Hybrid interaction method and device based on professional subject tasks and electronic device

Abstract

The invention relates to a mixed interaction method and device based on professional subject tasks and an electronic device, and is applied to the technical field of human-computer interaction. The method comprises the steps of calling a professional subject structured input template in response to a trigger operation of a user; receiving unstructured language input of a user and structured parameter input of the user in the professional subject structured input template; according to a preset verification rule associated with the professional subject structured input template, verifying the structured parameter input to obtain a verification result; and generating a professional subject task processing result according to the received unstructured language input and structured parameter input on the basis of the large language model under the condition that the verification result represents that the structured parameter input passes the verification. Through the application, the problem that accurate and efficient professional subject task processing is difficult to realize through an interaction mode of inputting a natural language is solved.

Description

Technical Field

[0001] This application relates to the field of human-computer interaction technology, and in particular to a hybrid interaction method, apparatus, and electronic device based on a professional subject task. Background Technology

[0002] With the widespread application of large language models, the mainstream approach for human-computer task interaction on task processing platforms is based on natural language dialog boxes. Users describe current task requirements, provide parameters, and obtain task processing results by entering natural language, such as "prompt words" or "prompt statements," into text boxes. This approach is open and flexible, can meet users' communication habits, and has a low learning cost.

[0003] However, interactive tasks for specialized disciplines typically involve structured tasks with multiple fixed parameters, requiring users to repeatedly input or describe the same parameters in the dialog box each time, which is cumbersome and inefficient. In addition, since specialized discipline tasks are generally quite complex, users may use imprecise terminology and expressions when issuing instructions through natural language, causing task processing platforms based on large language models to be unable to accurately recognize the natural language input by users, resulting in low accuracy of the final task results.

[0004] There is currently no effective solution to the problem that the interaction method of inputting natural language in related technologies makes it difficult to achieve accurate and efficient processing of professional subject tasks. Summary of the Invention

[0005] This embodiment provides a hybrid interaction method, apparatus, and electronic device based on professional subject tasks to solve the problem that it is difficult to achieve accurate and efficient processing of professional subject tasks through interaction methods that rely on inputting natural language in related technologies.

[0006] Firstly, this embodiment provides a hybrid interaction method based on professional subject tasks, applied to a professional subject task processing platform, which includes a large language model; the method includes:

[0007] In response to a user's trigger operation, a structured input template for a specific subject is invoked; the structured input template for a specific subject is obtained by matching it in a preset knowledge base of specific subjects based on the user's trigger operation.

[0008] Receive the user's unstructured language input, as well as the user's structured parameter input in the professional subject structured input template;

[0009] The structured parameter input is validated according to preset validation rules to obtain validation results; the validation rules are associated with the professional subject structured input template.

[0010] If the verification result indicates that the structured parameter input has passed the verification, based on the large language model, a professional subject task processing result is generated according to the received unstructured language input and the structured parameter input.

[0011] In some embodiments, if the verification result indicates that the structured parameter input has passed verification, generating a subject-specific task processing result based on the large language model, according to the received unstructured language input and the structured parameter input, includes:

[0012] The unstructured language input and the structured parameter input are integrated to obtain the integrated professional subject task;

[0013] Based on the large language model, the integrated professional subject tasks are processed to generate professional subject task processing results.

[0014] In some embodiments, the unstructured language input and the structured parameter input are integrated to obtain an integrated subject-specific task, including:

[0015] Subject-specific information is extracted from the unstructured language input to obtain first subject-specific prompt words; subject-specific information is extracted from the structured parameter input to obtain second subject-specific prompt words.

[0016] Match the first subject prompt word and the second subject prompt word to obtain the duplicate prompt words between the first subject prompt word and the second subject prompt word;

[0017] The integrated subject task is obtained by combining the first subject prompt words and the second subject prompt words after removing the duplicate prompt words, or by combining the second subject prompt words and the first subject prompt words after removing the duplicate prompt words.

[0018] In some embodiments, before matching the first subject prompt word and the second subject prompt word to obtain the repeating prompt word between the first subject prompt word and the second subject prompt word, the method further includes:

[0019] Determine whether there is a semantic conflict between the first subject-specific prompt and the second subject-specific prompt;

[0020] If it is determined that there is a semantic conflict between the first subject prompt and the second subject prompt, a first query instruction is sent to the user; the first query instruction is used to query the user for the subject information corresponding to the semantic conflict;

[0021] The system receives the user's first updated language input on the professional subject task processing platform; the first updated language input is the unstructured input obtained by the user after updating the unstructured language input and / or the structured parameter input based on the semantic conflict.

[0022] Based on the first updated language input, the corresponding first subject prompt words and / or second subject prompt words are updated.

[0023] In some embodiments, the process of processing the integrated subject-specific task based on the large language model to generate subject-specific task processing results includes:

[0024] Based on the integrated professional subject tasks, construct professional subject task processing instructions;

[0025] Using the large language model, determine whether there is missing information in the context of the instruction for processing the professional subject task;

[0026] If it is determined that there is no missing information in the context of the subject-specific task processing instruction, the subject-specific task processing result is generated.

[0027] In some embodiments, the method further includes:

[0028] If it is determined that there is missing information in the context of the subject-specific task processing instruction, a second query instruction is sent to the user; the second query instruction is used to query the user for missing subject-specific information in the context.

[0029] The system receives a second updated language input from the user on the professional subject task processing platform. The second updated language input is an unstructured input obtained by the user updating the unstructured language input and / or the structured parameter input based on the context relationship.

[0030] Based on the second updated language input, the integrated professional subject task is updated to obtain the updated professional subject task;

[0031] The updated subject-specific tasks are processed to generate subject-specific task processing results.

[0032] In some embodiments, the step of invoking a subject-specific structured input template in response to a user's triggering action further includes:

[0033] In response to a user's triggering action, receive the user's real-time unstructured language input;

[0034] Determine the subject-specific characteristics in the real-time unstructured language input;

[0035] Based on the characteristics of the professional discipline, a structured input template for the professional discipline is invoked; the structured input template for the professional discipline is obtained by matching the professional discipline characteristics in a preset professional discipline knowledge base.

[0036] Secondly, this embodiment provides a hybrid interactive device based on professional subject tasks, the device comprising: a receiving module, a processing module, and a generating module;

[0037] The processing module is used to respond to a user's trigger operation by calling a professional subject structured input template; the professional subject structured input template is obtained by matching in a preset professional subject knowledge base based on the user's trigger operation; it is also used to verify the structured parameter input according to preset verification rules and obtain verification results; the verification rules are associated with the professional subject structured input template.

[0038] The receiving module is used to receive the user's unstructured language input and the user's structured parameter input in the professional subject structured input template;

[0039] The generation module is used to generate a professional subject task processing result based on the large language model, according to the received unstructured language input and the structured parameter input, when the verification result indicates that the structured parameter input has passed the verification.

[0040] Thirdly, this embodiment provides an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the hybrid interaction method based on a subject-matter task as described in the first aspect above.

[0041] Fourthly, this embodiment provides a storage medium storing a computer program that, when executed by a processor, implements the hybrid interactive method based on professional subject tasks described in the first aspect above.

[0042] Compared with related technologies, the hybrid interactive method, apparatus, and electronic device based on professional subject tasks provided in this embodiment invoke a professional subject-specific structured input template through user-triggered operations on the platform, and then receive structured parameter input from the user based on the template. Subsequently, the structured parameter input is validated, improving the accuracy of subsequent input into the large language model. Simultaneously, by combining unstructured language input through a dialog box with the validated structured parameter input, the professional subject task processing result is generated based on the large language model. This combines the flexibility of the dialog box with the accuracy of the templated structured parameter input to achieve accurate and efficient professional subject task processing.

[0043] Details of one or more embodiments of this application are set forth in the following drawings and description to make other features, objects and advantages of this application more readily apparent. Attached Figure Description

[0044] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0045] Figure 1 This is a hardware structure block diagram of the terminal for the hybrid interaction method based on professional subject tasks provided in this embodiment;

[0046] Figure 2 This is a flowchart of a hybrid interaction method based on professional subject tasks provided in an embodiment of this application;

[0047] Figure 3 This is a flowchart illustrating the hybrid interaction method that integrates conversational and structured input provided in this specific embodiment.

[0048] Figure 4 This is a schematic diagram of the interactive task triggering interface provided in this specific embodiment;

[0049] Figure 5 This is a schematic diagram of a specialized subject-based structured input template for tumor CAR-T target mining provided in this specific embodiment;

[0050] Figure 6 This is a schematic diagram of a subject-specific structured input template for Mendelian randomization analysis provided in this specific embodiment;

[0051] Figure 7 This is a schematic diagram of the interactive interface variable input provided in this specific embodiment;

[0052] Figure 8 This is a schematic diagram illustrating the termination of the structured target input task provided in this specific embodiment;

[0053] Figure 9 This is a schematic diagram of the hybrid interactive system and its corresponding execution method provided in this specific embodiment;

[0054] Figure 10 This is a structural block diagram of the hybrid interactive device based on professional subject tasks in this embodiment. Detailed Implementation

[0055] To better understand the purpose, technical solution, and advantages of this application, the application is described and illustrated below in conjunction with the accompanying drawings and embodiments.

[0056] Unless otherwise defined, the technical or scientific terms used in this application shall have the general meaning understood by one of ordinary skill in the art to which this application pertains. Words such as “a,” “an,” “an,” “the,” “the,” and “these” used in this application do not indicate quantitative limitation and may be singular or plural. The terms “comprising,” “including,” “having,” and any variations thereof used in this application are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or device that comprises a series of steps or modules (units) is not limited to the listed steps or modules (units) but may include steps or modules (units) not listed, or may include other steps or modules (units) inherent to these processes, methods, products, or devices. Words such as “connected,” “linked,” and “coupled” used in this application are not limited to physical or mechanical connections but may include electrical connections, whether direct or indirect. “Multiple” used in this application refers to two or more. “And / or” describes the relationship between related objects, indicating that three relationships may exist; for example, “A and / or B” can represent: A alone, A and B simultaneously, and B alone. Normally, the character " / " indicates that the objects before and after it are in an "or" relationship. The terms "first," "second," "third," etc., used in this application are merely to distinguish similar objects and do not represent a specific order of objects.

[0057] The method embodiments provided in this example can be executed on a terminal, computer, or similar computing device. For example, it can run on a terminal. Figure 1 This is a hardware structure block diagram of the terminal for the hybrid interaction method based on professional subject tasks provided in this embodiment. For example... Figure 1 As shown, a terminal may include one or more ( Figure 1Only one is shown in the diagram. A processor 102 and a memory 104 for storing data are also included. The processor 102 may be, but is not limited to, a microprocessor (MCU) or a programmable logic device (FPGA). The terminal may also include a transmission device 106 for communication functions and an input / output device 108. Those skilled in the art will understand that… Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the terminal described above. For example, the terminal may also include components that are larger than... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown are illustrated.

[0058] The memory 104 can be used to store computer programs, such as application software programs and modules, like the computer program corresponding to the hybrid interactive method based on professional subject tasks in this embodiment. The processor 102 executes various functional applications and data processing by running the computer programs stored in the memory 104, thereby implementing the above-described method. The memory 104 may include high-speed random access memory and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor 102, and these remote memories can be connected to the terminal via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0059] The transmission device 106 is used to receive or send data via a network. This network includes a wireless network provided by the terminal's communication provider. In one example, the transmission device 106 includes a Network Interface Controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission device 106 can be a Radio Frequency (RF) module used for wireless communication with the Internet.

[0060] Currently, in human-computer interaction, users typically issue complex commands containing multiple precise parameters to specialized task processing platforms driven by artificial intelligence, such as large language models, in order to obtain task processing results for corresponding specialized disciplines, such as life science computing and data visualization analysis.

[0061] Current human-computer interaction methods primarily rely on users inputting natural language into dialog boxes to describe task requirements, provide parameters, and obtain results. The advantages of this approach are that its interaction format aligns with human communication habits, has a low learning curve, offers the most open-ended results, and allows for the greatest flexibility in input content and format. However, when handling structured or repetitive tasks, this method suffers from low operational efficiency and unstable results, failing to fully meet the needs of specialized fields.

[0062] For example, for structured tasks with multiple fixed parameters, users need to repeatedly input or describe the same parameters in a dialog box each time, which is cumbersome and inefficient. Furthermore, current artificial intelligence's understanding of natural language remains uncertain; slight differences in how users describe their needs or the use of imprecise terminology can lead to misunderstandings by the AI, resulting in incorrect or unsatisfactory output. The stability of the results highly depends on the user's "prompt engineering" ability.

[0063] Furthermore, to obtain accurate output, users need to learn and master how to construct high-quality prompts for specialized disciplines, raising the barrier to entry for specialized task platforms and making them unfriendly to non-specialist users. For tasks requiring the uploading of files such as gene data or tables, or tasks involving multiple selections, the current simple dialog box fails to provide intuitive and convenient operation support. In pure dialog box mode, the system cannot determine the validity of parameters during the user input phase. If the user input parameters are missing, incorrectly formatted, or out of range, the task instruction corresponding to the current prompt must be submitted to the platform's AI backend for calculation or processing before errors can be detected and returned to the user. This process involves a loop of "submit-wait-error return-modify," which not only wastes valuable computing resources but also significantly prolongs the user's waiting time, resulting in a poor user experience and high correction costs.

[0064] For unstructured input, users can directly input pure natural language text to describe the task content and send it to the AI. However, each user's description is unique, and the AI ​​needs to extract and understand the user's task content before executing the specific task. Furthermore, for users who want to use the method but lack knowledge of specialized disciplines such as Mendelian randomization analysis, they cannot accurately describe the required documents and parameters, leading to inefficiency and high computational resource consumption during AI dialogue. For structured input, the documents and parameters required for Mendelian randomization analysis are organized into easy-to-understand data templates. Users who want to use this analysis method only need to select the task template and interactively set the data and parameters used in the method. This significantly reduces the learning cost and usage threshold for users, while also improving the efficiency and reducing resource consumption in AI dialogue.

[0065] In summary, unstructured input is random; it feeds all user-inputted information to the AI ​​for processing. Structured input, on the other hand, follows rules. For a given method or question format, it can be organized into a template. After rendering, users can either fill in only the variable parts via interactive controls or supplement the template content before sending it to the AI ​​for dialogue. This process lowers the barrier to entry for users and reduces the resource consumption of the AI.

[0066] Therefore, addressing the inefficiencies, unstable results, and high user requirements inherent in existing technologies that rely solely on dialog box interactions, this application introduces a hybrid input mode combining dialog boxes and structured templates. Users can issue commands via natural language while simultaneously triggering operations to invoke a pre-defined structured template containing controls such as input boxes, dropdown menus, and file uploads, enabling precise and efficient definition of task parameters. This method combines the flexibility of natural language with the precision of template-based operations, significantly lowering the user's learning curve and improving the execution efficiency and reliability of complex tasks, making it particularly suitable for scientific computing and data analysis fields requiring precise parameter input.

[0067] Specifically, this embodiment provides a hybrid interaction method based on professional subject tasks, applied to a professional subject task processing platform, which includes a large language model; Figure 2 This is a flowchart of a hybrid interaction method based on professional subject tasks provided in an embodiment of this application, such as... Figure 2 As shown, the process includes the following steps:

[0068] Step S210: In response to the user's trigger operation, the professional subject structured input template is invoked; the professional subject structured input template is obtained by matching in the preset professional subject knowledge base based on the user's trigger operation.

[0069] The professional subject task processing platform provides users with an interactive interface that triggers the invocation of structured input templates for professional subjects. When a user needs to process a professional subject task, they trigger the operation through the scene tool invocation entry point in the interactive interface. At this time, the user clicks the scene tool corresponding to the professional subject task processing scene, and the professional subject task processing platform responds to the user's trigger operation on the interactive interface, presenting the professional subject structured input template that matches the current professional subject task (i.e., the scene task) in the user's current professional subject task processing scene to the interactive interface.

[0070] Furthermore, the entry point for scene tools can be placed inside or near the dialogue area in the interactive interface, appearing as a clickable button, icon, or text link. The current template can contain various forms of UI controls, such as text input boxes for entering numerical values ​​or strings, menus for making single or multiple selections from preset options, and file upload controls for uploading local data files. No restrictions are placed on the UI controls in the professional subject-specific structured input template.

[0071] The professional subject task processing platform stores a preset professional subject knowledge base, which includes different professional subjects, corresponding professional subject structured input templates, and mixed interactive input data formats. This allows users to determine the corresponding structured input template for different professional subject content, thereby improving user convenience.

[0072] As another implementation method, a professional subject-specific structured input template can be invoked in real time based on the user's current real-time unstructured language input. Specifically, in response to the user's trigger operation, the user's real-time unstructured language input is received; the professional subject-specific features in the real-time unstructured language input are determined; and the professional subject-specific structured input template is invoked based on the professional subject-specific features, which are matched against a preset professional subject-specific knowledge base. The professional subject-specific features include specific professional terms or specific professional parameters in the current professional subject, etc., which are not specifically limited here.

[0073] Step S220: Receive unstructured language input from the user, as well as structured parameter input from the user in the structured input template for the professional subject.

[0074] Among them, the professional subject task processing platform also provides users with an interactive interface that includes an input interface for inputting natural language; based on the professional subject task that needs to be solved, users can input unstructured language in the input interface, such as a dialog box.

[0075] Meanwhile, in the professional subject task processing platform, after presenting the professional subject structured input template matching the current user's scenario task to the interactive interface, the user will input the corresponding structured parameters in the professional subject structured input template based on the current professional subject task.

[0076] Step S230: According to the preset verification rules, the structured parameter input is verified to obtain the verification result; the verification rules are associated with the professional subject structured input template.

[0077] When a user completes unstructured language input and structured parameter input in the interactive interface, the structured parameters need to be validated according to preset validation rules. The preset validation rules are associated with the professional subject structured input templates and include the rules corresponding to the structured parameter input in each professional subject structured input template, such as the range of structured parameters and the format of structured parameters, etc., which are not specifically limited here.

[0078] The structured parameter input is validated according to the preset validation rules. If the validation result indicates that the structured parameter has failed the validation, the user needs to be prompted with an error message in the interactive interface to guide the user to correct the structured parameter input. At the same time, the subsequent integration steps of unstructured language input and structured parameter input are not performed to avoid processing the erroneous unstructured language input and structured parameter input, thus wasting computing resources.

[0079] Step S240: If the verification result indicates that the structured parameter input has passed the verification, based on the large language model, the professional subject task processing result is generated according to the received unstructured language input and structured parameter input.

[0080] Specifically, if the verification result indicates that the structured parameter input has passed the verification, the received unstructured language input and structured parameter input are integrated, and the integrated unstructured language input and structured parameter input are converted into instructions and sent to the large language model. The large language model executes the current instructions and generates the corresponding professional subject task processing results.

[0081] Through the above steps, user-triggered operations on the platform invoke a specialized subject-specific structured input template, thereby receiving structured parameter input from the user based on this template. Subsequently, the structured parameter input is validated, improving the accuracy of subsequent input into the large language model. Simultaneously, by combining unstructured language input via a dialog box with validated structured parameter input, specialized subject-specific task processing results are generated based on the large language model. This combines the flexibility of the dialog box with the precision of the templated structured parameter input to achieve accurate and efficient specialized subject-specific task processing.

[0082] In some embodiments, if the validation result indicates that the structured parameter input passes validation, a subject-specific task processing result is generated based on the large language model, according to the received unstructured language input and structured parameter input, including:

[0083] Unstructured language input and structured parameter input are integrated to obtain an integrated subject-specific task. Based on a large language model, the integrated subject-specific task is processed to generate the subject-specific task processing result.

[0084] In order to further improve the completeness of the professional subject task and the accuracy of subsequent task processing, after receiving the user's unstructured language input and structured parameter input, it is necessary to integrate the currently determined unstructured language input and the calibrated structured parameter input to eliminate duplicates, semantic conflicts, or missing input content. Then, by processing the integrated professional subject task, the completeness of the professional subject task processing is improved, thereby improving the accuracy of generating professional subject task processing results.

[0085] Furthermore, the method for integrating unstructured language input and structured parameter input includes: extracting subject-specific information from the unstructured language input to obtain first-subject-specific prompts; extracting subject-specific information from the structured parameter input to obtain second-subject-specific prompts; matching the first-subject-specific prompts and the second-subject-specific prompts to obtain duplicate prompts between them; integrating the first-subject-specific prompts and the second-subject-specific prompts after removing duplicate prompts, or integrating the second-subject-specific prompts and the first-subject-specific prompts after removing duplicate prompts, to obtain the integrated subject-specific task.

[0086] When integrating unstructured language input and structured parameter input, it is necessary to first extract subject-specific information from both. Further, for extracting subject-specific information from structured parameter input, it is necessary to first combine the structured parameter input with its corresponding preset subject-specific structured input template to form structured input content. Then, subject-specific information, including the subject-specific content and its corresponding parameters, is extracted from the structured input content to obtain the second subject-specific prompt words.

[0087] Because users may input duplicate information when performing language and parameter input for specialized subject tasks, this increases the amount of duplicate content in the integrated specialized subject task, wasting the computing resources of the specialized subject task processing platform and affecting its efficiency. Therefore, it is necessary to first eliminate duplicate content (repeated prompts) between the first subject prompts corresponding to the user's unstructured language input and the second subject prompts corresponding to the structured language input, resulting in a specialized subject task that can form an executable final task instruction.

[0088] In the process of matching the first and second subject-specific prompts, multiple features within the corresponding subject are compared sequentially. After identifying duplicate prompts, these duplicate prompts are removed from either the first or second subject-specific prompts, ensuring that the integrated subject-specific task weight contains only a single duplicate prompt. This further reduces the amount of data processed and improves the efficiency of subsequent task processing.

[0089] Before matching the first subject prompt and the second subject prompt to obtain duplicate prompts between them, it is necessary to determine whether there is a semantic conflict between the first subject prompt and the second subject prompt. If a semantic conflict is found between the first subject prompt and the second subject prompt, a first query instruction is sent to the user. The first query instruction is used to query the user for the subject information corresponding to the semantic conflict. The system receives the user's first updated language input on the subject task processing platform. The first updated language input is the unstructured input obtained by the user after updating the unstructured language input and / or structured parameter input based on the semantic conflict. The corresponding first subject prompt and / or second subject prompt are updated based on the first updated language input.

[0090] When integrating unstructured language input and structured parameter input, in addition to duplicate input, semantic conflicts may also occur. For example, if the first subject prompt contains a parameter value of 0.1 for feature A, while the second subject prompt, based on the parameter value of feature B and the relationship between feature B and feature A, determines the parameter value of feature A to be 0.01, then there is a conflict between the parameter value of feature B in the second subject prompt and the parameter value of feature A in the first prompt. The semantic conflict here is not limited to the parameter value of the feature; it can also be related to the corresponding string content, etc., which is not specifically limited here.

[0091] When a conflict is detected between the structured parameter input in the professional subject structured template and the parameter input in the unstructured natural language, the professional subject task processing platform will pause the execution of the corresponding professional subject task and proactively initiate a first inquiry command to the user in the dialog box, asking and guiding the user to make a final confirmation of the parameter value with semantic conflict, so as to ensure the accuracy of the command execution.

[0092] In some embodiments, the integrated subject-matter tasks are processed based on a large language model to generate subject-matter task processing results, including: constructing subject-matter task processing instructions based on the integrated subject-matter tasks; using the large language model to determine whether there is missing information in the context of the subject-matter task processing instructions; and generating subject-matter task processing results if there is no missing information in the context of the subject-matter task processing instructions.

[0093] If it is determined that there is missing information in the context of the subject-specific task processing instruction, a second query instruction is sent to the user. The second query instruction is used to query the user for the missing subject-specific information in the context. The system receives the user's second updated language input on the subject-specific task processing platform. The second updated language input is the unstructured input obtained by the user updating the unstructured language input and / or structured parameter input based on the context. Based on the second updated language input, the integrated subject-specific task is updated to obtain the updated subject-specific task. The updated subject-specific task is processed to generate the subject-specific task processing result.

[0094] Specifically, after receiving unstructured language input and structured parameter input from the user in the professional subject task processing platform, and after verifying and integrating them, the system constructs the professional subject task processing instruction corresponding to the integrated professional subject task, and sends it to the large language model in the professional subject task processing platform, so that the large language model can generate the corresponding professional subject task processing result based on the professional subject task processing instruction.

[0095] However, before receiving the subject-specific task processing instructions, the large language model in the subject-specific task processing platform only performs formal verification and processing on the task, but does not verify the accuracy of its specific content. Therefore, it is necessary to further determine whether there is any missing information in the context of the subject-specific task processing instructions through the large language model. Missing information here includes situations such as missing subject-specific information in the context or logical errors in the context.

[0096] When it is determined that there is missing information in the context of the subject-specific task processing instruction, a second query instruction needs to be sent to the user. This query is then sent to the user in the dialog interface of the interactive interface to ask for the missing subject-specific information in the context. The system also receives the second update language input (i.e., unstructured language input) sent by the user. Based on the second update language input, the current subject-specific task processing instruction is updated, which further improves the accuracy of the large language model in generating corresponding subject-specific task processing results for the subject-specific task processing instruction.

[0097] Additionally, the interactive interface provides a way to disable the structured input template for professional subject tasks. In response to the user's disable action, the template is removed from the interactive interface, allowing the user to continue interacting by only inputting unstructured language through the dialogue area.

[0098] The present embodiment will be described and explained below through specific examples.

[0099] This specific embodiment provides a hybrid interaction method that integrates conversational and structured input. Figure 3 This is a flowchart illustrating the hybrid interaction method that integrates conversational and structured input provided in this specific embodiment. (Reference) Figure 3 The method includes steps S310 to S340.

[0100] Step S310: Interactive task triggered.

[0101] Specifically, the professional subject task processing platform provides an interactive interface that includes a dialogue area and entry points for scenario tools. When a user generates a specific task or requirement, the user clicks the corresponding scenario tool, which triggers the template input mode; the professional subject task processing platform then calls up a professional subject structured input template based on the specific content in the scenario tool clicked by the user.

[0102] Figure 4 This is a schematic diagram of the interactive task triggering interface provided in this specific embodiment. (Reference) Figure 4 The interactive interface of the professional subject task processing platform includes a dialog box with a prompt: "Describe your needs, such as conducting a literature review on aging-related biomarkers." Additionally, there are "Network Search" and "Personal Knowledge Base" trigger buttons. When users need to perform a professional subject task, they can click on "Personal Knowledge Base" to obtain corresponding "Tumor CAR-T Target Mining" or "Mendelian Randomization Analysis" option buttons, allowing users to perform the corresponding trigger operation based on their current professional subject task processing needs.

[0103] Step S320: The hybrid interactive interface is presented.

[0104] Specifically, the professional subject task processing platform retains the dialog box while responding to user-triggered operations on the interface, presenting a structured input template that matches the scenario task. This template may include one or more of the following UI controls: text input boxes, drop-down / multi-select menus, and file upload controls, etc., without specific limitations.

[0105] The text input box is used to input parameters such as numerical values ​​and strings, for example, the p-value and kb in "Mendelian randomization analysis". The drop-down / multi-select menu is used to select one or more preset options, for example, the "screening dimensions" in "tumor CAR-T target mining". The file upload control is used to upload local files as input data, such as "exposure data" and "outcome data" involved in professional subject tasks.

[0106] Figure 5 This is a schematic diagram of a specialized subject-specific structured input template for tumor CAR-T target mining provided in this specific embodiment. (Refer to...) Figure 5 The structured input template for this major suggests that it can currently provide "a one-stop customized CAR-T target mining solution based on scRNA-seq, covering the entire process of cell type annotation, malignant cell sorting and other services". It also prompts users to upload single cell files, enter the target cell type, and select dimensions such as "cell surface genes, T cell immune compatibility, off-target toxicity avoidance, FDA safety assessment" from the drop-down menu.

[0107] After users upload / input relevant content according to the prompts of the professional subject structured input template, the professional subject task processing platform analyzes the single-cell file uploaded by the user and performs target mining on the target cell type entered by the user. It also uses cell surface genes, T cell immune compatibility, off-target toxicity avoidance, and FDA safety assessment as screening dimensions to achieve tumor CAR-T target mining.

[0108] Figure 6 This is a schematic diagram of a subject-specific structured input template for Mendelian randomization analysis provided in this specific embodiment. (Reference) Figure 6 The structured input template for this major offers the ability to "use the random assignment characteristics of genetic variations, such as single nucleotide polymorphisms (SNPs), as instrumental variables (IVs) to simulate randomized controlled trials and infer the causal association between exposure factors (such as biomarkers and environmental factors) and disease outcomes." It also prompts users to upload exposure data, outcome data, and confounding factor files, and provides parameter settings for each, including p-value, kb, r², and sample size. The default values ​​for p-value and kb are set to 5e-08, 10000, and 0.001.

[0109] like Figure 5 and Figure 6 As shown, for the "Tumor CAR-T Target Mining" task, the system provides a template with multiple checkboxes including "Click to Upload," "Please Enter," and "Filter Dimensions." For "Mendelian Randomization Analysis," multiple file upload controls and parameter input boxes are provided. This design clearly decomposes the user's intent into structured data fields, greatly improving the accuracy of the interaction.

[0110] Step S330: User input and real-time verification.

[0111] Specifically, users operate these Figure 5 and Figure 6The system utilizes UI controls to precisely set task parameters, while users can still input supplementary information, special instructions, or fine-tune the task in the dialog box. Furthermore, it can perform real-time validation of the input values ​​in the structured template before the user submits the task, based on preset validation rules. These preset validation rules can be customized according to the user's specific professional subject task requirements and professional knowledge.

[0112] Figure 7 This is a schematic diagram of the interactive interface variable input provided in this specific embodiment. (Reference) Figure 7 The variables include input, selection, and upload categories. The input category's UI is set as follows: hint: hint; value: 10000; activation: |10000; when empty, the prompt displays "None"; error message: Please enter. The selection category's UI is set as follows: hint: hint and its dropdown menu; value: value A, value B and their dropdown menus; activation: value A, value B, and their dropdown menus including surface genes, T cell low-expression genes, off-target genes, and FDA drug genes; when empty, the prompt displays "None"; error message: Please select and its dropdown menu. The upload category's UI is set as follows: hint: to be uploaded; value: file name, format; activation: clicking reopens the file selector; error message: Please upload the file.

[0113] When the structured parameters entered by the user pass validation, the user can proceed to the next step. If validation fails, for example, if the user-entered value 'p' is not a valid number, or if a required file is missing, the system will issue an error message and guide the user to correct it in real time. Real-time error message means that the system immediately provides a clear error message on the interactive interface without waiting for a response from the AI ​​backend. Guided correction means that the message clearly informs the user of the reason for the error and how to correct it, guiding the user to directly modify the settings on the current interface.

[0114] Step S340: Instruction execution and result return.

[0115] Specifically, after the user clicks the "OK" or "Send" button on the interactive interface, the platform integrates the validated structured parameters with the natural language instructions in the dialog box to form a complete and accurate task request. This request is then executed by the backend AI, and the execution result is returned to the user. If a conflict is detected between the template input and the natural language input parameters, the task execution will be paused, and the user will be prompted in the dialog box to confirm the conflicting parameter values, ensuring the accuracy of the instruction execution.

[0116] If a user wishes to exit mixed input during a conversation, i.e., to use only natural language input, they can simply turn off the Scene Tools tab in the dialog box. Figure 8This is a schematic diagram illustrating the completion of the structured target input task according to this specific embodiment. (Reference) Figure 8 After the user uploads / inputs relevant content according to the prompts in the professional subject-specific structured input template, the professional subject-specific task processing platform analyzes the uploaded single-cell file and performs target mining on the target cell type entered by the user. It uses cell surface genes, T-cell immune compatibility, off-target toxicity avoidance, and FDA safety assessment as screening dimensions to achieve tumor CAR-T target mining. If the user then initiates a new conversation and sends the instruction "prompt text formed by variable blocks, displaying 'None' when the variable is empty," the dialog box in the professional subject-specific task processing platform completes its processing and displays the "conversation results and output files." If the user wishes to exit mixed input at this point, they can click the "×" in "Tumor CAR-T Target Mining ×" to exit the structured parameter input.

[0117] To achieve the aforementioned hybrid interaction method that integrates conversational and structured input, a hybrid interaction system integrating conversational and structured input is set up in the professional subject task processing platform of this application, wherein, Figure 9 This is a schematic diagram of the hybrid interactive system and its corresponding execution method provided in this specific embodiment. (Reference) Figure 9 The hybrid interactive system includes: a data preprocessing module, a template triggering module, a data construction module, a data input and verification module, and a data output module.

[0118] The data preprocessing module is used to process the pre-configured prompt template for the scene task into a specific hybrid interactive input data format. The scene task here refers to the subject-specific task in the aforementioned embodiments. The prompt template refers to the subject-specific structured input template in the aforementioned embodiments.

[0119] The mixed interactive input data format consists of id (scene task ID), name (scene task name), and data (scene task data) attributes. The prompt template is a string configured for each scene tool in natural language. The locations where interactive controls need to be rendered are replaced with corresponding variable names, such as "malignant cell tag: #{malignantCellTag}", where malignantCellTag is the file variable that the user needs to upload, ultimately rendered as a file upload control. When using it, the prompt template string and the corresponding control configurations for the variables in the string need to be input into this module. This module segments the prompt template according to the input variable positions to obtain a mixed interactive input data array. Each object in the array can contain the following: name (control name attribute, corresponding to the variable name in the prompt template), type (control type), defaultValue (control default value), placeholder (control placeholder hint), rules (control validation rules), and other configurable control attributes (e.g., when the control is a selector, multiple and options can be configured). In the prompt template, the text before and after the interactive input variable is converted into elements of type "text" in an array. Each object in the array has properties that are configurable for the rendered control: for input controls, select controls, and file upload controls, common properties include type (input, select, file, text), name (variable name), default value, placeholder hints, and validation rules; properties specific to select controls include whether multiple selections are allowed and the number of options. It's important to note that the text data before and after the variable in the template is of type "text".

[0120] The template trigger module determines whether to trigger the hybrid interactive input mode. This module judges whether to trigger the hybrid interactive input mode based on the user's behavior in the natural language input interface. Initially, the user input interface is the natural language input interface. When the user selects a scene task by choosing a component, the interface switches to the hybrid interactive interface. In the hybrid interactive interface, clicking the close scene task tab returns to the default input interface.

[0121] The data construction module is used to render and display the scene task data. This module integrates the template data input from the hybrid interactive interface; specifically, its function is to render each element in the scene task template array obtained from the data preprocessing module as a corresponding control and construct a form display. Specifically, it iterates through the `data` array to render each element, and then uses the entire rendering result as a form control. <form>The child elements are displayed in the input interface. Rendering each element of the array involves the following steps:

[0122] Step 1: Select the input control corresponding to the current element based on the type attribute, such as text ( ), input ( <input> ), select( <select> ), file (< / select> <input type="file"> ).

[0123] Step two: Assign values ​​to other control properties contained in the element, such as placeholder, defaultValue, options, multiple, etc.

[0124] Step 3: If the element contains rules for validation, add it to the validation rules of the parent component's form.

[0125] For example, in the subject-specific structured objective corresponding to the subject-specific task of "Mendelian randomization analysis," the input variable r2 is required, can only contain numbers, and is specified to have a range of [0,1]. Therefore, the rules for verifying r2 are set as follows:

[0126] rules = [

[0127] {required: true, trigger; 'blur'},

[0128] {validator: checkR2, trigger; 'blur'}

[0129] ].

[0130] Add the above validation rules to the master rules file.

[0131] The data input and validation module is used for user input and data validation. In the mixed interactive input interface, users can input / modify variable values ​​through interactive controls, and also modify text. Users can delete interactive controls or text just like they would a regular text input area, and can also undo deletions. In the validation module, if parameter validation rules exist, form parameter validation is performed according to the rules when changes to control content or loss of focus are detected. If parameter validation fails, the rendering style on the interactive interface is immediately changed to an error message to remind the user to re-enter the data, allowing the user to restart the interactive input process based on the error message.

[0132] The data output module is used to construct a prompt text instruction based on the input content of the controls. In the hybrid interactive input interface, after the user completes the input, the structured parameter input, i.e., the current value of each interactive control, is filled back into the prompt template. By combining the structured parameter input and the user's natural language input, the final prompt text instruction sent to the large language model, i.e., AI, is constructed, and multiple rounds of dialogue are conducted based on the returned results.

[0133] This module constructs the output of the data input and validation module into natural language text. For output using natural language input, the content itself is the instruction sent to the AI ​​and does not require further construction. For output data input through mixed interactions, the current values ​​of each interactive control need to be filled back into the prompt template to obtain the final prompt text instruction sent to the AI. The AI ​​determines whether there are missing parameters or conflicts that could lead to ambiguity in the instruction information based on the received instruction and the context. If ambiguity exists, the AI ​​returns a message asking the user for a question, and the user sends the relevant information via natural language input before the AI ​​makes its judgment. If there is no ambiguity, the AI ​​directly returns the result.

[0134] Therefore, this specific embodiment leverages the complementary advantages of both structured and unstructured input methods: "dialogue" handles exploratory and ambiguous initial task concepts, while "templates" handle repetitive and structured parameter input. Structured template input incorporates form concepts, allowing for the setting of filling rules for variables within the template. Real-time feedback is provided to the user during interaction to check for errors in the entered information, enabling more precise definition of variable parameters within the template. By moving the rule-based information verification and validation to the front-end input interface, the traditional passive "submit-wait-error" interaction mode is transformed into a proactive "instant feedback, pre-emptive prevention" interaction mode, significantly improving the user experience. Instant verification reduces user difficulty, enhances the user experience, and increases the efficiency and reliability of communication with AI.

[0135] It should be noted that the steps shown in the above process or in the flowcharts in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions.

[0136] This embodiment also provides a hybrid interactive device based on a professional subject task. This device is used to implement the above embodiments and preferred embodiments, and will not be repeated as already described. The terms "module," "unit," "subunit," etc., used below can refer to a combination of software and / or hardware that performs a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0137] Figure 10 This is a structural block diagram of the hybrid interactive device based on professional subject tasks in this embodiment, as shown below. Figure 10 As shown, the device includes a receiving module 10, a processing module 20, and a generating module 30.

[0138] The processing module 20 is used to respond to the user's trigger operation and call the professional subject structured input template; the professional subject structured input template is obtained by matching in the preset professional subject knowledge base based on the user's trigger operation; it is also used to verify the structured parameter input according to the preset verification rules and obtain the verification result; the verification rules are associated with the professional subject structured input template.

[0139] The receiving module 10 is used to receive the user's unstructured language input, as well as the user's structured parameter input in the professional subject structured input template.

[0140] The generation module 30 is used to generate the professional subject task processing results based on the large language model, according to the received unstructured language input and structured parameter input, when the verification result characterizes the structured parameter input as passing the verification.

[0141] It should be noted that the above modules can be functional modules or program modules, and can be implemented through software or hardware. For modules implemented through hardware, the above modules can reside in the same processor; or the above modules can be located in different processors in any combination.

[0142] This embodiment also provides an electronic device including a memory and a processor, the memory storing a computer program and the processor being configured to run the computer program to perform the steps in any of the above method embodiments.

[0143] Optionally, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor and the input / output device is connected to the processor.

[0144] Optionally, in this embodiment, the processor can be configured to perform the following steps via a computer program:

[0145] S1, in response to the user's trigger operation, calls the professional subject structured input template; the professional subject structured input template is obtained by matching in the preset professional subject knowledge base based on the user's trigger operation.

[0146] S2 receives unstructured language input from the user, as well as structured parameter input from the user in a structured input template for a specific subject.

[0147] S3 verifies the structured parameter input according to the preset verification rules and obtains the verification results; the verification rules are associated with the professional subject structured input template.

[0148] S4, if the verification result indicates that the structured parameter input has passed the verification, based on the large language model, the professional subject task processing result is generated according to the received unstructured language input and structured parameter input.

[0149] It should be noted that the specific examples in this embodiment can refer to the examples described in the above embodiments and optional implementations, and will not be repeated in this embodiment.

[0150] Furthermore, in conjunction with the hybrid interaction method based on subject-specific tasks provided in the above embodiments, this embodiment can also provide a storage medium for implementation. This storage medium stores a computer program; when executed by a processor, the computer program implements any of the hybrid interaction methods based on subject-specific tasks in the above embodiments.

[0151] It should be understood that the specific embodiments described herein are merely illustrative of the application and not intended to limit it. All other embodiments derived by those skilled in the art based on the embodiments provided in this application without inventive effort are within the scope of protection of this application.

[0152] Obviously, the accompanying drawings are merely some examples or embodiments of this application. Those skilled in the art can apply this application to other similar situations based on these drawings without any creative effort. Furthermore, it is understood that although the work done in this development process may be complex and lengthy, for those skilled in the art, certain design, manufacturing, or production modifications made based on the technical content disclosed in this application are merely conventional technical means and should not be considered as insufficient disclosure of this application.

[0153] The term "embodiment" in this application refers to a specific feature, structure, or characteristic described in connection with an embodiment that may be included in at least one embodiment of this application. The appearance of this phrase in various places in the specification does not necessarily imply the same embodiment, nor does it imply that it is mutually exclusive with or independent of other embodiments. It will be clearly or implicitly understood by those skilled in the art that the embodiments described in this application may be combined with other embodiments without conflict.

[0154] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of patent protection. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the appended claims. < / form>

Claims

1. A hybrid interactive method based on professional subject tasks, characterized in that, The method is applied to a professional subject task processing platform, which includes a large language model; the method includes: In response to a user's trigger operation, a structured input template for a specific subject is invoked; the structured input template for a specific subject is obtained by matching it in a preset knowledge base of specific subjects based on the user's trigger operation. Receive the user's unstructured language input, as well as the user's structured parameter input in the professional subject structured input template; The structured parameter input is validated according to preset validation rules to obtain validation results; the validation rules are associated with the professional subject structured input template. If the verification result indicates that the structured parameter input has passed the verification, based on the large language model, a professional subject task processing result is generated according to the received unstructured language input and the structured parameter input.

2. The hybrid interaction method based on professional subject tasks according to claim 1, characterized in that, If the verification result indicates that the structured parameter input has passed verification, based on the large language model, and according to the received unstructured language input and the structured parameter input, a subject-specific task processing result is generated, including: The unstructured language input and the structured parameter input are integrated to obtain the integrated professional subject task; Based on the large language model, the integrated professional subject tasks are processed to generate professional subject task processing results.

3. The hybrid interaction method based on professional subject tasks according to claim 2, characterized in that, The unstructured language input and the structured parameter input are integrated to obtain the integrated subject-specific task, including: Subject-specific information is extracted from the unstructured language input to obtain first subject-specific prompt words; subject-specific information is extracted from the structured parameter input to obtain second subject-specific prompt words. Match the first subject prompt word and the second subject prompt word to obtain the duplicate prompt words between the first subject prompt word and the second subject prompt word; The integrated subject task is obtained by combining the first subject prompt words and the second subject prompt words after removing the duplicate prompt words, or by combining the second subject prompt words and the first subject prompt words after removing the duplicate prompt words.

4. The hybrid interaction method based on professional subject tasks according to claim 3, characterized in that, Before matching the first subject prompt word and the second subject prompt word to obtain the duplicate prompt words between the first subject prompt word and the second subject prompt word, the method further includes: Determine whether there is a semantic conflict between the first subject-specific prompt and the second subject-specific prompt; If it is determined that there is a semantic conflict between the first subject prompt and the second subject prompt, a first query instruction is sent to the user; the first query instruction is used to query the user for the subject information corresponding to the semantic conflict; The system receives the user's first updated language input on the professional subject task processing platform; the first updated language input is the unstructured input obtained by the user after updating the unstructured language input and / or the structured parameter input based on the semantic conflict. Based on the first updated language input, the corresponding first subject prompt words and / or second subject prompt words are updated.

5. The hybrid interaction method based on professional subject tasks according to claim 2, characterized in that, The process of processing the integrated subject-specific tasks based on the large language model to generate subject-specific task processing results includes: Based on the integrated professional subject tasks, construct professional subject task processing instructions; Using the large language model, determine whether there is missing information in the context of the instruction for processing the professional subject task; If it is determined that there is no missing information in the context of the subject-specific task processing instruction, the subject-specific task processing result is generated.

6. The hybrid interaction method based on professional subject tasks according to claim 5, characterized in that, The method further includes: If it is determined that there is missing information in the context of the subject-specific task processing instruction, a second query instruction is sent to the user; the second query instruction is used to query the user for missing subject-specific information in the context. The system receives a second updated language input from the user on the professional subject task processing platform. The second updated language input is an unstructured input obtained by the user updating the unstructured language input and / or the structured parameter input based on the context relationship. Based on the second updated language input, the integrated professional subject task is updated to obtain the updated professional subject task; The updated subject-specific tasks are processed to generate subject-specific task processing results.

7. The hybrid interactive method based on professional subject tasks according to any one of claims 1 to 6, characterized in that, The method of responding to a user's triggered action by invoking a structured input template for a specific subject also includes: In response to a user's triggering action, receive the user's real-time unstructured language input; Determine the subject-specific characteristics in the real-time unstructured language input; Based on the characteristics of the professional discipline, a structured input template for the professional discipline is invoked; the structured input template for the professional discipline is obtained by matching the professional discipline characteristics in a preset professional discipline knowledge base.

8. A hybrid interactive device based on a professional subject task, characterized in that, The device includes: a receiving module, a processing module, and a generating module; The processing module is used to respond to a user's trigger operation by calling a professional subject structured input template; the professional subject structured input template is obtained by matching in a preset professional subject knowledge base based on the user's trigger operation; it is also used to verify the structured parameter input according to preset verification rules and obtain verification results; the verification rules are associated with the professional subject structured input template. The receiving module is used to receive the user's unstructured language input and the user's structured parameter input in the professional subject structured input template; The generation module is used to generate a professional subject task processing result based on the large language model, according to the received unstructured language input and the structured parameter input, when the verification result indicates that the structured parameter input has passed the verification.

9. An electronic device comprising a memory and a processor, characterized in that, The memory stores a computer program, and the processor is configured to run the computer program to perform the hybrid interactive method based on a subject-matter task as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the hybrid interactive method based on a subject-specific task as described in any one of claims 1 to 7.

Images