Training method of script link generation large model and script link generation method

By obtaining the structural information of action scripts and training large language models to generate script links, the problems of low development efficiency and duplication of scripts in the existing technology are solved, and efficient and intelligent script link generation and development are achieved.

CN120104114APending Publication Date: 2025-06-06CHENGDU DESAY SV KAWA TECHNOLOGY CO LTD
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202510181065.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-02-19
Publication Date
2025-06-06

AI Technical Summary

Technical Problem

When developing intelligent applications, the existing technology requires complete full-process scripts, which are large and not flexible enough, resulting in low development efficiency and a large number of similar or repeated scripts, wasting development manpower and storage resources.

Method used

By obtaining the structure information of the action script, including execution actions, preconditions and neighboring script relationship information, the prompt text is generated and input into the large language model for training, and the script link is obtained to generate a large model. This model is used to generate script links, thereby determining the execution order of different scripts.

Benefits of technology

It improves the intelligence of system development, saves development time and manpower, reduces the use of storage resources, and realizes flexible management and efficient development of scripts.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120104114A_ABST
    Figure CN120104114A_ABST
Patent Text Reader

Abstract

The invention discloses a training method of a script link generation large model and a script link generation method. The training method comprises the steps that structure information of at least one action script is obtained, the structure information of the action script comprises an execution action, a precondition and adjacent script relation information, and the adjacent script relation information is used for indicating a next action script which can be executed after the action script is executed; generating at least one first prompt text according to the structure information of each action script; each first prompt text is input into the large language model for training, a script link generation large model is obtained, the script link generation large model is used for generating script links, the system development intelligence degree is improved, time is saved, a complete process script executed by operation is split into independent action scripts, and the operation efficiency is improved. The script link is generated through the large model, a full-process script does not need to be compiled completely, the workload is small, development manpower and storage resources are saved, management is convenient, and different functions are more flexible and changeable to achieve.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a training method for a script link generation large model and a script link generation method. Background Art

[0002] At present, there are many intelligent applications for script execution tasks on smart devices, such as voice assistant multi-step tasks, trigger conditions and execution actions of smart scene task masters. The script consists of one or more actions, and a single action simulates the user's manual operation (clicking text, icons, sliding lists, entering text into text boxes, etc.). For example, the voice grabbing red envelope function of the mobile phone voice assistant, when the user says to the voice assistant: "Grab the WeChat red envelope", the voice assistant will call out the WeChat red envelope grabbing script after receiving the command, and execute it step by step according to the script. The first step is to click on the red envelope message, the second step is to enter WeChat, the third step is to click on the red envelope, the fourth step is to open the red envelope, and finally end the entire process of grabbing the red envelope. In the traditional solution, the script is pre-edited, and each step has a fixed process execution order. This solution requires a complete script for the entire process for each function, which is a large workload, and the control action is single and not flexible enough. The more functions, the more script data, which is also a big problem for storage management; and all control functions need to rely on manual exhaustive execution steps of all functions, and the development efficiency is low. In addition, there may be a large number of similar or repeated scripts for different functions, which greatly wastes development manpower and storage resources, and the development process is not intelligent enough. Summary of the invention

[0003] The present invention provides a script link generation large model training method and a script link generation method, so as to improve the intelligent degree of system development and save time.

[0004] According to one aspect of the present invention, a method for training a script link to generate a large model is provided, comprising:

[0005] Acquire structural information of at least one action script, wherein the structural information of the action script includes an execution action, a precondition, and adjacent script relationship information, wherein the adjacent script relationship information is used to indicate a next action script that can be executed after the action script is executed;

[0006] Generate at least one first prompt text according to the structural information of each of the action scripts;

[0007] Each of the first prompt texts is input into a large language model for training to obtain a script link generation large model, and the script link generation large model is used to generate a script link.

[0008] According to another aspect of the present invention, a script link generation method is provided, comprising:

[0009] Acquire target operation requirement information, and generate a second prompt text according to the target operation requirement information;

[0010] Inputting the second prompt text into a pre-trained script link generation model, wherein the script link generation model is trained using the script link generation model training method described in any embodiment of the present application;

[0011] The target script link is determined according to the output of the script link generation large model.

[0012] According to another aspect of the present invention, there is provided a training device for generating a large model by script link, comprising:

[0013] An action script acquisition module, used to acquire structural information of at least one action script, wherein the structural information of the action script includes an execution action, a precondition, and adjacent script relationship information, wherein the adjacent script relationship information is used to indicate the next action script that can be executed after the action script is executed;

[0014] A first text generation module, used for generating at least one first prompt text according to the structural information of each action script;

[0015] The model training module is used to input each of the first prompt texts into a large language model for training to obtain a script link generation large model, and the script link generation large model is used to generate a script link.

[0016] According to another aspect of the present invention, there is provided a script link generating device, comprising:

[0017] A second text generation module, used to obtain target operation requirement information and generate a second prompt text according to the target operation requirement information;

[0018] A model input module, used for inputting the second prompt text into a pre-trained script link generation model, wherein the script link generation model is trained using the script link generation model training method described in any embodiment of the present invention;

[0019] The script link generation module is used to determine the target script link according to the output of the script link generation large model.

[0020] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising:

[0021] at least one processor, and a memory communicatively coupled to the at least one processor;

[0022] In which, the memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the script link generation large model training method or script link generation method described in any embodiment of the present invention.

[0023] According to another aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the script link generation large model training method or script link generation method described in any embodiment of the present invention when executed.

[0024] According to another aspect of the present invention, a computer program product is provided, which includes a computer program, and when the computer program is executed by a processor, it implements the script link generation large model training method or script link generation method described in any embodiment of the present invention.

[0025] The technical solution of the embodiment of the present invention obtains structural information of at least one action script, wherein the structural information of the action script includes execution actions, preconditions and adjacent script relationship information, wherein the adjacent script relationship information is used to indicate the next action script that can be executed after the action script is executed; generates at least one first prompt text according to the structural information of each action script; inputs each first prompt text into a large language model for training to obtain a script link generation large model, wherein the script link generation large model is used to generate a script link. The method provided by the present application can effectively improve the intelligence level of system development and save time, and splits a complete process script of an operation execution into separate action scripts, wherein the structural information of the action script includes execution actions, preconditions and adjacent script relationship information, wherein the adjacent script relationship information is used to indicate the next action script that can be executed after the action script is executed; and generates at least one first prompt text according to the structural information of each action script. Generate at least one first prompt text, train a large language model through the first prompt text, the large language model learns the meaning of the data contained in the structural information of the action script, and learns the execution order of each action script according to the adjacent script relationship information, and obtains a script link generation large model. The script link generation large model is used to generate script links, and the execution order of different scripts is determined through script links. There is no need to write a complete full-process script, and the workload is small. In addition, generating script links through the script link generation large model can effectively improve the intelligence level of system development. There is no need to develop all scripts one by one, and the scripts are disassembled into separate action scripts. There is no need to enumerate the execution steps of each function, and there will be no large number of similar or repeated scripts. It can effectively save development time, improve development efficiency, save development manpower and storage resources, facilitate script management, and be more flexible and changeable when implementing different functions.

[0026] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0028] Figure 1 This is a flowchart of a training method for generating a large model from a script link according to Embodiment 1 of the present invention;

[0029] Figure 2 is an example diagram of a script sequence provided according to Embodiment 1 of the present invention;

[0030] Figure 3 is an example diagram of another script sequence provided according to the first embodiment of the present invention;

[0031] Figure 4 This is a flowchart of a training method for generating a large model using a script link according to Embodiment 2 of the present invention;

[0032] Figure 5 is an example diagram of structural information of an action script provided according to Embodiment 2 of the present invention;

[0033] Figure 6 is an example diagram of an action script library provided according to Embodiment 2 of the present invention;

[0034] Figure 7 is a flowchart of a script link generation method provided in Embodiment 3 of the present invention;

[0035] Figure 8 is a flowchart of a script link generation method provided according to Embodiment 4 of the present invention;

[0036] Fig. 9 is a schematic diagram of dynamically generating a script sequence based on a large model according to a fourth embodiment of the present invention;

[0037] Fig.10 It is a structural diagram of a training device for generating a large model of a script link provided in Embodiment 5 of the present invention;

[0038] Fig.11 is a structural schematic diagram of a script link generating device provided according to Embodiment 6 of the present invention;

[0039] Fig.12 It is a structural schematic diagram of an electronic device provided according to Embodiment 7 of the present invention. DETAILED DESCRIPTION

[0040] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0041] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0042] Embodiment 1

[0043] Figure 1 This is a flowchart of a training method for generating a large model from a script link provided in the first embodiment of the present invention. This embodiment is applicable to the case of developing scripts efficiently. The method can be executed by a training device for generating a large model from a script link. The training device for generating a large model from a script link can be implemented in the form of hardware and / or software. The training device for generating a large model from a script link can be configured in an electronic device. Figure 1 As shown, the method includes:

[0044] S101. Obtain structural information of at least one action script, where the structural information of the action script includes an execution action, a precondition, and adjacent script relationship information, where the adjacent script relationship information is used to indicate the next action script that can be executed after the action script is executed.

[0045] In this embodiment, the action script can be understood as a script that simulates the user's operation action, for example, simulating the action of the user clicking on the APP of an application on the main interface. The execution action can be operations such as clicking, sliding, and inputting; the precondition can be understood as the conditions that need to be satisfied for the action to be executed, for example, the click operation can be performed on the main interface at present; the adjacent script relationship information can be understood as the information describing other action scripts adjacent to the action script, and the adjacent script relationship information is used to indicate the next action script that can be executed after the action script is executed. The adjacent script relationship information may include information of one or more action scripts. The number of the next action script that can be executed after the action script is executed can be 0, 1, or more. 0 means that there is no next action script that can be executed after the execution of the action script, 1 means that there is only 1 next action script that can be executed after the execution of the action script, and more means that the number of the next action script that can be executed after the execution of the action script is more than one, and one action script can be selected for execution.

[0046] Pre-analyze the execution of different actions. Take the application of the method provided in the embodiment of the present application to the vehicle system as an example. Different applications can be installed on the vehicle system, and users can operate different applications through different actions, such as clicking on the interface of the application, entering information in different windows, etc. By analyzing the different actions of the user, the prerequisites for the execution of the action, the next action after the action is executed, and other information, the structural information of the action script is generated, and the structural information of the action script is saved to the corresponding storage space, such as a database. When performing model training, the structural information of the action script is obtained from the corresponding storage space; the model training in the embodiment of the present application can be triggered by user operation, or training conditions can be set, and model training is automatically triggered when the training conditions are met, and so on.

[0047] Each action script has unique identification information to distinguish different action scripts. For example, each action script is numbered from 1 to N to distinguish different action scripts.

[0048] S102: Generate at least one first prompt text according to the structural information of each action script.

[0049] In this embodiment, the first prompt text can be understood as a text used to prompt the large language model, for example, prompt. The structural information of each action script is written into the prompt text, and at least one first prompt text is generated according to the sample generated by the model provided by the action script.

[0050] S103. Input each first prompt text into a large language model for training to obtain a script link generation large model, and the script link generation large model is used to generate a script link.

[0051] In this embodiment, the script link generation large model can be understood as a large model for generating script links. The script link includes different action scripts. The action scripts that need to be executed and the execution order of the action scripts can be determined according to the script link.

[0052] Each first prompt text is sequentially input into the large language model to guide the large language model to be trained. During the training process, the large language model learns the meaning of the data contained in the action script, and learns the execution order of each action script according to the adjacent script relationship information, and obtains the script link generation model. The trained script link generation model can be used to generate a script link, for example, the prompt text is input into the script link generation model, and the script link generation model is inferred according to the knowledge learned during the training process to generate a script link. The generated script link can be composed of different action scripts, and the action scripts are sequentially sorted according to the execution order. The action scripts in the script link are executed sequentially to complete the corresponding operation. For example, the script link includes the identification of different action scripts, and the execution process of different action scripts is pre-written into executable code and stored in the execution script library. After the script link is generated, the action script is obtained from the execution script library according to the identification and executed. The method provided in the embodiment of the present application disassembles the script into a separate action script, does not need to exhaustively enumerate the execution steps of each function, and does not have a large number of similar or repeated scripts, which can effectively save development time, improve development efficiency, solve the problem of low system development intelligence and low development efficiency, and the whole development process is more intelligent.

[0053] The embodiments of the present application can be applied to the "see-and-say" function in the in-car voice. The process in which the in-car voice assistant automatically helps the car owner search for movies in application A is taken as an example to illustrate the implementation principle of the prior art. The in-car voice assistant simulates the user's operation and automatically opens application A to help the car owner search for the movie he wants to watch. The software helps the user complete the relatively long action chain of "searching and playing a specified movie from the mobile phone home page to application A". It is necessary to write the following steps that imitate the user's manual operation: 1. Enter the application from the mobile phone home page; 2. Click on the search interface; 3. Then enter the power name; 4. Finally, click on the four steps of play. Each of the above steps can be described by an action script that can be understood by a machine, and there is a fixed execution relationship between the scripts. These four actions and the series relationship constitute the movie script sequence of this playing application A. Exemplarily, Figure 2 An example diagram of a script sequence is provided. The script for entering application A from the home page to play a movie needs to include the four tasks shown in the diagram.

[0054] Similarly, for the process of "going from the mobile phone home page to iQiyi and continuing to play the movie in the history that was not finished last time", a machine execution script sequence is written. The first action of this script sequence is to enter iQiyi from the home page, the second action is to click the history entry, and the third action is to click the power of the most recent history to play. If the voice assistant needs to support the above two functions, two script sequences corresponding to these two functions need to be saved. For example, Figure 3 Another example diagram of a script sequence is provided. The script for entering iQiyi from the home page to play historical films needs to include the three actions shown in the figure.

[0055] By comparing the script sequences of the two functions mentioned above, we can know that the first function and the second function both need to enter iQiyi from the home page in the first step. The two saved action script sequences have an action script with exactly the same function, "From the mobile phone home page to iQiyi", so the content of the script needs to be repeatedly written during the development process. In the traditional solution, there are a large number of similar or repeated contents in each script, which greatly wastes development manpower and storage resources. The embodiment of the present application generates a large model of script links by decomposing each action into independent scripts and obtaining a script link generation model through model training. The script link generation model can generate a script link, and the script link includes a plurality of action scripts. The order relationship of arrangement and combination.

[0056] The embodiment of the present application provides a training method for generating a large model of a script link, which effectively improves the intelligence level of system development, splits a complete process script of an operation execution into separate action scripts, wherein the action script includes an execution action, a precondition and adjacent script relationship information, and the adjacent script relationship information is used to indicate the next action script that can be executed after the action script is executed; generates at least one first prompt text according to each action script, and trains the large language model through the first prompt text. The large language model learns the meaning of the data contained in the action script, and learns the execution order of each action script according to the adjacent script relationship information to obtain a large model for generating a script link. The large model for generating a script link is used to generate a script link, and the execution order of different scripts is determined through the script link. There is no need to write a complete full-process script, and the workload is small. Moreover, generating a script link through the script link generation large model can effectively improve the intelligence level of system development, and there is no need to develop all scripts one by one. The script is disassembled into separate action scripts, and there is no need to enumerate the execution steps of each function, and there will be no large number of similar or repeated scripts. It can effectively save development time, improve development efficiency, save development manpower and storage resources, facilitate script management, and be more flexible and changeable when implementing different functions.

[0057] Embodiment 2

[0058] Figure 4This is a flowchart of a training method for generating a large model from a script link provided in the second embodiment of the present invention. This embodiment is refined on the basis of the above embodiment. Figure 4 As shown, the method includes:

[0059] S201. Obtain historical user operation information and analyze it to determine all execution actions.

[0060] In this embodiment, historical user operation information can be understood as information formed by different operations performed by the user, for example, the user clicks on the operation interface of the application, the user enters information in the search box, etc., the execution time of different operations, etc. Taking the car system as an example, different types of applications can be installed on the car system, and the user can operate on the car system, for example, by clicking and other methods to open different applications. Inside the application, the control of different functions of the application can be achieved by clicking, sliding, inputting and other operations. The embodiment of the present application can collect the above user operations, form historical user operation information and save it. During model training, the historical user operation information is obtained and analyzed to determine the actions performed by the user, the preconditions met when performing the action, the next execution action adjacent to the execution action, etc., and the actions performed by all users are determined by analyzing the historical user operation information. Scripts are written for each execution action, or a pre-written script is obtained to determine the script information for executing this action. The execution action in the embodiment of the present application includes a specific action, and can also include the script information for executing this action.

[0061] S202: for each execution action, determine at least one precondition corresponding to the execution action, and generate corresponding initial action script information according to the execution action and each corresponding precondition, wherein the initial action script information includes the execution action and the precondition.

[0062] In the present embodiment, the initial action script information can be understood as the initially determined script structure information, the initial action script information does not contain all the structure information of the action script, and the initial action script information includes the execution action and the precondition. For each execution action, the precondition corresponding to the execution action is determined, and the precondition corresponding to the execution action can be one or more, that is, each execution action can be executed under different circumstances. The execution action and its corresponding precondition are generated into an initial action script information, and the initial action script information includes an execution action and a precondition. When there are multiple preconditions, the corresponding initial action script information is generated according to the execution action and each precondition.

[0063] S203. For each initial action script information, determine the next execution action according to the execution action and preconditions included in the initial action script information, generate adjacent script relationship information according to the next execution action, write the adjacent script relationship information into the initial action script information, and form the structural information of the action script.

[0064] For each initial action script information, the next execution action to be executed after the execution of the current execution action is completed is determined based on the execution action and preconditions included in the initial action script information. The number of the next execution action can be one or more. Adjacent script relationship information is generated based on the next execution action and written into the initial action script information to form the structural information of the action script.

[0065] The embodiment of the present application can analyze different execution actions and different preconditions corresponding to the same execution action, determine one or preconditions corresponding to each execution action, and the next execution action executed under the precondition, and then generate the structural information of the corresponding action script. The execution action, the preconditions for execution, and the next executable action script are described through the structural information of the action script. The embodiment of the present application can determine the structural information of different action scripts through the above method.

[0066] After the structural information of the action script is generated, manual review may be performed to check whether the structural information of the generated action script is correct.

[0067] The embodiment of the present application can develop and write original independent script units for the system that needs to be controlled. There are enough independent scripts to cover all user-operable page controls or objects, and the actions that users can manually operate in all interfaces of all applications. An action script is written to describe each independent action.

[0068] Optionally, the adjacent script relationship information includes: an array, the length of the array is the total number of action scripts, and each element in the array is used to indicate whether the action script corresponding to the element can be executed after the current action script is executed.

[0069] The adjacent script relationship information includes an array, which indicates whether each action script can be executed; wherein the length of the array is the total number of action scripts, each element in the array corresponds to an action script, and the element can be associated with different action scripts by position. Each element in the array is used to indicate whether the action script corresponding to the element can be executed after the current action script is executed.

[0070] Exemplarily, all action scripts are numbered from 1 to N, and the length of the array is the number of action scripts N. The i-th element in the array is 0, indicating that the i-th script cannot be directly executed after the current action script is executed, and vice versa, the i-th element in the array is 1, indicating that the i-th script can be executed immediately after the current action script is executed. The adjacent script relationship information can be a pointer chain relationship, indicating the next executable script pointer. Figure 5 An example diagram of the structural information of an action script is provided, wherein Action: click(X, Y) may indicate that the execution action is to perform a click operation at the (X, Y) coordinates on the screen, and adjacent script relationship information is represented by an array, namely the adjacent script relationship array in the diagram, wherein the first bit in the array is 0 indicating that the first action script cannot be executed after the execution of this action script is completed. Figure 6 An example diagram of an action script library is provided. Different action scripts are stored in the action script library. Each action script has a unique corresponding identifier. For example, different action scripts are identified by numbers 1, 2, ..., N. Each step that a user can manually operate can be represented by a script unit and stored in the action script library. Figure 6 As shown, each click, slide, and input operation on each interface is represented by an action script, and the execution relationship between each script and other scripts is determined in combination with the actual operation.

[0071] S204: Generate at least one first prompt text according to the structural information of each action script.

[0072] S205. Input each first prompt text into a large language model for training to obtain a script link generation large model, and the script link generation large model is used to generate a script link.

[0073] Optionally, the first prompt text includes: structural information of each action script, script link generation sample and operation requirement information.

[0074] In this embodiment, the script link generation example can be understood as an example indicating how to generate a script link. For example, the script link generation example includes: a precondition, a target end point, and a script link composed of "script 1-script 16-script 8-script 4". Through this script link, the operation from the starting point to the end point can be completed. The precondition is used to determine the starting point of the script, and the target end point is used to determine the end point of the script. The operation requirement information is the information required to simulate the user operation, for example, the precondition and the end point information of the operation. The precondition can determine the starting point of the operation.

[0075] During the training of the large language model, the data structure of a single action script and the meaning of the relationship between adjacent scripts are learned based on each action script. Based on all independent action scripts, the large model reasoning ability is used to find the relationship between adjacent scripts. Through the script link generation examples, the correct script link is learned, and the corresponding script link is generated according to the operation requirement information. The large language model is continuously adjusted by judging whether the generated script link is correct, and finally the script link generation large model is obtained. The script link generation large model obtained after training can accurately generate script links.

[0076] Since the script link generation large model may not perform specific actions during the learning and reasoning process, and may only use preconditions and adjacent script relationship information for learning and reasoning, the first prompt text may only include the preconditions and adjacent script relationship information in the structural information of the action script, and does not include the execution action; that is, the action script includes: the preconditions and adjacent script relationship information in the structural information of each action script, script link generation samples, and operation requirement information. In the process of training the script link generation large model in the embodiment of the present application, all information that affects the script link generation can be input into the large model to guide the large model to learn; in the large model prediction process, all information that affects the script link generation can also be input into the large model so that the large model can perform correct reasoning on the script link.

[0077] Optionally, the structural information of the action script also includes at least one of the following information: user feature information and scenario information.

[0078] In this embodiment, user characteristic information can be understood as information describing the unique characteristics of the user, such as the user's preferences and habits; scene judgment information can be understood as information in different application scenarios, such as environmental information. The structural information of the action script also includes at least one of user characteristic information and scene information, and the user characteristic information and scene information are used to assist the action script in executing the corresponding action.

[0079] After the update conditions are met, the structural information and / or action script of the action script is updated. The updated action script and / or structural information of the action script can directly store or replace the original action script and / or structural information of the action script; the script link generation model can directly generate a script link based on the updated action script, or the script link generation model can be retrained and updated based on the updated action script, and so on. There is no need to change other action scripts. When the number of scripts is large, other action scripts are not affected, which reduces the workload of developers. Among them, the update condition can be that the software function has changed. The change of the software function may cause the action of the user operation to change, so the action script needs to be updated.

[0080] For example, after the software function changes, the traditional script execution scheme of the voice assistant needs to invest in products to design each function before the function is developed, and it is also necessary to invest manpower to write each function script sequence, and then hand it over to the application developer for development. The more functions there are, the workload increases exponentially, and the later function changes require re-modifying the script and updating the software version. The method provided in the embodiment of the present application forms an action script for each execution action separately, which can greatly reduce the script design and script writing workload and manpower investment, and can dynamically generate a large model in combination with the user's actual use process information, without the need for manual modification to add script sequences, and without the need to re-update the software for function upgrades.

[0081] The embodiment of the present application provides a training method for generating a large model of a script link, which can effectively improve the intelligent degree of system development, save time, split a complete process script of an operation execution into separate action scripts, the action script includes execution actions, preconditions and adjacent script relationship information, and the adjacent script relationship information is used to indicate the next action script that can be executed after the action script is executed; the large language model learns the meaning of the data contained in the action script, and learns the execution order of each action script according to the adjacent script relationship information, and obtains a large model for generating a script link, generates a script link through the script link, and determines the execution order of different scripts through the script link, and does not need to fully write the full process script, and the workload is small. In addition, the script link is generated by generating a large model through the script link, which can effectively improve the intelligent degree of system development, and does not need to develop all scripts one by one. The script is disassembled into separate action scripts, and the execution steps of each function do not need to be exhaustive, and there will be no large number of similar or repeated scripts, which can effectively save development time, improve development efficiency, save development manpower and storage resources, facilitate the management of scripts, and be more flexible and changeable when implementing different functions. Reduce the manpower investment in iterative updates of function development; reduce the function update process and shorten the update cycle. Breaking the constraints of fixed scripts, each time a new function is added or modified, there is no need to manually rewrite the script to release the version upgrade, only the large model background needs to be dynamically changed and generated. This greatly reduces the workload and human investment in script design and script writing.

[0082] Embodiment 3

[0083] Figure 7 This is a flowchart of a script link generation method provided in Embodiment 3 of the present invention. This embodiment is applicable to the case of quickly generating a script link. The method can be executed by a script link generation device. The script link generation device can be implemented in the form of hardware and / or software. The script link generation device can be configured in an electronic device. Figure 7 As shown, the method includes:

[0084] S301: Obtain target operation requirement information, and generate a second prompt text according to the target operation requirement information.

[0085] In this embodiment, the target operation requirement information can be understood as the requirement information for simulating the user operation, for example, the preconditions for determining the starting point, and the information for the operation end point. The second prompt text can be understood as the text for prompting the large language model, for example, prompt.

[0086] The target operation requirement information can be determined based on the user's voice command, action command, brain-computer command, the current operating status of the device, and other information. The target operation requirement information is written into the prompt text to generate a second prompt text. The target operation requirement is used to determine the preconditions and the end point of the operation; or the preconditions and the end point of the operation are determined according to the target operation requirement, and written into the prompt text to generate a second prompt text. The second prompt text also includes various action scripts and can also include script link generation samples.

[0087] Exemplarily, the user can output voice commands through voice; or the user can output action commands through hand movements (for example, swinging the wrist, making specified movements with fingers, making specified movements with elbows, etc.), leg movements (for example, stretching legs, stretching feet, etc.), facial movements (for example, blinking, opening mouth, etc.), etc.; or, the user can output brain-computer commands by wearing a dedicated device (such as an EEG cap), and the dedicated device can obtain brain-computer commands by collecting the user's brain wave signals.

[0088] Exemplarily, the method provided in the embodiment of the present application can be deployed on a vehicle system, a smart phone, a tablet computer, a smart watch, an eye tracking device, a VR glasses and other smart devices. Taking deployment on a vehicle system as an example, the driver may have navigation, music playing, phone calls and other needs during driving, but the driver's direct operation of the vehicle system will distract attention and cannot ensure driving safety. Therefore, the driver can control the vehicle system to automatically play music through voice "Please help me play xx music". After the vehicle system obtains the driver's voice command, it analyzes the voice command to determine the target operation demand information. Taking deployment on a smart phone as an example, the user can also output the target operation demand information through voice commands, or the user wears a sensor to collect the user's wrist information. The user can collect the corresponding wrist information in any scene such as work, exercise, doing housework, etc., and determine the target operation demand information through the wrist information. For example, it is detected that the user turns his wrist inward and opens the xx application to continue playing the last played video. The method provided in the embodiment of the present application can also be deployed on an eye movement device. For users who are inconvenient to operate and cannot make voice, the user can wear an eye movement device, collect the user's eye movement instructions through the eye movement device, and then determine the target operation requirement information through the eye movement instructions. For example, the user blinks his left eye, opens the music software to continue playing the music played last time, and the user blinks his left eye twice in succession, opens the music software to continue playing the music played last time and turns up the volume. S302, input the second prompt text into the pre-trained script link generation model, and the script link generation model is trained using the script link generation model training method described in any embodiment of the present invention.

[0089] The second prompt text is input into the pre-trained script link generation model to guide the script link generation model to perform reasoning. The script link generation model is trained using the script link generation model training method described in any embodiment of the present invention.

[0090] S303, determine the target script link based on the output of the script link to generate the large model.

[0091] In this embodiment, the target script link can be understood as a script link obtained by model reasoning. The script link generation model can directly output the script link, and the script link output by the script link generation model is used as the target script link.

[0092] During the training of the large language model, the data structure of a single action script and the meaning of the chain pointer are learned based on each action script. Based on all independent action scripts, the large model reasoning ability is used to find the adjacent script relationship. During the reasoning process, the action script of the starting point and the action script of the end point are first determined, and then the script link from the starting point to the end point is determined. The number of script links from the starting point to the end point can be one or more. The script link generation large model can directly output one of the script links as the target script link, or the script link generation large model can output all script links and select one of them as the target script link.

[0093] The target script link includes different action scripts and the execution order of the action scripts. Through the target script link, different action scripts and their execution order can be determined so that the action scripts can be called to execute corresponding actions and realize different functions. When realizing different functions, there is no need to write a complete execution script, but only to infer the script link through the large model. For the same actions of different functions, there is no need to write them repeatedly, which saves development resources and improves development efficiency.

[0094] The embodiment of the present application provides a script link generation method, which improves the intelligence level of system development. A second prompt text is generated according to the target operation requirement information, and the trained script link generation large model is guided by the second prompt text. The script link generation large model is guided to infer according to the learned instructions to generate a target script link. The execution order of different scripts can be determined through the target script link. There is no need to write a complete full-process script, and the workload is small. Generating script links through the script link generation large model can effectively improve the intelligence level of system development, and there is no need to develop all scripts one by one; the script is disassembled into separate action scripts, and there is no need to enumerate the execution steps of each function, and there will be no large number of similar or repeated scripts, which can effectively save development time, improve development efficiency, save development manpower and storage resources, facilitate script management, and be more flexible and changeable when implementing different functions.

[0095] Embodiment 4

[0096] Figure 8 This is a flowchart of a script link generation method provided by Embodiment 4 of the present invention. This embodiment is refined on the basis of the above embodiment. Figure 8 As shown, the method includes:

[0097] S401: Receive user operation information.

[0098] In this embodiment, the user operation information may be the user's voice command information, for example, the user issues a command through voice, "Please help me use application A to play XX movie." The user operation information is collected through a microphone or other sensors.

[0099] S402: Determine target interface state information according to user operation information, and determine target preconditions according to the state of the current interface.

[0100] In this embodiment, the target interface state information can be understood as the information of the interface that needs to be reached at the end of the execution operation, for example, the target interface state information is to play a specified movie. The target precondition can be understood as the precondition satisfied by the actual interface during this reasoning, which is used to match the precondition in the action script.

[0101] Analyze the user operation information, determine the target point that the user wants to reach, and use it as the target interface state information. For example, if the user operation information is "Please help me use application A to play XX movie", the target interface state information is to play the movie. When analyzing the user operation information, you can also extract all useful information related to the user's needs, and determine the target interface state information based on the extracted information. Read the state of the current interface and get the target precondition. For example, if the current interface state is the main interface, the target precondition is the main interface.

[0102] S403: Generate target operation requirement information according to the target interface state information and target preconditions.

[0103] The target interface state information and the target precondition are used as the target operation requirement information, or the target interface state information and the target precondition are written into the information to form the target operation requirement information.

[0104] S404: Generate a second prompt text according to the target operation requirement information.

[0105] S405. Input the second prompt text into a pre-trained script link generation model. The script link generation model is trained using the script link generation model training method of any embodiment of the present invention.

[0106] S406, determining the target script link according to the output of the script link-generated large model.

[0107] The target precondition is used to determine the starting point, and the target interface status information is used to determine the end point;

[0108] The target script link is the shortest path link from the start point to the end point.

[0109] There can be multiple links from the starting point to the end point. The embodiment of the present application selects the shortest path link as the target script link to ensure that the script execution time is the shortest, execute the corresponding operations at the fastest speed, and improve the response speed.

[0110] Different actions can be performed under the same preconditions. For example, any application can be opened on the main interface, and operations such as search and play can be performed on the interface of the application. Therefore, by inputting the target preconditions into the script link generation model, one or more starting points can be determined; by inputting the target interface status information into the script link generation model, a unique end point can be determined, and the reachable paths between the starting point and the end point can be analyzed to finally determine the shortest path link. The execution action Action can include different parameters, which can correspond to different objects. The object can refer to a single object, or it can refer to a class of objects by dividing the objects according to categories; user operation information can determine the execution end point in combination with the above information; Act ion includes different slots, for example, X and Y in Action: click (X, Y) represent different slots respectively; the slots can be filled according to the user operation information, for example, to play the first episode of "The Legend of Zhen Huan", "The Legend of Zhen Huan" and "The First Episode" are respectively filled into the corresponding slots in the Action of the corresponding action script. When the script link generation model generates the target script link, the corresponding parameters can be directly written into the corresponding parameter positions in the action script according to the user operation information. When executing the action script, it can be executed directly according to the written parameters. For example, when executing the script corresponding to the playback operation, the first episode is played according to the parameters in the slot.

[0111] S407: Determine the action scripts to be called and the calling order of the action scripts to be called according to the target script link.

[0112] The target script link includes different action scripts and a calling order. The action scripts form a link in order. The order of the action scripts in the link is the calling order of the action scripts. That is, all the action scripts connected in series on the target script path can be understood as several steps of operation on the vehicle computer required to complete this function. The order of each action script from the beginning is the order of each operation. The action scripts included in the target script link are used as the action scripts to be called, and the calling order of each action script to be called is determined. Alternatively, the target script link includes the identifiers of different action scripts and the calling order of the action scripts. The action script can be obtained from the action script library according to the identifier of the action script. The calling order of the action script is the link pointer relationship between each execution script.

[0113] S408 , calling each action script to be called in sequence according to the calling order of each action script to be called to execute corresponding actions.

[0114] The action scripts to be called are called in sequence according to the calling order of the action scripts to be called. After each action script to be called is called, the corresponding action corresponding to the action script is executed.

[0115] For example, Fig. 9 A schematic diagram for dynamically generating a script sequence based on a large model is provided; the diagram shows different action scripts and different chain relationships, the action scripts can be stored in an action script library, and each color of the chain relationship is a script link to achieve different functions.

[0116] In the scenario of voice assistant interaction, after the script link generation model generates the script link, it can be dynamically sent to the voice assistant. The voice assistant reads the action script corresponding to each point in the action script library in sequence according to the generated script link, and executes each action script. The whole process can be completed from opening application A to finally playing the movie.

[0117] Optionally, the structural information of the action script includes at least one of the following information: user feature information and scenario information.

[0118] Optionally, the action script to be called performs corresponding actions according to user feature information and / or scene information.

[0119] When the action script to be called is called, it can perform the corresponding action according to at least one of the user characteristic information and the scene information; the user characteristic information and the scene information can be used as information in the execution process of the action Action to assist the action script in performing the corresponding action. For example, the user characteristic information includes at least one of the following: the number of historical uses, the number of times the user actively modifies the action, and the user's like feedback; the scene information is the environmental parameter, and the environmental parameter includes at least one of the following: weather, vehicle body status, date.

[0120] By adding at least one of user feature information and scenario information to the action script, it is possible to generate requirements such as user preference analysis, user-defined scenarios, AI push, etc., and realize global arbitrary multi-step action combination control of the terminal through dynamic script sequence relationship generation.

[0121] When the software function changes, the action script that executes the action can be directly updated without modifying other action scripts. This can greatly reduce the workload and manpower investment in script design and script writing, and can dynamically generate large models based on the user's actual usage process information. There is no need for manual modification to add script sequences, nor is there a need to re-update the software to upgrade functions.

[0122] The structural information of the action script also includes: execution actions, preconditions and adjacent script relationship information.

[0123] The embodiment of the present application provides a script link generation method, which effectively improves the intelligent degree of system development and saves time. The target interface state information is determined through user operation information, the target precondition is determined according to the state of the current interface, and the target operation requirement information is generated according to the target interface state information and the target precondition, so that the script link generation model determines the starting point according to the target precondition, determines the end point according to the target interface state information, and generates the target script link with the shortest path based on the starting point and the end point. The execution order of different scripts is determined through the script link, and there is no need to write a complete full-process script, and the workload is small. The script link is generated through the script link generation model, which can effectively improve the intelligent degree of system development, and there is no need to develop all scripts one by one; the script is disassembled into separate action scripts, and there is no need to exhaustively enumerate the execution steps of each function, and there will be no large number of similar or repeated scripts, which can effectively save development time, improve development efficiency, save development manpower and storage resources, facilitate the management of scripts, and be more flexible and changeable when implementing different functions. Reduce the manpower investment in iterative updates of function development; reduce the function update process and shorten the update cycle. Breaking the constraints of fixed scripts, each time a new function is added or modified, there is no need to manually rewrite the script to release the version upgrade, only the large model background needs to be dynamically changed and generated. This greatly reduces the workload and human investment in script design and script writing.

[0124] Embodiment 5

[0125] Fig.10 This is a schematic diagram of the structure of a training device for generating a large model using a script link provided in Embodiment 5 of the present invention. Fig.10 As shown, the device includes: an action script acquisition module 51, a first text generation module 52 and a model training module 53.

[0126] An action script acquisition module 51, used to acquire structural information of at least one action script, wherein the structural information of the action script includes an execution action, a precondition, and adjacent script relationship information, wherein the adjacent script relationship information is used to indicate the next action script that can be executed after the action script is executed;

[0127] A first text generating module 52, configured to generate at least one first prompt text according to the structural information of each of the action scripts;

[0128] The model training module 53 is used to input each of the first prompt texts into a large language model for training to obtain a script link generation large model, and the script link generation large model is used to generate a script link.

[0129] The embodiment of the present application provides a training device for generating a large model of a script link, which effectively improves the intelligence level of system development and saves time. The complete process script of an operation execution is split into separate action scripts, wherein the action script includes an execution action, a precondition and adjacent script relationship information, and the adjacent script relationship information is used to indicate the next action script that can be executed after the action script is executed; at least one first prompt text is generated according to each action script, and the large language model is trained through the first prompt text. The large language model learns the meaning of the data contained in the action script, and learns the execution order of each action script according to the adjacent script relationship information to obtain a script link generation large model. The script link generation large model is used to generate a script link, and the execution order of different scripts is determined through the script link. There is no need to completely write the full process script, and the workload is small. In addition, by generating a script link through the script link generation large model, the intelligence level of system development can be effectively improved, and all scripts do not need to be developed one by one. The script is disassembled into separate action scripts, and there is no need to exhaustively enumerate the execution steps of each function, and there will be no large number of similar or repeated scripts. It can effectively save development time, improve development efficiency, save development manpower and storage resources, facilitate script management, and be more flexible and changeable when implementing different functions.

[0130] Optionally, the action script acquisition module 51 includes:

[0131] The historical information acquisition module is used to obtain historical user operation information and analyze it to determine all executed actions;

[0132] An initial script generation module, used for determining, for each execution action, at least one precondition corresponding to the execution action, and generating corresponding initial action script information according to the execution action and each corresponding precondition, wherein the initial action script information includes the execution action and the precondition;

[0133] An action script generation module is used to determine the next execution action for each initial action script information based on the execution action and preconditions included in the initial action script information, and generate adjacent script relationship information based on the next execution action, and write the adjacent script relationship information into the initial action script information to form the structural information of the action script.

[0134] Optionally, the adjacent script relationship information includes: an array, the length of the array is the total number of action scripts, and each element in the array is used to indicate whether the action script corresponding to the element can be executed after the current action script is executed.

[0135] Optionally, the first prompt text includes: structural information of each action script, script link generation sample and operation requirement information.

[0136] Optionally, the structural information of the action script also includes at least one of the following information: user feature information and scenario information.

[0137] The training device for generating a large model from a script link provided in an embodiment of the present invention can execute the training method for generating a large model from a script link provided in any embodiment of the present invention, and has the functional modules and beneficial effects corresponding to the execution method.

[0138] Embodiment 6

[0139] Fig.11 This is a schematic diagram of the structure of a training device for generating a large model using a script link provided in Embodiment 6 of the present invention. Fig.11 As shown, the device includes: a second text generating module 61, a model input module 62 and a script link generating module 63.

[0140] A second text generation module 61 is used to obtain target operation requirement information and generate a second prompt text according to the target operation requirement information;

[0141] A model input module 62, used for inputting the second prompt text into a pre-trained script link generation model, wherein the script link generation model is trained by using the script link generation model training method described in any embodiment of the present invention;

[0142] The script link generation module 63 is used to determine the target script link according to the output of the script link generation large model.

[0143] The embodiment of the present application provides a script link generation device, which effectively improves the intelligence level of system development and saves time. A second prompt text is generated according to the target operation requirement information, and the trained script link generation large model is guided by the second prompt text. The script link generation large model is guided to infer according to the learned instructions to generate a target script link. The execution order of different scripts can be determined through the target script link. There is no need to write a complete full-process script, and the workload is small. In addition, by generating a script link through the script link generation large model, the intelligence level of system development can be effectively improved. There is no need to develop all scripts one by one, and the scripts can be disassembled into separate action scripts. There is no need to enumerate the execution steps of each function, and there will be no large number of similar or repeated scripts. It can effectively save development time, improve development efficiency, save development manpower and storage resources, facilitate script management, and be more flexible and changeable when implementing different functions.

[0144] Optionally, the second text generation module 61 includes:

[0145] An operation information receiving unit, used to receive user operation information;

[0146] A target information determination unit, configured to determine target interface state information according to the user operation information, and determine target preconditions according to the state of the current interface;

[0147] The target requirement generating unit is used to generate target operation requirement information according to the target interface state information and the target precondition.

[0148] The target precondition is used to determine the starting point, and the target interface state information is used to determine the end point;

[0149] The target script link is the shortest path link from the starting point to the end point.

[0150] Optionally, the device further comprises:

[0151] A script-to-be-called determining unit, used to determine the action scripts to be called and the calling order of the action scripts to be called according to the target script link;

[0152] The script calling unit is used to call the action scripts to be called in sequence according to the calling order of the action scripts to be called to execute corresponding actions.

[0153] Optionally, the structural information of the action script includes at least one of the following information: user feature information and scenario information;

[0154] The action script to be called executes corresponding actions according to the user feature information and / or scenario information.

[0155] The script link generation device provided in the embodiment of the present invention can execute the script link generation method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0156] Embodiment 7

[0157] Fig.12 A schematic diagram of the structure of an electronic device provided for implementing Embodiment VII of the present invention. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required herein.

[0158] like Fig.12As shown, the electronic device 70 includes at least one processor 71, and a memory connected to the at least one processor 71, such as a read-only memory (ROM) 72, a random access memory (RAM) 73, etc., wherein the memory stores a computer program that can be executed by at least one processor, and the processor 71 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 72 or the computer program loaded from the storage unit 78 to the random access memory (RAM) 73. In the RAM 73, various programs and data required for the operation of the electronic device 70 can also be stored. The processor 71, the ROM 72, and the RAM 73 are connected to each other via a bus 74. An input / output (I / O) interface 75 is also connected to the bus 74.

[0159] A number of components in the electronic device 70 are connected to the I / O interface 75, including: an input unit 76, such as a keyboard, a mouse, etc.; an output unit 77, such as various types of displays, speakers, etc.; a storage unit 78, such as a disk, an optical disk, etc.; and a communication unit 79, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 79 allows the electronic device 70 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0160] The processor 71 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the processor 71 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The processor 71 executes the various methods and processes described above, such as a training method for generating a large model of script links or a script link generation method.

[0161] In some embodiments, the training method of the script link generation large model or the script link generation method may be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 78. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 70 via the ROM 72 and / or the communication unit 79. When the computer program is loaded into the RAM 73 and executed by the processor 71, one or more steps of the training method of the script link generation large model or the script link generation method described above may be executed. Alternatively, in other embodiments, the processor 71 may be configured to execute the training method of the script link generation large model or the script link generation method by any other appropriate means (e.g., by means of firmware).

[0162] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0163] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when the computer program is executed by the processor, the functions / operations specified in the flow chart and / or block diagram are implemented. The computer program may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0164] An embodiment of the present invention provides a computer program product, which includes a computer program. When the computer program is executed by a processor, it implements the script link generation large model training method or script link generation method described in any embodiment of the present invention.

[0165] In the context of the present invention, a computer-readable storage medium may be a tangible medium that may contain or store a computer program for use by or in combination with an instruction execution system, device or equipment. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0166] To provide interaction with a user, the systems and techniques described herein may be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form (including acoustic input, voice input, or tactile input).

[0167] The systems and techniques described herein may be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0168] A computing system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The client and server relationship is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services.

[0169] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of the present invention can be achieved, and this document does not limit this.

[0170] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A training method for generating a large model using a script link, characterized in that: include: Acquire structural information of at least one action script, wherein the structural information of the action script includes an execution action, a precondition, and adjacent script relationship information, wherein the adjacent script relationship information is used to indicate a next action script that can be executed after the action script is executed; Generate at least one first prompt text according to the structural information of each of the action scripts; Each of the first prompt texts is input into a large language model for training to obtain a script link generation large model, and the script link generation large model is used to generate a script link.

2. The method according to claim 1, characterized in that The obtaining of structural information of at least one action script includes: Obtain historical user operation information and analyze it to determine all executed actions; For each execution action, determine at least one precondition corresponding to the execution action, and generate corresponding initial action script information according to the execution action and each corresponding precondition, wherein the initial action script information includes the execution action and the precondition; For each initial action script information, the next execution action is determined according to the execution action and preconditions included in the initial action script information, and adjacent script relationship information is generated according to the next execution action, and the adjacent script relationship information is written into the initial action script information to form the structural information of the action script.

3. The method according to claim 1, characterized in that The adjacent script relationship information includes: an array, the length of the array is the total number of action scripts, and each element in the array is used to indicate whether the action script corresponding to the element can be executed after the current action script is executed.

4. The method according to claim 1, characterized in that The first prompt text includes: structural information of each action script, script link generation sample and operation requirement information.

5. The method according to any one of claims 1 to 4, characterized in that The structural information of the action script also includes at least one of the following information: user feature information and scene information.

6. A script link generation method, characterized in that: include: Acquire target operation requirement information, and generate a second prompt text according to the target operation requirement information; Inputting the second prompt text into a pre-trained script link generation model, wherein the script link generation model is trained using the script link generation model training method according to any one of claims 1 to 5; The target script link is determined according to the output of the script link generation large model.

7. The method according to claim 6, characterized in that Obtain target operation requirement information, including: Receive user operation information; Determine the target interface state information according to the user operation information, and determine the target precondition according to the state of the current interface; Target operation requirement information is generated according to the target interface state information and the target precondition.

8. The method according to claim 7, characterized in that The target precondition is used to determine the starting point, and the target interface state information is used to determine the end point; The target script link is the shortest path link from the starting point to the end point.

9. The method according to any one of claims 6 to 8, characterized in that: Also includes: Determine the action scripts to be called according to the target script link, and the calling order of the action scripts to be called; The action scripts to be called are called in sequence according to the calling order of the action scripts to be called to execute corresponding actions.

10. The method according to claim 9, characterized in that The structural information of the action script includes at least one of the following information: user feature information and scenario information; The action script to be called executes corresponding actions according to the user feature information and / or scenario information.

11. A training device for generating a large model using a script link, characterized in that: include: An action script acquisition module, used to acquire structural information of at least one action script, wherein the structural information of the action script includes an execution action, a precondition, and adjacent script relationship information, wherein the adjacent script relationship information is used to indicate the next action script that can be executed after the action script is executed; A first text generation module, used for generating at least one first prompt text according to the structural information of each action script; The model training module is used to input each of the first prompt texts into a large language model for training to obtain a script link generation large model, and the script link generation large model is used to generate a script link.

12. A script link generation device, characterized in that: include: A second text generation module, used to obtain target operation requirement information and generate a second prompt text according to the target operation requirement information; A model input module, used for inputting the second prompt text into a pre-trained script link generation model, wherein the script link generation model is trained by the script link generation model training method according to any one of claims 1 to 5; The script link generation module is used to determine the target script link according to the output of the script link generation large model.

13. An electronic device, characterized in that: The electronic device comprises: at least one processor, and a memory communicatively coupled to the at least one processor; Wherein, the memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the training method for the script link generation large model described in any one of claims 1-5 or the script link generation method described in any one of claims 6-10.

14. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, which are used to enable a processor to implement the script link generation large model training method described in any one of claims 1-5 or the script link generation method described in any one of claims 6-10 when executed.

15. A computer program product, characterized in that The computer program product comprises a computer program which, when executed by a processor, implements the training method for generating a large script link model according to any one of claims 1 to 5 or the script link generation method according to any one of claims 6 to 10.

Citation Information

Cited By

  • Function link display method, debugging device, intelligent wearable device and debugging system

    CN120929181A