Task processing method and device based on large model intelligent agent and electronic equipment

Abstract

The invention provides a task processing method and device based on large model intelligent proxy and electronic equipment, and relates to the technical field of artificial intelligence. The method comprises the steps of obtaining input information of a user to determine task content and a task type of a to-be-processed task; determining one or more candidate application programs associated with the task content and an execution time sequence of the candidate application programs; the candidate application programs are called according to the execution time sequence, prompt information and XML description of the target application programs are constructed for the called target application programs, a target interaction process between the target application programs is determined based on the prompt information and the XML description, and the target application programs are called to execute the target interaction process; and obtaining a task result of the to-be-processed task in response to the completion of calling of the candidate application program according to the scheduling time sequence. Therefore, according to the scheme, the large model automatically schedules a plurality of application programs to realize cooperative processing according to the input information of the user, so that more efficient task cooperation is realized.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, and in particular to a task processing method, device and electronic device based on a large model intelligent agent. Background Art

[0002] In the internal office business scenario of an enterprise, since there are multiple office applications in the internal office of the enterprise, and the completion of some business scenarios requires the collaboration across multiple applications. The current general processing method still requires employees to operate multiple applications on their mobile phones or computers to complete the processing in sequence, which is somewhat inconvenient. Summary of the invention

[0003] The present invention provides a task processing method, device and electronic device based on a large model intelligent agent, so as to solve the problem that multiple applications cannot be called automatically.

[0004] To this end, one purpose of the present disclosure is to propose a task processing method based on a large model intelligent agent.

[0005] The second objective of the present disclosure is to propose a task processing device based on a large model intelligent agent.

[0006] A third objective of the present disclosure is to provide an electronic device.

[0007] A fourth object of the present disclosure is to provide a non-transitory computer-readable storage medium.

[0008] A fifth object of the present disclosure is to provide a computer program product.

[0009] To achieve the above-mentioned purpose, the first aspect of the present disclosure proposes a task processing method based on a large model intelligent agent, including: obtaining user input information, and determining the task content and task type of the task to be processed based on the input information; determining, by the target large model, one or more candidate applications associated with the task content and the execution sequence of the candidate applications according to the task content and the task type; calling the candidate applications according to the execution sequence, constructing prompt information and extensible markup language XML description of the target application for the called target application, and inputting the prompt information, XML description and mask page image of the target application into the target large model, determining the target interaction process between the target applications, and calling the target application to execute the target interaction process; in response to the candidate application ending the call according to the scheduling sequence, obtaining the task result of the task to be processed.

[0010] According to an embodiment of the present disclosure, it is characterized in that the process of obtaining the XML description of the target application includes: determining the task sub-content corresponding to the target application based on the task content; obtaining a first page image and a second page image of the target application, wherein the first page image is an image of a start page of the target application, and the second image is an image of an end page after the target application executes the task sub-content; based on the first page image and the second page image, obtaining the interactive elements of the target application, and determining the XML description of the target application based on the interactive elements.

[0011] According to one embodiment of the present disclosure, obtaining the interactive elements of the target application based on the first page image and the second page image includes: parsing the first page image and the second page image to determine a first candidate page element corresponding to the first page image, and a second candidate page element corresponding to the second page image; obtaining attribute information of the first candidate page element and the second candidate page element, and determining the interactive elements in the first candidate page element and the second candidate page element based on the attribute information.

[0012] According to one embodiment of the present disclosure, determining the XML description of the target application based on the interactive element includes: determining the temporal relationship between the first page image and the second page image; determining the correspondence between the first page image, the second page image and the interactive element; and performing structured processing on the temporal relationship between the first page image and the second page image and the correspondence to obtain an XML description of the target application.

[0013] According to one embodiment of the present disclosure, determining the correspondence between the first page image, the second page image and the interactive element includes: determining path information for jumping from the first page image to the second page image, the path information including the interactive element used during the jump; monitoring the path information to determine the correspondence between the first page image, the second page image and the interactive element.

[0014] According to one embodiment of the present disclosure, the process of obtaining prompt information of the target application includes: predicting the interaction actions of the target application by the target big model to obtain the interaction prediction result of the target application; determining the target application of the previous call based on the execution timing, and obtaining the historical interaction process of the target application of the previous call; determining the prompt information of the target application based on the interaction prediction result and the historical interaction process.

[0015] According to one embodiment of the present disclosure, the masked page image acquisition process of the target application includes: for any page image of the first page image and the second page image, traversing the interactive elements in the XML description, and masking any page image based on the interactive elements to obtain the masked page image.

[0016] According to one embodiment of the present disclosure, the prompt information, XML description and mask page image of the target application are input into the target big model to determine the target interaction process between the target applications, including: the target big model parses the XML description according to the prompt information to determine the interaction logic of the target application; based on the prompt information, the mask page image is parsed to determine the page layout information of the target application, the page layout information includes location information of the interaction elements; based on the interaction logic and the location information, the candidate interaction process corresponding to the target application is determined; the candidate interaction process of each target application in the execution sequence is obtained, and the target interaction process between the target applications is determined according to the candidate interaction process.

[0017] According to one embodiment of the present disclosure, determining the task content and task type of the task to be processed based on the input information includes: performing intent recognition on the input information, determining invalid information in the input information, and filtering the invalid information to obtain valid information in the input information; performing task recognition on the valid information to determine the task content and task type of the task to be processed.

[0018] According to one embodiment of the present disclosure, the training process of the target large model includes: determining multiple sample applications corresponding to the sample interaction process, and a sample candidate interaction process corresponding to each sample application; based on the sample candidate interaction process, training the initial large model until the training is completed to obtain the target large model.

[0019] According to one embodiment of the present disclosure, obtaining the sample interaction process includes: obtaining a first sample page image and a second sample page image of a sample application corresponding to the sample candidate interaction process; obtaining a sample interaction element of the sample application based on the first sample page image and the second sample page image, and determining a sample XML description of the sample application based on the sample interaction element; for any sample page image of the first sample page image and the second sample page image, traversing the sample interaction elements in the sample XML description, and masking any sample page image based on the sample interaction elements to obtain a sample masked page image; determining the sample interaction process based on the sample XML description and the sample masked page image.

[0020] According to one embodiment of the present disclosure, the initial large model is trained based on the sample candidate interaction process until the training is completed to obtain the target large model, including: executing the sample candidate interaction process on the sample application, and encapsulating the interaction process to obtain a sample encapsulation result; inputting the sample encapsulation result into the initial large model, and the initial large model traverses the sample encapsulation result to obtain operation information of each sample candidate interaction process; generating training prompt information based on the operation information, and training the initial large model based on the training prompt information to obtain a training result; judging whether to continue training the initial large model according to the training result, until the training result meets the training end condition to obtain the target large model.

[0021] To achieve the above-mentioned purpose, the second aspect of the present disclosure proposes a task processing device based on a large model intelligent agent, including: a first determination module, used to obtain user input information, and determine the task content and task type of the task to be processed based on the input information; a second determination module, used to determine one or more candidate applications associated with the task content and the execution sequence of the candidate applications according to the task content and the task type by the target large model; a third determination module, used to call the candidate application according to the execution sequence, and for the called target application, construct prompt information and extensible markup language XML description of the target application, and input the prompt information, XML description and mask page image of the target application into the target large model, determine the target interaction process between the target applications, and call the target application to execute the target interaction process; a fourth determination module, used to obtain the task result of the task to be processed in response to the completion of the call of the candidate application according to the execution sequence.

[0022] According to an embodiment of the present disclosure, it is characterized in that the third determination module is also used to: determine the task sub-content corresponding to the target application based on the task content; obtain a first page image and a second page image of the target application, wherein the first page image is an image of a start page of the target application, and the second image is an image of an end page after the target application executes the task sub-content; obtain interactive elements of the target application based on the first page image and the second page image, and determine an XML description of the target application based on the interactive elements.

[0023] According to one embodiment of the present disclosure, the third determination module is also used to: parse the first page image and the second page image to determine a first candidate page element corresponding to the first page image and a second candidate page element corresponding to the second page image; obtain attribute information of the first candidate page element and the second candidate page element, and determine the interactive elements in the first candidate page element and the second candidate page element based on the attribute information.

[0024] According to one embodiment of the present disclosure, the third determination module is also used to: determine the temporal relationship between the first page image and the second page image; determine the correspondence between the first page image, the second page image and the interactive element; and perform structured processing on the temporal relationship between the first page image and the second page image and the correspondence to obtain an XML description of the target application.

[0025] According to one embodiment of the present disclosure, the third determination module is also used to: determine the path information for jumping from the first page image to the second page image, the path information including the interactive elements used during the jump; monitor the path information to determine the correspondence between the first page image, the second page image and the interactive elements.

[0026] According to one embodiment of the present disclosure, the third determination module is also used to: predict the interaction actions of the target application by the target large model to obtain the interaction prediction result of the target application; determine the target application of the previous call based on the execution timing, and obtain the historical interaction process of the target application of the previous call; determine the prompt information of the target application based on the interaction prediction result and the historical interaction process.

[0027] According to one embodiment of the present disclosure, the third determination module is also used to: for any page image between the first page image and the second page image, traverse the interactive elements in the XML description, and mask any page image based on the interactive elements to obtain the masked page image.

[0028] According to one embodiment of the present disclosure, the third determination module is also used for: the target large model parses the XML description according to the prompt information to determine the interaction logic of the target application; based on the prompt information, the mask page image is parsed to determine the page layout information of the target application, the page layout information includes the location information of the interactive elements; based on the interaction logic and the location information, the candidate interaction process corresponding to the target application is determined; the candidate interaction process of each target application in the execution sequence is obtained, and the target interaction process between the target applications is determined according to the candidate interaction process.

[0029] According to one embodiment of the present disclosure, the first determination module is further used to: perform intent recognition on the input information, determine invalid information in the input information, and filter the invalid information to obtain valid information in the input information; perform task recognition on the valid information to determine the task content and task type of the task to be processed.

[0030] According to one embodiment of the present disclosure, the third determination module is also used to: determine multiple sample applications corresponding to the sample interaction process, and a sample candidate interaction process corresponding to each sample application; based on the sample candidate interaction process, train the initial large model until the training is completed to obtain the target large model.

[0031] According to one embodiment of the present disclosure, the third determination module is further used to: obtain a first sample page image and a second sample page image of a sample application corresponding to the sample candidate interaction process; obtain a sample interaction element of the sample application based on the first sample page image and the second sample page image, and determine a sample XML description of the sample application based on the sample interaction element; for any sample page image of the first sample page image and the second sample page image, traverse the sample interaction elements in the sample XML description, and mask any sample page image based on the sample interaction elements to obtain a sample masked page image; determine the sample interaction process based on the sample XML description and the sample masked page image.

[0032] According to one embodiment of the present disclosure, the third determination module is also used to: execute the sample candidate interaction process on the sample application, and encapsulate the interaction process to obtain a sample encapsulation result; input the sample encapsulation result into the initial large model, and the initial large model traverses the sample encapsulation result to obtain operation information of each sample candidate interaction process; generate training prompt information based on the operation information, and train the initial large model based on the training prompt information to obtain a training result; determine whether to continue training the initial large model according to the training result until the training result meets the training end condition to obtain the target large model.

[0033] To achieve the above-mentioned purpose, the third aspect embodiment of the present disclosure proposes an electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to implement the task processing method based on large model intelligent agent as described in the first aspect embodiment of the present disclosure.

[0034] To achieve the above-mentioned purpose, the fourth aspect embodiment of the present disclosure proposes a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to implement the task processing method based on large model intelligent agent as described in the first aspect embodiment of the present disclosure.

[0035] To achieve the above-mentioned purpose, the fifth aspect of the present disclosure proposes a computer program product, including a computer program, which, when executed by a processor, is used to implement the task processing method based on a large model intelligent agent as described in the first aspect of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the present disclosure.

[0037] Figure 1 A flowchart of a task processing method based on a large model intelligent agent provided in an embodiment of the present disclosure;

[0038] Figure 2 An example flow chart of a process for obtaining an XML description of a target application in a task processing method based on a large model intelligent agent provided in an embodiment of the present disclosure;

[0039] Figure 3 A flowchart of another task processing method based on a large model intelligent agent provided in an embodiment of the present disclosure;

[0040] Figure 4An example flow chart of a training process of a target large model in a task processing method based on a large model intelligent agent provided in an embodiment of the present disclosure;

[0041] Figure 5 An example diagram of a process for calling an application according to input information provided in an embodiment of the present disclosure;

[0042] Figure 6 A structural example diagram of a task processing device based on a large model intelligent agent provided in an embodiment of the present disclosure;

[0043] Figure 7 This is a diagram showing an example structure of an electronic device provided in an embodiment of the present disclosure.

[0044] The above drawings show clear embodiments of the present disclosure, which will be described in more detail below. These drawings and text descriptions are not intended to limit the scope of the present disclosure in any way, but to illustrate the concepts of the present disclosure to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0045] In order to enable ordinary persons in the art to better understand the technical solutions of the present disclosure, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the accompanying drawings.

[0046] It should be noted that the terms "first", "second", etc. in the specification and claims of the present disclosure 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 disclosure described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of devices and methods consistent with some aspects of the present disclosure as detailed in the appended claims.

[0047] It should be noted that the acquisition, storage, use, and processing of data in the technical solution of this application are in compliance with the provisions of relevant laws and regulations.

[0048] It should be noted that the information (including but not limited to user device information, user personal information, etc.), data (including but not limited to data used for analysis, stored data, displayed data, etc.) and signals involved in this application are all authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards.

[0049] It should be noted that in the embodiments of the present application, certain software, components, models and other existing solutions in the industry may be mentioned, and they should be regarded as exemplary. Their purpose is only to illustrate the feasibility of implementing the technical solution of the present application, but it does not mean that the applicant has or will necessarily use the relevant content of the solution.

[0050] Figure 1 This is a flowchart of a task processing method based on a large model intelligent agent provided in an embodiment of the present disclosure. Figure 1 As shown, the task processing method based on the large model intelligent agent includes:

[0051] S101, obtaining user input information, and determining the task content and task type of the task to be processed based on the input information.

[0052] It should be noted that the execution subject of the task processing method based on the large model intelligent agent provided in the embodiment of the present disclosure is an electronic device, which can be a terminal device. Optionally, the terminal device can be a mobile electronic device or a non-mobile electronic device. Exemplarily, the mobile electronic device can be a mobile phone, a tablet computer, a laptop computer, a PDA, an in-vehicle electronic device, a wearable device, an ultra-mobile personal computer (ultra-mobile personal computer, UMPC), a netbook or a personal digital assistant (personal digital assistant, PDA), etc., and the non-mobile electronic device can be a network attached storage (Network Attached Storage, NAS), a personal computer (personal computer, PC), a television (television, TV), a teller machine or a self-service machine, etc. The embodiment of the present disclosure is not specifically limited.

[0053] In some implementations, the information input by the user at a set position of the electronic device can be used as the user's input information. Optionally, the user can input voice information, text information, etc. The embodiment of the present disclosure does not specifically limit the form of the user's input information.

[0054] Optionally, when the user's input information is voice information, the voice information can be converted into text information, and the text information is used as the input information; when the user's input information is text information, the input text information is the input information.

[0055] In some implementations, the task information contained in the input information may be identified to determine the task content and task type of the task to be processed. Optionally, the input information may be identified based on keywords. For example, let the keywords be: perform XX operation on XX, open XX, etc., and the task content may be determined by identifying whether the input information contains the above keywords.

[0056] Optionally, if the input information contains a keyword, part of the information containing the keyword is used as the task content.

[0057] Optionally, the task type includes a single task and a combined task. A single task refers to a task content that includes one task indication information; a combined task refers to a task content that includes multiple task indication information. Optionally, the task content may be identified to determine one or more task indication information in the task content.

[0058] For example, if the task content is: send an email to A, then the task instruction information is sending an email, and the task type is determined to be a single-type task; if the task content is: send an email to A and send a message to B, then the task instruction information is sending an email and sending a message, and the task type is determined to be a combined task.

[0059] S102, the target large model determines one or more candidate applications associated with the task content and the execution sequence of the candidate applications according to the task content and the task type.

[0060] In some implementations, the target big model can determine whether the candidate application associated with the task content is a single candidate application or multiple candidate applications based on the task type, and perform semantic recognition on the task content to determine one or more candidate applications associated with the task content and the execution timing of the candidate applications.

[0061] It can be understood that if the task type is a combined task, it is determined that the task content is associated with multiple candidate applications. If the task content is: send an email to A and send a message to B, multiple candidate applications are identified as email applications and SMS applications, and the execution sequence is to use the email application first and then the SMS application.

[0062] S103, calling the candidate applications according to the execution sequence, constructing the prompt information and extensible markup language XML description of the target application for the called target application, and inputting the prompt information, XML description and mask page image of the target application into the target large model, determining the target interaction process between the target applications, and calling the target application to execute the target interaction process.

[0063] In some implementations, based on the execution sequence of the task, the candidate application is called, and the candidate application currently called is determined as the target application. For example, if the execution sequence is to use the mail application first, then the SMS application, and the SMS application is currently called, the SMS application is determined as the target application.

[0064] In some implementations, when determining the target application, the prompt information of the target application can be constructed based on the interaction information of the target application and the interaction information of the historical target application. The interaction information can be the interaction process, and optionally, the target macro model can predict the interaction process of the target application based on the content.

[0065] Optionally, the interaction information of the target application and the interaction information of the historical target application may be combined to obtain the prompt information of the target application.

[0066] In some implementations, interactive elements of the target application may be obtained, and an Extensible Markup Language (XML) description may be determined based on the interactive elements. The interactive elements refer to elements with interactive properties, such as buttons, icons, input boxes, sliders, and other elements in the target application.

[0067] Furthermore, the page image of the target application may be masked according to the interactive element, and a masked page image of the target application may be obtained.

[0068] Furthermore, the prompt information, XML description and mask page image of the target application are input into the target large model, and the target large model parses the input information to obtain the interaction process of the target application. By determining the interaction process of each target application and splicing the interaction process based on the execution time sequence, the target interaction process between the target applications is obtained.

[0069] Furthermore, the target application may be called according to the target interaction process, so that the target application executes the target interaction process, thereby completing the task content.

[0070] S104, in response to the candidate application program ending its call according to the execution sequence, obtaining a task result of the task to be processed.

[0071] In some implementations, after the target large model calls and ends the candidate application according to the execution sequence, the execution process, execution result and other information can be evaluated to obtain the task result of the task to be processed. In other words, the completion quality and efficiency of the task and whether the target large model has completed the task content can be determined based on the task result.

[0072] Optionally, if the task result indicates that the completion quality of the task is low, the target large model can be retrained until the target large model can complete the task content with high quality.

[0073] It is understandable that intelligent agent refers to the combination of big model and intelligent agent technology, and the big model can be used as the core of the intelligent agent to enhance its understanding and processing capabilities of language, text and dialogue. This technology combines the natural language processing capabilities of the big model with the decision-making and action capabilities of the intelligent agent. In the disclosed embodiment, the big model can obtain the target interaction process between applications based on the user's input information, and the intelligent agent can call multiple applications based on the target interaction process to execute the target interaction process.

[0074] In the task processing method based on the large model intelligent agent provided in the embodiment of the present disclosure, the task content and task type of the task to be processed are determined by identifying the user's input information. Then, the candidate application and the execution sequence can be determined according to the task content and task type. For the target application called according to the execution sequence, the prompt information and XML description of the target application can be obtained. By inputting the prompt information, XML description and the mask page image of the target application into the target large model, the target interaction process between the target applications can be obtained, and the target application is called to execute the target interaction process. The large model can analyze the collaborative tasks according to the input information given by the user, and automatically schedule multiple applications to realize collaborative processing, so as to realize more efficient task collaboration, which can further improve the convenience of internal office processing in the enterprise and the user experience of employees, thereby improving work efficiency.

[0075] Based on the above embodiments, the embodiments of the present disclosure can explain the process of obtaining the XML description of the target application, such as Figure 2 As shown, the process of obtaining the XML description of the target application includes:

[0076] S201, determining the task sub-content corresponding to the target application based on the task content.

[0077] In some implementations, by decomposing the task content, multiple independent subtasks can be obtained, each subtask corresponds to a task subcontent, wherein the subtasks can be arranged according to the execution sequence of the candidate applications. That is, by determining the currently scheduled target application, the task subcontent corresponding to the target application can be determined based on the execution sequence.

[0078] S202, obtaining a first page image and a second page image of a target application, wherein the first page image is an image of a start page of the target application, and the second image is an image of an end page after the target application completes execution of a task sub-content.

[0079] Optionally, when opening the target application, you can take a screenshot of the target application to obtain an image of the target application's start page as the first page image, and execute the task sub-content on the target application to obtain an image of the target application's end page after the target application completes executing the task sub-content as the second page image.

[0080] S203: Acquire interactive elements of the target application according to the first page image and the second page image, and determine an XML description of the target application based on the interactive elements.

[0081] In some implementations, the interactive elements of the target application can be determined based on the candidate page elements included in the page image and whether the candidate page elements have interactive properties. In other words, the candidate page elements with interactive properties are used as interactive elements, thereby realizing a more direct way to interact with the page, so that the task content can be completed more conveniently.

[0082] Optionally, the first page image and the second page image may be parsed to determine a first candidate page element corresponding to the first page image and a second candidate page element corresponding to the second page image. Optionally, image recognition may be performed on the first page image and the second page image based on a target detection algorithm to determine candidate page elements in the page images.

[0083] Further, the attribute information of the first candidate page element and the second candidate page element is obtained, and the interactive element in the first candidate page element and the second candidate page element is determined based on the attribute information. Optionally, the attribute information of the first candidate page element and the second candidate page element can be determined by obtaining the identification information of the candidate page element and querying the database storing the attribute information based on the identification information. Further, the candidate page element with interactive properties can be determined as the interactive element based on the attribute information.

[0084] Optionally, in order to facilitate the target large model to call the application and enhance the interoperability between applications, the XML description of the target application may be determined based on the interactive elements.

[0085] Optionally, by determining the temporal relationship between the first page image and the second page image, and determining the correspondence between the first page image, the second page image and the interactive element, the temporal relationship and the correspondence between the first page image and the second page image can be structured to obtain an XML description of the target application.

[0086] In some implementations, in order to intuitively understand the jump logic and interaction mode between pages, the correspondence between the first page image, the second page image and the interactive element can be determined according to the interactive element that jumps from the first page image to the second page image.

[0087] Optionally, by determining path information for jumping from the first page image to the second page image, the path information includes interactive elements used during the jump, and then monitoring the path information to determine the corresponding relationship between the first page image, the second page image and the interactive elements.

[0088] In the task processing method based on the large model intelligent agent provided in the embodiment of the present disclosure, by acquiring the first page image and the second page image of the target application and determining the interactive elements from the page images, further based on the correspondence between the interactive elements and the page images, an XML description of the target application can be obtained, so that the target large model can call the application based on the XML description, thereby improving the convenience of the call and enhancing the interoperability between applications.

[0089] Figure 3 This is a flowchart of a task processing method based on a large model intelligent agent provided in an embodiment of the present disclosure. Figure 3 As shown, the task processing method based on the large model intelligent agent includes:

[0090] S301, obtaining user input information, and determining the task content and task type of the task to be processed based on the input information.

[0091] In some implementations, the input information can be used to identify the intent, determine the invalid information in the input information, and filter the invalid information to obtain the valid information in the input information. Then, the task content and task type can be identified from the valid information. In other words, the task identification can be performed on the valid information to determine the task content and task type of the task to be processed. By clarifying the task content and task type, the processing efficiency of the large model processing task can be improved and the accuracy of task processing can be enhanced.

[0092] S302: The target large model determines one or more candidate applications associated with the task content and the execution sequence of the candidate applications according to the task content and the task type.

[0093] S303, calling the candidate application according to the execution sequence, and constructing prompt information and an extensible markup language XML description of the called target application.

[0094] The relevant contents of step S302 to step S303 can be found in the above embodiment and will not be repeated here.

[0095] S304, the target large model parses the XML description according to the prompt information to determine the interaction logic of the target application.

[0096] In some implementations, the target large model can predict the interaction action and generate prompt information for the currently called target application based on the historical interaction actions of the target application. This can obtain clear and accurate prompt information and improve the accuracy of calling the application for interaction.

[0097] That is, the target large model can predict the interaction actions of the target application to obtain the interaction prediction result of the target application. Furthermore, the target application that was previously called can be determined based on the execution sequence, and the historical interaction process of the target application that was previously called can be obtained to determine the prompt information of the target application based on the interaction prediction result and the historical interaction process.

[0098] In some implementations, after the prompt information of the target large model is determined, the target large model can parse the XML description according to the prompt information to obtain the interaction logic of the target application contained in the XML description. Optionally, the target large model can obtain the interaction prediction result and the historical interaction process of the target application from the prompt information, and then parse the XML description according to the interaction prediction result and the historical interaction process to obtain the interaction logic of the target application.

[0099] S305 , parsing the mask page image based on the prompt information to determine page layout information of the target application, where the page layout information includes location information of interactive elements.

[0100] In some implementations, for any page image of the first page image and the second page image, the interactive elements in the XML description are traversed, and based on the interactive elements, any page image is masked to obtain a masked page image. Optionally, the interactive elements can be numbered during the traversal process, and based on the numbering information of the interactive elements, the mask of the interactive elements is added to any page image to obtain a masked page image. Optionally, based on the masked page image, the area that needs to be processed or focused on in the page image can be accurately defined and controlled. Such as buttons, links or other interactive elements, to improve image processing efficiency.

[0101] Furthermore, the target large model can clarify information such as interactive elements used in the interaction process based on the prompt information, and parse the mask page image to determine the position information of the interactive elements in the mask page image, and use the position information as the page layout information of the target application.

[0102] S306: Determine a candidate interaction process corresponding to the target application based on the interaction logic and the location information.

[0103] In some implementations, the path information after the interaction is triggered can be determined based on the interaction logic, and the location that triggers the interaction can be determined based on the location information. The path information after the interaction is triggered and the location that triggers the interaction can then be used as candidate interaction processes corresponding to the target application.

[0104] S307, obtaining candidate interaction processes of each target application in the execution sequence, and determining target interaction processes between target applications according to the candidate interaction processes.

[0105] In some implementations, the above steps S303-S306 may be repeatedly executed to obtain a candidate interaction process of each target application in the execution sequence, and each candidate interaction process may be combined according to the execution sequence to obtain a target interaction process.

[0106] S308, calling the target application to execute the target interaction process.

[0107] S309, in response to the candidate application program ending its call according to the execution sequence, obtaining a task result of the task to be processed.

[0108] The relevant contents of step S308 to step S309 can be found in the above embodiment and will not be repeated here.

[0109] In the task processing method based on the big model intelligent agent provided in the embodiment of the present disclosure, the task content and task type of the task to be processed are determined by identifying the input information of the user. Then, the candidate application and the execution sequence can be determined according to the task content and task type. For the target application called according to the execution sequence, the prompt information and XML description of the target application can be obtained. The target big model can parse the XML description and the mask page image according to the prompt information to determine the target interaction process between the target applications, and further call the target application to execute the target interaction process. The big model can analyze the collaborative tasks according to the input information given by the user, and automatically schedule multiple applications to realize collaborative processing, so as to realize more efficient task collaboration, and further improve the convenience of internal office processing in the enterprise and the user experience of employees, thereby improving work efficiency.

[0110] Based on the above embodiments, the present disclosure can explain the training process of the target large model, such as Figure 4 As shown, the training process of the target large model includes:

[0111] S401, determining a plurality of sample application programs corresponding to a sample interaction process, and a sample candidate interaction process corresponding to each sample application program.

[0112] In some implementations, in order to enable the big model to learn the real interaction process and enhance the big model's understanding of the real scene, sample interaction processes can be obtained and the big model can be trained based on the sample interaction processes to obtain a target big model that predicts the target interaction process with high accuracy.

[0113] In some implementations, the sample candidate interaction process can be processed by obtaining a first sample page image and a second sample page image of a sample application corresponding to the sample candidate interaction process, and obtaining sample interaction elements of the sample application based on the first sample page image and the second sample page image, and determining a sample XML description of the sample application based on the sample interaction elements.

[0114] Furthermore, for any sample page image of the first sample page image and the second sample page image, the sample interaction elements in the sample XML description are traversed, and based on the sample interaction elements, any sample page image is masked to obtain a sample masked page image.

[0115] Optionally, after obtaining the sample XML description and the sample mask page image, the sample interaction process can be determined based on the sample XML description and the sample mask page image. That is, the sample interaction logic can be parsed from the XML description, and the location of the sample interaction element can be parsed from the sample mask page image, and then the sample interaction process can be determined based on the sample interaction logic and the location of the sample interaction element.

[0116] In some implementations, the sample interaction process may be parsed to determine multiple sample applications used in the sample interaction process and sample candidate interaction processes corresponding to each sample application.

[0117] S402, based on the sample candidate interaction process, the initial large model is trained until the training is completed to obtain the target large model.

[0118] Optionally, the sample candidate interaction process is executed on the sample application, and the interaction process is encapsulated to obtain a sample encapsulation result, which is then input into the initial large model. The initial large model traverses the sample encapsulation result to obtain the operation information of each sample candidate interaction process.

[0119] Further, training prompt information can be generated based on the operation information, and the initial large model can be trained based on the training prompt information to obtain a training result. Optionally, it can be determined whether to continue training the initial large model based on the training result until the training result meets the training end condition to obtain the target large model.

[0120] Optionally, the training result may include the predicted interaction process obtained by the initial large model according to the training prompt information, and it can be judged whether the predicted interaction process is consistent with the sample candidate interaction process. If not, the loss function of the model can be determined according to whether the predicted interaction process is consistent with the sample candidate interaction process, and the model parameters can be adjusted based on the loss function, and the adjusted initial large model can be used for training until a target large model that can output a predicted interaction process that is consistent with the sample candidate interaction process is obtained.

[0121] In the task processing method based on the large model intelligent agent provided in the embodiment of the present disclosure, by obtaining the sample interaction process and training the initial large model based on the sample interaction process to obtain the target large model, the model is trained according to the sample interaction process, which can accelerate the model training and optimization process and improve the adaptability and generalization ability of the model, so that the target large model can accurately determine the target interaction process based on the user input information to realize the call of the application in the target interaction process.

[0122] Figure 5 The flowchart shown is a flowchart of calling an application according to input information. By performing intent recognition on the user's input information, the task type and task content of the task to be processed contained in the input information can be determined. The task type includes single-type tasks and combined-type tasks. When the task type is determined to be a single-type task, it can be determined that the target large model only needs to call one candidate application; when the task type is determined to be a combined-type task, it can be determined that the target large model needs to call multiple candidate applications.

[0123] For the candidate application currently called, it is taken as the target application, and the prompt information and XML description of the target application are constructed. By inputting the prompt information, XML description and mask page image of the target application into the target large model, the target interaction process between the target applications can be determined, and then the target application can be called to execute the target interaction process.

[0124] Corresponding to the task processing methods based on large model intelligent agents provided in the above-mentioned embodiments, an embodiment of the present disclosure also provides a task processing device based on large model intelligent agents. Since the task processing device based on large model intelligent agents provided in the embodiment of the present disclosure corresponds to the task processing methods based on large model intelligent agents provided in the above-mentioned embodiments, the implementation methods of the above-mentioned task processing methods based on large model intelligent agents are also applicable to the task processing device based on large model intelligent agents provided in the embodiment of the present disclosure, and will not be described in detail in the following embodiments.

[0125] Figure 6 This is a diagram showing an example of the structure of a task processing device based on a large model intelligent agent provided in an embodiment of the present disclosure. Figure 6 As shown, the task processing device based on large model intelligent agent includes: a first determination module 601, a second determination module 602, a third determination module 603 and a fourth determination module 604.

[0126] The first determination module 601 is used to obtain user input information and determine the task content and task type of the task to be processed based on the input information;

[0127] A second determination module 602 is used to determine, by the target large model, one or more candidate applications associated with the task content and the execution sequence of the candidate applications according to the task content and the task type;

[0128] The third determination module 603 is used to call the candidate application according to the execution sequence, construct prompt information and an extensible markup language XML description of the called target application, input the prompt information, XML description and the mask page image of the target application into the target large model, determine the target interaction process between the target applications, and call the target application to execute the target interaction process;

[0129] The fourth determining module 604 is configured to obtain the task result of the task to be processed in response to the candidate application program being called to end according to the execution sequence.

[0130] According to an embodiment of the present disclosure, it is characterized in that the third determination module 603 is also used to: determine the task sub-content corresponding to the target application based on the task content; obtain a first page image and a second page image of the target application, wherein the first page image is an image of a start page of the target application, and the second image is an image of an end page after the target application executes the task sub-content; obtain interactive elements of the target application based on the first page image and the second page image, and determine an XML description of the target application based on the interactive elements.

[0131] According to one embodiment of the present disclosure, the third determination module 603 is also used to: parse the first page image and the second page image to determine a first candidate page element corresponding to the first page image and a second candidate page element corresponding to the second page image; obtain attribute information of the first candidate page element and the second candidate page element, and determine the interactive elements in the first candidate page element and the second candidate page element based on the attribute information.

[0132] According to one embodiment of the present disclosure, the third determination module 603 is also used to: determine the temporal relationship between the first page image and the second page image; determine the correspondence between the first page image, the second page image and the interactive element; and perform structured processing on the temporal relationship between the first page image and the second page image and the correspondence to obtain an XML description of the target application.

[0133] According to one embodiment of the present disclosure, the third determination module 603 is also used to: determine the path information for jumping from the first page image to the second page image, the path information including the interactive elements used during the jump; monitor the path information to determine the correspondence between the first page image, the second page image and the interactive elements.

[0134] According to one embodiment of the present disclosure, the third determination module 603 is also used to: predict the interaction actions of the target application by the target large model to obtain the interaction prediction result of the target application; determine the target application of the previous call based on the execution timing, and obtain the historical interaction process of the target application of the previous call; determine the prompt information of the target application based on the interaction prediction result and the historical interaction process.

[0135] According to one embodiment of the present disclosure, the third determination module 603 is also used to: for any page image between the first page image and the second page image, traverse the interactive elements in the XML description, and mask any page image based on the interactive elements to obtain the masked page image.

[0136] According to one embodiment of the present disclosure, the third determination module 603 is also used for: the target large model parses the XML description according to the prompt information to determine the interaction logic of the target application; based on the prompt information, the mask page image is parsed to determine the page layout information of the target application, the page layout information includes the location information of the interactive elements; based on the interaction logic and the location information, the candidate interaction process corresponding to the target application is determined; the candidate interaction process of each target application in the execution sequence is obtained, and the target interaction process between the target applications is determined according to the candidate interaction process.

[0137] According to one embodiment of the present disclosure, the first determination module 601 is further used to: perform intent recognition on the input information, determine invalid information in the input information, and filter the invalid information to obtain valid information in the input information; perform task recognition on the valid information to determine the task content and task type of the task to be processed.

[0138] According to one embodiment of the present disclosure, the third determination module 603 is also used to: determine multiple sample applications corresponding to the sample interaction process, and a sample candidate interaction process corresponding to each sample application; based on the sample candidate interaction process, train the initial large model until the training is completed to obtain the target large model.

[0139] According to one embodiment of the present disclosure, the third determination module 603 is further used to: obtain a first sample page image and a second sample page image of a sample application corresponding to the sample candidate interaction process; obtain a sample interaction element of the sample application based on the first sample page image and the second sample page image, and determine a sample XML description of the sample application based on the sample interaction element; for any sample page image in the first sample page image and the second sample page image, traverse the sample interaction elements in the sample XML description, and mask any sample page image based on the sample interaction elements to obtain a sample masked page image; determine the sample interaction process based on the sample XML description and the sample masked page image.

[0140] According to one embodiment of the present disclosure, the third determination module 603 is also used to: execute the sample candidate interaction process on the sample application, and encapsulate the interaction process to obtain a sample encapsulation result; input the sample encapsulation result into the initial large model, and the initial large model traverses the sample encapsulation result to obtain operation information of each sample candidate interaction process; generate training prompt information based on the operation information, and train the initial large model based on the training prompt information to obtain a training result; determine whether to continue training the initial large model according to the training result until the training result meets the training end condition to obtain the target large model.

[0141] In the task processing device based on the large model intelligent agent provided by the embodiment of the present disclosure, the task content and task type of the task to be processed are determined by identifying the input information of the user. Then, the candidate application and the execution sequence can be determined according to the task content and task type. For the target application called according to the execution sequence, the prompt information and XML description of the target application can be obtained. By inputting the prompt information, XML description and the mask page image of the target application into the target large model, the target interaction process between the target applications can be obtained, and the target application is called to execute the target interaction process. The large model can analyze the collaborative tasks according to the input information given by the user, and automatically schedule multiple applications to realize collaborative processing, so as to realize more efficient task collaboration, and further improve the convenience of internal office processing in the enterprise and the user experience of employees, thereby improving work efficiency.

[0142] Figure 7 This is a diagram showing an example of the structure of an electronic device provided in an embodiment of the present disclosure. Figure 7 As shown, the electronic device 700 may include: a transceiver 701 , a processor 702 , and a memory 703 .

[0143] The processor 702 executes the computer execution instructions stored in the memory, so that the processor 702 executes the scheme in the above embodiment. The processor 702 can be a general-purpose processor, including a central processing unit CPU, a network processor (network processor, NP), etc.; it can also be a digital signal processor DSP, an application-specific integrated circuit ASIC, a field programmable gate array FPGA or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components.

[0144] The memory 703 is connected to the processor 702 via a system bus and completes communication between them. The memory 703 is used to store computer program instructions.

[0145] The transceiver 701 may be used to obtain tasks to be executed and configuration information of the tasks to be executed.

[0146] The system bus can be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The system bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus. The transceiver is used to realize the communication between the database access device and other computers (such as clients, read-write libraries, and read-only libraries). The memory may include random access memory (RAM) and may also include non-volatile memory.

[0147] The electronic device provided in the embodiments of the present disclosure may be the terminal device of the above embodiments.

[0148] The disclosed embodiment also provides a chip for running instructions, which is used to execute the technical solution of the task processing method based on large model intelligent agent in the above embodiment.

[0149] The embodiment of the present disclosure also provides a computer-readable storage medium, which stores computer instructions. When the computer instructions are executed on a computer, the computer executes the technical solution of the task processing method based on the large model intelligent agent in the above embodiment.

[0150] The embodiments of the present disclosure also provide a computer program product, which includes a computer program stored in a computer-readable storage medium. At least one processor can read the computer program from the computer-readable storage medium. When at least one processor executes the computer program, it can implement the technical solution of the task processing method based on the large model intelligent agent in the above-mentioned embodiment.

[0151] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the specification and practicing the invention disclosed herein. The present disclosure is intended to cover any variations, uses or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary techniques in the art that are not disclosed in the present disclosure. The description and examples are to be considered exemplary only, and the true scope and spirit of the present disclosure are indicated by the following claims.

[0152] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. A task processing method based on a large model intelligent agent, characterized in that: include: Obtaining user input information, and determining the task content and task type of the task to be processed based on the input information; The target big model determines one or more candidate applications associated with the task content and the execution timing of the candidate applications according to the task content and the task type; Calling the candidate application according to the execution sequence, constructing prompt information and an extensible markup language XML description of the called target application, inputting the prompt information, the XML description and the mask page image of the target application into the target macro model, determining a target interaction process between the target applications, and calling the target application to execute the target interaction process; In response to the candidate application program ending its call according to the scheduling sequence, a task result of the task to be processed is obtained.

2. The method according to claim 1, characterized in that The process of obtaining the XML description of the target application includes: Based on the task content, determining the task sub-content corresponding to the target application; Acquire a first page image and a second page image of the target application, wherein the first page image is an image of a start page of the target application, and the second image is an image of an end page after the target application executes the task sub-content; According to the first page image and the second page image, the interactive elements of the target application are acquired, and the XML description of the target application is determined based on the interactive elements.

3. The method according to claim 2, characterized in that The acquiring the interactive element of the target application according to the first page image and the second page image includes: Parsing the first page image and the second page image to determine a first candidate page element corresponding to the first page image and a second candidate page element corresponding to the second page image; Attribute information of the first candidate page element and the second candidate page element is acquired, and based on the attribute information, an interactive element among the first candidate page element and the second candidate page element is determined.

4. The method according to claim 3, characterized in that The step of determining the XML description of the target application based on the interactive element comprises: determining a timing relationship between a first page image and a second page image; Determining a correspondence between the first page image, the second page image, and the interactive element; The temporal relationship and the corresponding relationship between the first page image and the second page image are structured to obtain an XML description of the target application.

5. The method according to claim 4, characterized in that The determining the correspondence between the first page image, the second page image, and the interactive element includes: Determine path information for jumping from the first page image to the second page image, wherein the path information includes interactive elements used when jumping; The path information is monitored to determine a corresponding relationship between the first page image, the second page image, and the interactive element.

6. The method according to claim 1, characterized in that The process of obtaining the prompt information of the target application includes: The target large model predicts the interaction action of the target application to obtain the interaction prediction result of the target application; Based on the execution sequence, determining the target application that was previously called, and obtaining the historical interaction process of the target application that was previously called; Based on the interaction prediction result and the historical interaction process, prompt information of the target application is determined.

7. The method according to claim 2, characterized in that The mask page image acquisition process of the target application comprises: For any page image of the first page image and the second page image, the interactive elements in the XML description are traversed, and based on the interactive elements, the any page image is masked to obtain the masked page image.

8. The method according to claim 1, characterized in that The step of inputting the prompt information, the XML description and the mask page image of the target application into the target macro model and determining the target interaction process between the target applications comprises: The target big model parses the XML description according to the prompt information to determine the interaction logic of the target application; Based on the prompt information, the mask page image is parsed to determine page layout information of the target application, wherein the page layout information includes location information of interactive elements; Determining a candidate interaction process corresponding to the target application based on the interaction logic and the location information; A candidate interaction process of each target application in the execution sequence is obtained, and a target interaction process between the target applications is determined according to the candidate interaction process.

9. The method according to claim 1, characterized in that: The determining the task content and task type of the task to be processed based on the input information includes: Performing intent recognition on the input information, determining invalid information in the input information, and filtering the invalid information to obtain valid information in the input information; Perform task identification on the valid information to determine the task content and task type of the task to be processed.

10. The method according to claim 1, characterized in that The training process of the target large model includes: Determine a plurality of sample applications corresponding to the sample interaction process, and a sample candidate interaction process corresponding to each sample application; Based on the sample candidate interaction process, the initial large model is trained until the training is completed to obtain the target large model.

11. The method according to claim 10, characterized in that The sample interaction process is obtained, including: Acquire a first sample page image and a second sample page image of a sample application corresponding to the sample candidate interaction process; Acquire a sample interactive element of the sample application according to the first sample page image and the second sample page image, and determine a sample XML description of the sample application based on the sample interactive element; For any sample page image of the first sample page image and the second sample page image, traverse the sample interaction elements in the sample XML description, and mask the any sample page image based on the sample interaction elements to obtain a sample masked page image; The sample interaction process is determined based on the sample XML description and the sample mask page image.

12. The method according to claim 10, characterized in that The initial large model is trained based on the sample candidate interaction process until the training is completed to obtain the target large model, including: Executing the sample candidate interaction process on the sample application, and encapsulating the interaction process to obtain a sample encapsulation result; Input the sample encapsulation results into the initial large model, and the initial large model traverses the sample encapsulation results to obtain operation information of each sample candidate interaction process; Generate training prompt information based on the operation information, and train the initial large model based on the training prompt information to obtain a training result; Whether to continue training the initial large model is determined according to the training result, until the training result meets the training end condition, thereby obtaining the target large model.

13. A task processing device based on a large model intelligent agent, characterized in that: include: A first determination module is used to obtain user input information and determine the task content and task type of the task to be processed based on the input information; A second determination module is used to determine, by the target large model, one or more candidate applications associated with the task content and the execution timing of the candidate applications according to the task content and the task type; a third determination module, configured to call the candidate application according to the execution sequence, construct prompt information and an extensible markup language XML description of the called target application, input the prompt information, the XML description and the mask page image of the target application into the target macro model, determine a target interaction process between the target applications, and call the target application to execute the target interaction process; The fourth determining module is used to obtain the task result of the task to be processed in response to the candidate application being called to end according to the execution sequence.

14. The device according to claim 13, characterized in that The third determining module is further used for: Based on the task content, determining the task sub-content corresponding to the target application; Acquire a first page image and a second page image of the target application, wherein the first page image is an image of a start page of the target application, and the second image is an image of an end page after the target application executes the task sub-content; According to the first page image and the second page image, the interactive elements of the target application are acquired, and the XML description of the target application is determined based on the interactive elements.

15. The device according to claim 14, characterized in that The third determining module is further used for: Parsing the first page image and the second page image to determine a first candidate page element corresponding to the first page image and a second candidate page element corresponding to the second page image; Attribute information of the first candidate page element and the second candidate page element is acquired, and based on the attribute information, an interactive element among the first candidate page element and the second candidate page element is determined.

16. The device according to claim 15, characterized in that The third determining module is further used for: determining a timing relationship between a first page image and a second page image; Determining a correspondence between the first page image, the second page image, and the interactive element; The temporal relationship and the corresponding relationship between the first page image and the second page image are structured to obtain an XML description of the target application.

17. The device according to claim 16, characterized in thatThe third determining module is further used for: Determine path information for jumping from the first page image to the second page image, wherein the path information includes interactive elements used when jumping; The path information is monitored to determine a corresponding relationship between the first page image, the second page image, and the interactive element.

18. The device according to claim 13, characterized in that The third determining module is further used for: The target large model predicts the interaction action of the target application to obtain the interaction prediction result of the target application; Based on the execution sequence, determining the target application that was previously called, and obtaining the historical interaction process of the target application that was previously called; Based on the interaction prediction result and the historical interaction process, prompt information of the target application is determined.

19. The device according to claim 14, characterized in that The third determining module is further used for: For any page image of the first page image and the second page image, the interactive elements in the XML description are traversed, and based on the interactive elements, the any page image is masked to obtain the masked page image.

20. The device according to claim 13, characterized in that The third determining module is further used for: The target big model parses the XML description according to the prompt information to determine the interaction logic of the target application; Based on the prompt information, the mask page image is parsed to determine page layout information of the target application, wherein the page layout information includes location information of interactive elements; Determining a candidate interaction process corresponding to the target application based on the interaction logic and the location information; A candidate interaction process of each target application in the execution sequence is obtained, and a target interaction process between the target applications is determined according to the candidate interaction process.

21. The device according to claim 13, characterized in that The first determining module is further used for: Performing intent recognition on the input information, determining invalid information in the input information, and filtering the invalid information to obtain valid information in the input information; Perform task identification on the valid information to determine the task content and task type of the task to be processed.

22. The device according to claim 13, characterized in that The third determining module is further used for: Determine a plurality of sample applications corresponding to the sample interaction process, and a sample candidate interaction process corresponding to each sample application; Based on the sample candidate interaction process, the initial large model is trained until the training is completed to obtain the target large model.

23. The device according to claim 22, characterized in that The third determining module is further used for: Acquire a first sample page image and a second sample page image of a sample application corresponding to the sample candidate interaction process; Acquire a sample interactive element of the sample application according to the first sample page image and the second sample page image, and determine a sample XML description of the sample application based on the sample interactive element; For any sample page image of the first sample page image and the second sample page image, traverse the sample interaction elements in the sample XML description, and mask the any sample page image based on the sample interaction elements to obtain a sample masked page image; The sample interaction process is determined based on the sample XML description and the sample mask page image.

24. The device according to claim 22, characterized in that The third determining module is further used for: Executing the sample candidate interaction process on the sample application, and encapsulating the interaction process to obtain a sample encapsulation result; Input the sample encapsulation results into the initial large model, and the initial large model traverses the sample encapsulation results to obtain operation information of each sample candidate interaction process; Generate training prompt information based on the operation information, and train the initial large model based on the training prompt information to obtain a training result; Whether to continue training the initial large model is determined according to the training result, until the training result meets the training end condition, thereby obtaining the target large model.

25. An electronic device, characterized in that: include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory to implement the method according to any one of claims 1 to 12.

26. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are executed by a processor, they are used to implement the method according to any one of claims 1 to 12.

27. A computer program product, characterized in that The invention comprises a computer program, which implements the method according to any one of claims 1 to 12 when being executed by a processor.

Images