process usually vary. Moreover, in the same industry, the starting operations are usually similar, such as reading table data, creating new table items, etc. Therefore, in this application, the user's historical RPA design can be obtained, the number of various starting operations can be counted, and the graph control corresponding to the starting operation with the largest number can be recommended to the user for selection.

[0030] In this step, the pre-trained LLM model can be used to perform a structured representation of the completed RPA code; the information corresponding to the target structure is determined as the operation information or parameter information of the completed RPA code.

[0031] Among them, the LLM model is trained to be able to understand the code semantics and identify the key operations and parameters in the completed RPA code of the user. After the identification is completed, the identified content can be converted into a structured representation form. This process can refer to the use of abstract syntax tree parsing technology, which will not be elaborated here.

[0032] After being converted into a structured representation form, it will contain multiple structures, and each structure will correspond to a piece of code information, such as operation information like operation type and operation object, or parameter information like parameter value.

[0033] Therefore, the corresponding information can be obtained from specific structures, so as to achieve the extraction of operation information or parameter information.

[0034] In a specific example, the user is using an RPA tool to automate the processing of Excel table data. The user has written code to read Excel data and now hopes to perform the next operation, such as writing the data to a database.

[0035] Among them, the code snippet completed by the user is as follows:

[0036] import pandas as pd;

[0037] data = pd.read_excel("data.xlsx");

[0038] After structuring, the following structure is obtained:

[0039] Operation type: Read Excel;

[0040] Operation object: data.xlsx;

[0041] Data type: DataFrame.

[0042] Step 102: Identify the user's intention based on the operation information and parameter information, obtaining at least one intention operation category and the corresponding intention parameters for each intention operation category.

[0043] Specifically, in this step, the user's intention can be identified first based on the operation information and parameter information to obtain at least one intention operation category; then, the intention parameters can be filled according to the intention operation category, operation information, and parameter information to obtain the corresponding intention parameters for each intention operation category.

[0044] Among them, the user intention recognition can also be achieved through a neural network model (such as an LLM model). Specifically, a pre-trained semantic understanding model can be used to analyze and infer the operation information and parameter information to obtain the user's potential intention keywords; the user's potential intention keywords are classified into pre-defined operation categories to obtain at least one intention operation category.

[0045] It should be noted that the user's historical RPA designs can be used as training data to train the LLM model, so that it can better fit the user's design scenarios and design habits, and make the inferred user intentions closer to the user's true intentions.

[0046] The results obtained by the LLM model for identifying the user's intention are at least one intention operation category. Multiple different operation categories can be preset to make the output of the LLM model more standardized.

[0047] In a specific example, the following operation categories can be included:

[0048] Data processing: including data reading, data writing, data conversion, data cleaning, etc.;

[0049] Process control: including loops, conditional judgments, exception handling, etc.;

[0050] User interaction: including sending emails, pop-up prompts, obtaining user input, etc.;

[0051] System operations: including opening applications, operating files, executing commands, etc.;

[0052] Web operations: including web scraping, web automation, etc.

[0053] The potential intention keywords of the user obtained based on the above semantic understanding model can be classified according to the semantic similarity with the above operation categories and their related explanations, so as to find at least one operation category that is closest to the potential intention keywords of the user, which is the intention operation category.

[0054] After obtaining any of the above intention operation categories, the intention parameters can be filled. Generally, the parameters required for different operation categories are different. Therefore, a set of parameter slots can be set for each operation category, and then according to the operation information and parameter information extracted above, these parameter slots can be filled to obtain the intention parameters for each intention operation category.

[0055] For example, the parameter slots of an operation category include:

[0056] Data source slot: such as Excel file, CSV file, database, web page, etc.;

[0057] Data target slot: such as Excel file, CSV file, database, variable, etc.;

[0058] Operation type slot: such as read, write, update, delete, etc.;

[0059] Condition slot: such as judgment condition, loop times, etc.;

[0060] Target application slot: such as browser, specific application program, etc.

[0061] Still taking the example in the above step 101 as an example, the potential intention keywords of the user obtained by the semantic understanding model can be data processing and data writing. Through classification, it can be known that the intention operation category is data writing in data processing, and the corresponding parameter slots can be data source, data type, and operation type. After filling these parameter slots based on the parameter information and operation information extracted above, the following can be obtained:

[0062] Data source: data.xlsx (Excel file);

[0063] Data type: DataFrame;

[0064] Operation type: write.

[0065] Step 103: For any intention operation category, determine the similarity between the intention operation category and each graph control in the preset knowledge graph according to the intention operation category and the corresponding intention parameters.

[0066] In this step, the graph control is a code control stored in advance in the preset knowledge graph for realizing the RPA target. In this embodiment, the knowledge graph will be maintained and expanded, and the knowledge graph will involve entities, relationships, and attributes.

[0067] In a specific example, the knowledge graph entities in this embodiment may include the following:

[0068] Data source entities: Excel, CSV, databases (MySQL, SQLServer, Oracle), websites, PDFs, emails, APIs, etc.;

[0069] RPA control entities: Excel reading control, Excel writing control, email sending control, web page opening control, button clicking control, text input control, data scraping control, OCR control, loop control, conditional judgment control, waiting control, etc.;

[0070] RPA operation entities: data reading, data writing, email sending, web page opening, button clicking, text input, data scraping, OCR recognition, loop operation, conditional judgment, waiting operation, etc.;

[0071] Data type entities: strings, numbers, dates, booleans, etc.;

[0072] Application scenario entities: financial automation, human resources automation, IT automation, customer service automation, etc.

[0073] In addition, the knowledge graph relationships can, but are not limited to, include the following relationships:

[0074] Operation object: RPA control -> data source (e.g., Excel reading control -> Excel file);

[0075] Input / output: RPA control -> data type (e.g., Excel reading control -> string);

[0076] Dependency relationship: RPA control -> RPA control (e.g., loop control -> Excel reading control);

[0077] Function implementation: RPA control -> RPA operation (e.g., Excel reading control -> data reading);

[0078] Application scenario: RPA control -> application scenario (e.g., Excel reading control -> financial automation);

[0079] Similar functions: RPA control -> RPA control (e.g., CSV reading control -> Excel reading control).

[0080] It should be noted that for the attributes of entities and relationships, reference can be made to the above-explained and exemplified content of entities and relationships.

[0081] In this step, when determining the similarity, for any intention operation category, generate a user intention vector from the intention operation category and the corresponding intention parameters; for any user intention vector, calculate the similarity between the user intention vector and each graph control vector in the preset knowledge graph to obtain the similarity between the intention operation category and each graph control.

[0082] Among them, when generating the user intention vector, the intention operation category vector and the intention parameter vector can be concatenated or weighted averaged to obtain the final user intention vector. Among them, the intention operation category vector is obtained based on the intention operation category, and the intention parameter vector is obtained based on the intention parameters, and both can be obtained using methods such as one-hot encoding or word embedding.

[0083] In the above example, assuming that a word embedding model (such as Word2Vec) is used to convert the intention, slot value, and knowledge graph entity into vectors, the following can be obtained:

[0084] Intention operation type vector: The vector representation of "data writing" is [0.2, 0.5, 0.8, -0.1, 0.3];

[0085] Data source vector: The vector representation of "Excel file" is [0.7, 0.1, -0.2, 0.6, 0.4];

[0086] Data type vector: The vector representation of "DataFrame" is [0.9, -0.3, 0.2, 0.1, 0.5];

[0087] Operation type vector: The vector representation of "write" is [-0.1, 0.7, 0.6, -0.2, 0.8].

[0088] After concatenating the above vectors, the following is obtained:

[0089] User intention vector = [0.2, 0.5, 0.8, -0.1, 0.3, 0.7, 0.1, -0.2, 0.6, 0.4, 0.9, -0.3, 0.2, 0.1, 0.5, -0.1, 0.7, 0.6, -0.2, 0.8]

[0090] In the knowledge graph, each graph space is also converted into a vector. For example, a graph embedding algorithm (such as Node2Vec, TransE) is used to convert the entities and relationships in the knowledge graph into vectors. In a specific example, the vector representation of the control for writing to the database in the knowledge graph is: Write to database vector = [0.1, 0.6, 0.9, -0.2, 0.2, 0.8, 0.2, -0.1, 0.5, 0.3, 0.8, -0.2, 0.3, 0.2, 0.6, -0.3, 0.8, 0.7, -0.1, 0.9].

[0091] In this embodiment, the cosine similarity calculation method can be used to calculate the similarity between the intended operation category and each graph control in the preset knowledge graph. The similarity between the above-mentioned user intention vector and the vector written into the database is cos(user intention vector, vector written into the database). In a specific example, the obtained similarity is 0.85.

[0092] Step 104: Recommend the graph controls corresponding to the similarities that meet the preset similarity conditions to assist the user in designing the RPA process.

[0093] Through the foregoing calculation, the similarities between each graph control in the knowledge graph and the intended operation category have been calculated. Therefore, in this step, for any intended operation category, the graph control with the highest similarity to the intended operation category is recommended.

[0094] In a specific example, for the intended operation category of data writing, if 0.85 is the highest similarity, then the corresponding control of writing into the database can be recommended to the user.

[0095] In order to further improve the user's design efficiency, in this embodiment, when it is detected that the user selects any recommended graph control, the graph control is pre-filled with intention parameters; the pre-filled graph control is edited and displayed, and the pre-filled content is marked in a preset manner to facilitate the user to re-edit the pre-filled graph control.

[0096] It should be noted that the intention parameters can be directly filled into the corresponding parameter positions of the graph control to obtain a graph control containing parameter pairs. In order to facilitate the user to re-edit, the filled part can be marked, such as highlighting, underlining, etc.

[0097] In this embodiment, the completed RPA code of the user is obtained, and the operation information and parameter information in the completed RPA code are extracted; user intention recognition is performed according to the operation information and parameter information to obtain at least one intended operation category and the intention parameters corresponding to each intended operation category; for any intended operation category, the similarity between the intended operation category and each graph control in the preset knowledge graph is determined according to the intended operation category and the corresponding intention parameters; the graph control is a code control stored in advance in the preset knowledge graph for realizing the RPA target; the graph controls corresponding to the similarities that meet the preset similarity conditions are recommended to assist the user in designing the RPA process. Based on this, the present application combines the intention recognition mechanism and the knowledge graph recommendation mechanism, and recommends the required graph controls to the user during the process of the user designing the RPA process. The user can directly select the recommended graph controls to design the subsequent process, thereby improving the development and design efficiency of the RPA process, reducing the development difficulty, and improving the development accuracy.

[0098] Example Two

[0099] Figure 2 The following is a schematic structural diagram of an RPA process assisted design device for collaborative knowledge graphs of large models provided in Example Two of this application. The RPA process assisted design device for collaborative knowledge graphs of large models provided in the embodiments of this application can execute the RPA process assisted design method for collaborative knowledge graphs of large models provided in any embodiment of this application, and has the corresponding functional modules and beneficial effects for executing the method. This device can be implemented in software and / or hardware, such as Figure 2 shown, the device includes:

[0100] An extraction module 201, configured to obtain the completed RPA code of the user and extract the operation information and parameter information in the completed RPA code;

[0101] An intention recognition module 202, configured to perform user intention recognition based on the operation information and parameter information to obtain at least one intention operation category and the intention parameters corresponding to each intention operation category;

[0102] A similarity comparison module 203, configured to, for any intention operation category, determine the similarity between the intention operation category and each graph control in the preset knowledge graph according to the intention operation category and the corresponding intention parameters; the graph control is a code control pre-stored in the preset knowledge graph for achieving the RPA target;

[0103] A recommendation assistance module 204, configured to recommend the graph controls corresponding to the similarities that meet the preset similarity conditions to assist the user in designing the RPA process.

[0104] Furthermore, the extraction module is specifically configured to:

[0105] Perform a structured representation of the completed RPA code using a pre-trained LLM model;

[0106] Determine the information corresponding to the target structure as the operation information or parameter information of the completed RPA code.

[0107] Furthermore, the intention recognition module is specifically configured to:

[0108] Perform user intention recognition based on the operation information and parameter information to obtain at least one intention operation category;

[0109] Perform intention parameter filling according to the intention operation category, operation information, and parameter information to obtain the intention parameters corresponding to each intention operation category.

[0110] Furthermore, the intention recognition module is specifically configured to:

[0111] Analyze and infer the operation information and parameter information using a pre-trained semantic understanding model to obtain potential user intention keywords;

[0112] Classify the potential user intention keywords into predefined operation categories to obtain at least one intended operation category.

[0113] Furthermore, the similarity comparison module is specifically used for:

[0114] For any intended operation category, generate a user intention vector from the intended operation category and the corresponding intended parameters;

[0115] For any user intention vector, calculate the similarity between the user intention vector and each graph control vector in the preset knowledge graph to obtain the similarity between the intended operation category and each graph control.

[0116] Furthermore, the recommendation assistance module is specifically used for:

[0117] For any intended operation category, recommend the graph control with the highest similarity to the intended operation category.

[0118] Furthermore, the device is specifically further used for:

[0119] In the case of detecting that the user selects any recommended graph control, pre-fill the graph control with the intended parameters;

[0120] Edit and display the pre-filled graph control, and label the pre-filled content in a preset manner to facilitate the user to re-edit the pre-filled graph control.

[0121] Embodiment III

[0122] Figure 3 For the RPA process assisted design device of a large model collaborative knowledge graph provided in Embodiment IV of this application, as Figure 3 shown, the RPA process assisted design device of the large model collaborative knowledge graph further includes a processor 310, a memory 320, an input device 330, and an output device 340; the number of processors 310 in the RPA process assisted design device of the large model collaborative knowledge graph can be one or more, Figure 3 taking one processor 310 as an example; the processor 310, the memory 320, the input device 330, and the output device 340 in the RPA process assisted design device of the large model collaborative knowledge graph can be connected through a bus or other means, Figure 3 taking connection through a bus as an example.

[0123] The memory 320, as a computer-readable storage medium, can be used to store software programs, computer-executable programs, and modules, such as the program instructions / modules corresponding to the RPA process assisted design method of the large model collaborative knowledge graph in the embodiments of the present invention. The processor 310 executes various functional applications and data processing of the RPA process assisted design device of the large model collaborative knowledge graph by running the software programs, instructions, and modules stored in the memory 320, that is, implements the above-mentioned RPA process assisted design method of the large model collaborative knowledge graph:

[0124] Obtain the completed RPA code of the user, and extract the operation information and parameter information in the completed RPA code;

[0125] Perform user intention recognition according to the operation information and parameter information to obtain at least one intention operation category and the respective intention parameters corresponding to each intention operation category;

[0126] For any intention operation category, determine the similarity between the intention operation category and each graph control in the preset knowledge graph according to the intention operation category and the corresponding intention parameters; the graph control is a code control stored in advance in the preset knowledge graph for implementing the RPA target;

[0127] Recommend the graph controls corresponding to the similarities that meet the preset similarity conditions to assist the user in designing the RPA process.

[0128] Further, extracting the operation information and parameter information in the completed RPA code includes:

[0129] Use the pre-trained LLM model to perform a structured representation of the completed RPA code;

[0130] Determine the information corresponding to the target structure as the operation information or parameter information of the completed RPA code.

[0131] Further, performing user intention recognition according to the operation information and parameter information to obtain at least one intention operation category and the respective intention parameters corresponding to each intention operation category includes:

[0132] Perform user intention recognition according to the operation information and parameter information to obtain at least one intention operation category;

[0133] Fill in the intention parameters according to the intention operation category, operation information, and parameter information to obtain the respective intention parameters corresponding to each intention operation category.

[0134] Further, performing user intention recognition according to the operation information and parameter information to obtain at least one intention operation category includes:

[0135] Analyze and infer the operation information and parameter information using a pre-trained semantic understanding model to obtain potential user intention keywords;

[0136] Classify the potential user intention keywords into pre-defined operation categories to obtain at least one intention operation category.

[0137] Further, for any intention operation category, determine the similarity between the intention operation category and each graph control in the pre-set knowledge graph according to the intention operation category and the corresponding intention parameters, including:

[0138] For any intention operation category, generate a user intention vector from the intention operation category and the corresponding intention parameters;

[0139] For any user intention vector, calculate the similarity between the user intention vector and each graph control vector in the pre-set knowledge graph to obtain the similarity between the intention operation category and each graph control.

[0140] Further, recommend the graph controls corresponding to the similarities that meet the pre-set similarity conditions, including:

[0141] For any intention operation category, recommend the graph control with the highest similarity to the intention operation category.

[0142] Further, the method further includes:

[0143] When it is monitored that the user selects any recommended graph control, pre-fill the graph control with the intention parameters;

[0144] Edit and display the pre-filled graph control, and label the pre-filled content in a pre-set manner to facilitate the user to re-edit the pre-filled graph control.

[0145] The memory 320 may mainly include a program storage area and a data storage area. Among them, the program storage area may store an operating system and application programs required for at least one function; the data storage area may store data created according to the use of the terminal, etc. In addition, the memory 320 may include high-speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, a flash memory device, or other non-volatile solid-state storage devices. In some instances, the memory 320 may further include a memory remotely set relative to the processor 310, and these remote memories may be connected to the RPA process-assisted design device of the large model collaborative knowledge graph through a network. Examples of the above network include but are not limited to the Internet, an enterprise intranet, a local area network, a mobile communication network, and combinations thereof.

[0146] Embodiment Four

[0147] Embodiment 4 of this application further provides a storage medium containing computer-executable instructions. When the computer-executable instructions are executed by a computer processor, they are used to execute an RPA process assisted design method for a large model collaborative knowledge graph. The method includes:

[0148] Obtain the completed RPA code of the user, and extract the operation information and parameter information in the completed RPA code;

[0149] Perform user intention recognition based on the operation information and parameter information to obtain at least one intention operation category and the intention parameters corresponding to each intention operation category;

[0150] For any intention operation category, determine the similarity between the intention operation category and each graph control in the preset knowledge graph according to the intention operation category and the corresponding intention parameters; the graph control is a code control stored in advance in the preset knowledge graph for realizing the RPA target;

[0151] Recommend the graph controls corresponding to the similarities that meet the preset similarity conditions to assist the user in designing the RPA process.

[0152] Further, extracting the operation information and parameter information in the completed RPA code includes:

[0153] Use the pre-trained LLM model to perform a structured representation of the completed RPA code;

[0154] Determine the information corresponding to the target structure as the operation information or parameter information of the completed RPA code.

[0155] Further, performing user intention recognition based on the operation information and parameter information to obtain at least one intention operation category and the intention parameters corresponding to each intention operation category includes:

[0156] Perform user intention recognition based on the operation information and parameter information to obtain at least one intention operation category;

[0157] Fill in the intention parameters according to the intention operation category, operation information and parameter information to obtain the intention parameters corresponding to each intention operation category.

[0158] Further, performing user intention recognition based on the operation information and parameter information to obtain at least one intention operation category includes:

[0159] Use the pre-trained semantic understanding model to analyze and infer the operation information and parameter information to obtain the user's potential intention keywords;

[0160] Classify the user's potential intention keywords into the predefined operation categories to obtain at least one intention operation category.

[0161] Further, for any intended operation category, determining the similarity between the intended operation category and each graph control in the preset knowledge graph according to the intended operation category and the corresponding intended parameters includes:

[0162] For any intended operation category, generating a user intention vector from the intended operation category and the corresponding intended parameters;

[0163] For any user intention vector, calculating the similarity between the user intention vector and each graph control vector in the preset knowledge graph to obtain the similarity between the intended operation category and each graph control.

[0164] Further, recommending the graph controls corresponding to the similarities that meet the preset similarity conditions includes:

[0165] For any intended operation category, recommending the graph control with the highest similarity to the intended operation category.

[0166] Further, the method further includes:

[0167] When it is detected that the user selects any recommended graph control, pre-filling the graph control with the intended parameters;

[0168] Presenting the pre-filled graph control for editing and annotating the pre-filled content in a preset manner to facilitate the user to re-edit the pre-filled graph control.

[0169] Of course, for a storage medium containing computer-executable instructions provided by an embodiment of the present application, the computer-executable instructions are not limited to the above method operations, and can also execute relevant operations in the RPA process-assisted design method of the large model collaborative knowledge graph provided by any embodiment of the present application.

[0170] From the above description of the embodiments, those skilled in the art can clearly understand that the present application can be implemented by means of software and necessary general-purpose hardware. Of course, it can also be implemented by hardware, but in many cases, the former is a better implementation method. Based on such an understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as a floppy disk, read-only memory (ROM), random access memory (RAM), flash memory (FLASH), hard disk, or optical disc of a computer, and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods of various embodiments of the present application.

[0171] It should be noted that in the embodiments of the above search device, the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other and are not used to limit the protection scope of the present application.

[0172] Note that the above is only a preferred embodiment of the present application and the technical principles applied. Those skilled in the art will understand that the present application is not limited to the specific embodiments here, and various obvious changes, re-adjustments and substitutions can be made by those skilled in the art without departing from the protection scope of the present application. Therefore, although the present application has been described in more detail through the above embodiments, the present application is not limited to the above embodiments only. Without departing from the concept of the present application, more other equivalent embodiments can be included, and the scope of the present application is determined by the scope of the appended claims.

Claims

1. An RPA process assisted design method for large model collaborative knowledge graphs, characterized in that, The method includes: Obtaining the completed RPA code of the user, and extracting the operation information and parameter information in the completed RPA code; Performing user intent recognition based on the operation information and parameter information to obtain at least one intent operation category and the respective intent parameters corresponding to each intent operation category; For any intent operation category, determining the similarity between the intent operation category and each graph control in the preset knowledge graph according to the intent operation category and the corresponding intent parameters; the graph control is a code control stored in advance in the preset knowledge graph for realizing the RPA target; Recommending the graph controls corresponding to the similarities that meet the preset similarity conditions to assist the user in designing the RPA process.

2. The method according to claim 1, characterized in that, The extracting the operation information and parameter information in the completed RPA code includes: Using a pre-trained LLM model to perform a structured representation of the completed RPA code; Determining the information corresponding to the target structure as the operation information or parameter information of the completed RPA code.

3. The method according to claim 1, wherein The performing user intent recognition based on the operation information and parameter information to obtain at least one intent operation category and the respective intent parameters corresponding to each intent operation category includes: Performing user intent recognition based on the operation information and parameter information to obtain at least one intent operation category; Performing intent parameter filling according to the intent operation category, the operation information, and the parameter information to obtain the respective intent parameters corresponding to each intent operation category.

4. The method according to claim 3, wherein The performing user intent recognition based on the operation information and parameter information to obtain at least one intent operation category includes: Using a pre-trained semantic understanding model to analyze and infer the operation information and parameter information to obtain potential user intent keywords; Classifying the potential user intent keywords into predefined operation categories to obtain at least one intent operation category.

5. The method according to claim 1, characterized in that The determining the similarity between the intent operation category and each graph control in the preset knowledge graph according to the intent operation category and the corresponding intent parameters for any intent operation category includes: For any intent operation category, generating a user intent vector from the intent operation category and the corresponding intent parameters; For any user intent vector, calculating the similarity between the user intent vector and each graph control vector in the preset knowledge graph to obtain the similarity between the intent operation category and each graph control.

6. The method according to claim 1, wherein The recommending the graph controls corresponding to the similarities that meet the preset similarity conditions includes: For any intent operation category, recommending the graph control with the highest similarity to the intent operation category.

7. The method according to claim 1, characterized in that The method further includes: In the case of monitoring that the user selects any recommended graph control, pre-filling the graph control with the intent parameters; Editing and displaying the pre-filled graph control, and labeling the pre-filled content in a preset manner to facilitate the user to re-edit the pre-filled graph control.

8. An RPA process assisted design device for a large model collaborative knowledge graph, characterized in that, including: An extraction module for obtaining the completed RPA code of the user and extracting the operation information and parameter information in the completed RPA code; An intention recognition module, configured to perform user intention recognition based on the operation information and parameter information, so as to obtain at least one intention operation category and the intention parameters corresponding to each intention operation category respectively; A similarity comparison module, configured to, for any intention operation category, determine the similarity between the intention operation category and each graph control in a preset knowledge graph according to the intention operation category and the corresponding intention parameters; the graph control is a code control pre-stored in the preset knowledge graph for realizing the RPA target; A recommendation assistance module, configured to recommend the graph controls corresponding to the similarities that meet the preset similarity conditions to assist the user in designing the RPA process.

9. An RPA process assisted design device for a large model collaborative knowledge graph, characterized in that, Comprising: One or more processors; A storage device, configured to store one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors implement the RPA process assisted design method of the large model collaborative knowledge graph according to any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the RPA process assisted design method of the large model collaborative knowledge graph according to any one of claims 1-7.

Images